Initial commit

.gitattributes

+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bin.* filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zstandard filter=lfs diff=lfs merge=lfs -text
+*.tfevents* filter=lfs diff=lfs merge=lfs -text
+*.db* filter=lfs diff=lfs merge=lfs -text
+*.ark* filter=lfs diff=lfs merge=lfs -text
+**/*ckpt*data* filter=lfs diff=lfs merge=lfs -text
+**/*ckpt*.meta filter=lfs diff=lfs merge=lfs -text
+**/*ckpt*.index filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.gguf* filter=lfs diff=lfs merge=lfs -text
+*.ggml filter=lfs diff=lfs merge=lfs -text
+*.llamafile* filter=lfs diff=lfs merge=lfs -text
+model-00001-of-00008.safetensors filter=lfs diff=lfs merge=lfs -text
+model-00002-of-00008.safetensors filter=lfs diff=lfs merge=lfs -text
+model-00003-of-00008.safetensors filter=lfs diff=lfs merge=lfs -text
+model-00004-of-00008.safetensors filter=lfs diff=lfs merge=lfs -text
+model-00005-of-00008.safetensors filter=lfs diff=lfs merge=lfs -text
+model-00006-of-00008.safetensors filter=lfs diff=lfs merge=lfs -text
+model-00007-of-00008.safetensors filter=lfs diff=lfs merge=lfs -text
+model-00008-of-00008.safetensors filter=lfs diff=lfs merge=lfs -text

LICENSE

+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

README.md

+# JIUTIAN-139MoE-Chat
+
+## 环境配置
+
+### Docker
+
+拉取镜像，启动并进入容器
+
+```
+docker pull image.sourcefind.cn:5000/dcu/admin/base/pytorch:2.1.0-ubuntu20.04-dtk24.04.1-py3.10
+docker run -it  --shm-size 80g --network=host --name=jiutian --privileged  --device /dev/m--device=/dev/kfd --device=/dev/dri --group-add video --cap-add=SYS_PTRACE --security-opt seccomp=unconfined -v /opt/hyhal/:/opt/hyhal/:ro image.sourcefind.cn:5000/dcu/admin/base/pytorch:2.1.0-ubuntu20.04-dtk24.04.1-py3.10 /bin/bash
+```
+
+下载模型权重：[JIUTIAN-139MoE-Chat · 模型库 (modelscope.cn)](https://www.modelscope.cn/models/jiutian-ai/jiutian-139moe-chat/files)
+
+```
+#克隆项目
+git clone http://developer.hpccube.com/codes/modelzoo/jiutian-139moe-chat.git
+cd jiutian-139moe-chat
+```
+
+### 模型推理
+
+```
+python inference.py
+```
+
+### FastAPI调用模型
+
+```
+python app.py
+```
+
+测试容器内部是否能够正常调用：
+
+另起一个终端，输入
+
+```
+curl -X POST "http://localhost:8000/predict/" -H "Content-Type: application/json" -d '{"text": "Please introduce the Great Wall."}'
+```
+
+成功调用会在终端生成内容：
+
+```
+{"response":"Human:\nPlease introduce the Great Wall.\n\nAssistant:\n The Great Wall of China is a series of fortifications built along the northern borders of China to protect against invasions and raids from various nomadic groups. It is one of the most famous landmarks in China and is also one of the largest construction projects in human history.\n\nThe Great Wall stretches"}
+```
+

app.py

+from fastapi import FastAPI, Request
+from pydantic import BaseModel
+from transformers import AutoTokenizer, AutoModelForCausalLM
+import torch
+
+app = FastAPI()
+
+# 加载模型和分词器
+model_id = "/workspace/jiutian/JIUTIAN-139MoE-chat"  # 修改为你的模型路径
+tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
+model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", torch_dtype=torch.bfloat16,
+                                             trust_remote_code=True)
+
+
+# 定义请求体
+class ModelInput(BaseModel):
+    text: str
+
+
+@app.post("/predict/")
+async def predict(request: Request, model_input: ModelInput):
+    # 处理输入文本
+    text = "Human:\n" + model_input.text + "\n\nAssistant:\n"
+
+    # 分词和生成输出
+    inputs = tokenizer(text, return_tensors="pt", add_special_tokens=False, padding_side='left', truncation_side='left')
+    outputs = model.generate(**inputs, max_new_tokens=64, repetition_penalty=1.03, do_sample=False, eos_token_id=0)
+
+    # 解码输出
+    response_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+    return {"response": response_text}
+
+
+if __name__ == "__main__":
+    import uvicorn
+
+    uvicorn.run(app, host="0.0.0.0", port=8000)
\ No newline at end of file

config.json

+{
+  "_name_or_path": "/apps/sharedstorage/oceanstor-a800/sft/deepspeed_chat_moe/Pretrained_model/final_with_router/",
+  "activation": "silu",
+  "architectures": [
+    "JiutianForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "auto_map": {
+    "AutoConfig": "configuration_jiutian.JiutianConfig",
+    "AutoModelForCausalLM": "modeling_jiutian.JiutianForCausalLM"
+  },
+  "bos_token_id": 1,
+  "eos_token_id": 0,
+  "hidden_act": "silu",
+  "hidden_size": 5120,
+  "initializer_range": 0.02,
+  "intermediate_size": 13824,
+  "max_position_embeddings": 4096,
+  "model_type": "jiutian",
+  "num_attention_heads": 40,
+  "num_experts_per_tok": 2,
+  "num_hidden_layers": 40,
+  "num_key_value_heads": 40,
+  "num_local_experts": 8,
+  "output_router_logits": true,
+  "rms_norm_eps": 1e-08,
+  "rope_theta": 10000.0,
+  "router_aux_loss_coef": 0.01,
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.36.2",
+  "use_cache": true,
+  "use_cope": false,
+  "vocab_size": 69120
+}

configuration_jiutian.py

+# coding=utf-8
+# Copyright 2023 Mixtral AI and the HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" Jiutian-MoE model configuration"""
+
+from transformers.configuration_utils import PretrainedConfig
+from transformers.utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+
+class JiutianConfig(PretrainedConfig):
+
+    model_type = "jiutian"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    def __init__(
+        self,
+        vocab_size=69120,
+        hidden_size=5120,
+        intermediate_size=13824,
+        num_hidden_layers=40,
+        num_attention_heads=40,
+        num_key_value_heads=40,
+        hidden_act="silu",
+        max_position_embeddings=4096,
+        initializer_range=0.02,
+        rms_norm_eps=1e-8,
+        use_cache=True,
+        pad_token_id=None,
+        bos_token_id=1,
+        eos_token_id=0,
+        tie_word_embeddings=False,
+        rope_theta=1e4,
+        sliding_window=None,
+        attention_dropout=0.0,
+        num_experts_per_tok=2,
+        num_local_experts=8,
+        output_router_logits=False,
+        router_aux_loss_coef=0.01,
+        use_cope=False,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.sliding_window = sliding_window
+
+        # for backward compatibility
+        if num_key_value_heads is None:
+            num_key_value_heads = num_attention_heads
+
+        self.num_key_value_heads = num_key_value_heads
+        self.hidden_act = hidden_act
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.attention_dropout = attention_dropout
+
+        self.num_experts_per_tok = num_experts_per_tok
+        self.num_local_experts = num_local_experts
+        self.output_router_logits = output_router_logits
+        self.router_aux_loss_coef = router_aux_loss_coef
+        self.use_cope = use_cope
+        super().__init__(
+            pad_token_id=pad_token_id,
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )

generation_config.json

+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 0,
+  "transformers_version": "4.36.2"
+}

inference.py

+from transformers import AutoModelForCausalLM, AutoTokenizer
+import torch
+model_id = "/workspace/JIUTIAN-139MoE-chat"
+tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
+model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", torch_dtype=torch.bfloat16, trust_remote_code=True)
+text = "Please introduce the Great Wall."
+text = "Human:\n" + text + "\n\nAssistant:\n"
+inputs = tokenizer(text, return_tensors="pt", add_special_tokens=False,padding_side='left',truncation_side='left')
+outputs = model.generate(**inputs, max_new_tokens=64, repetition_penalty=1.03,do_sample=False,eos_token_id=0)
+print(tokenizer.decode(outputs[0],skip_special_tokens=True))
\ No newline at end of file

model.safetensors.index.json

+{
+    "metadata": {
+        "total_size": 77756016640
+    },
+    "weight_map": {
+        "lm_head.weight": "model-00008-of-00008.safetensors",
+        "model.embed_tokens.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.0.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.0.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.0.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.2.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.2.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.2.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.3.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.3.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.3.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.4.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.4.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.4.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.5.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.5.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.5.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.6.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.6.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.6.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.7.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.7.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.experts.7.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.block_sparse_moe.gate.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.input_layernorm.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.0.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.0.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.0.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.2.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.2.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.2.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.3.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.3.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.3.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.4.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.4.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.4.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.5.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.5.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.5.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.6.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.6.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.6.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.7.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.7.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.experts.7.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.block_sparse_moe.gate.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.input_layernorm.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.0.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.0.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.0.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.2.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.2.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.2.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.3.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.3.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.3.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.4.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.4.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.4.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.5.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.5.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.5.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.6.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.6.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.6.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.7.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.7.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.experts.7.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.block_sparse_moe.gate.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.input_layernorm.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.post_attention_layernorm.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.self_attn.k_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.10.self_attn.o_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.10.self_attn.q_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.10.self_attn.v_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.0.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.0.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.0.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.2.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.2.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.2.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.3.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.3.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.3.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.4.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.4.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.4.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.5.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.5.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.5.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.6.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.6.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.6.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.7.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.7.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.experts.7.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.block_sparse_moe.gate.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.input_layernorm.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.post_attention_layernorm.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.self_attn.k_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.self_attn.o_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.self_attn.q_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.11.self_attn.v_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.0.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.0.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.0.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.2.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.2.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.2.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.3.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.3.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.3.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.4.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.4.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.4.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.5.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.5.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.5.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.6.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.6.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.6.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.7.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.7.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.experts.7.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.block_sparse_moe.gate.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.input_layernorm.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.post_attention_layernorm.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.self_attn.k_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.self_attn.o_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.self_attn.q_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.12.self_attn.v_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.0.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.0.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.0.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.2.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.2.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.2.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.3.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.3.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.3.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.4.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.4.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.4.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.5.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.5.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.5.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.6.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.6.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.6.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.7.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.7.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.experts.7.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.block_sparse_moe.gate.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.input_layernorm.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.post_attention_layernorm.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.self_attn.k_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.self_attn.o_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.self_attn.q_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.13.self_attn.v_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.0.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.0.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.0.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.2.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.2.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.2.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.3.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.3.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.3.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.4.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.4.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.4.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.5.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.5.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.5.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.6.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.6.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.6.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.7.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.7.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.experts.7.w3.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.block_sparse_moe.gate.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.input_layernorm.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.post_attention_layernorm.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.self_attn.k_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.self_attn.o_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.self_attn.q_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.14.self_attn.v_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.0.w1.weight": "model-00003-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.0.w2.weight": "model-00003-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.0.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.2.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.2.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.2.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.3.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.3.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.3.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.4.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.4.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.4.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.5.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.5.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.5.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.6.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.6.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.6.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.7.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.7.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.experts.7.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.block_sparse_moe.gate.weight": "model-00003-of-00008.safetensors",
+        "model.layers.15.input_layernorm.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.post_attention_layernorm.weight": "model-00004-of-00008.safetensors",
+        "model.layers.15.self_attn.k_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.15.self_attn.o_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.15.self_attn.q_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.15.self_attn.v_proj.weight": "model-00003-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.0.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.0.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.0.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.2.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.2.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.2.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.3.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.3.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.3.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.4.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.4.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.4.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.5.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.5.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.5.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.6.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.6.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.6.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.7.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.7.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.experts.7.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.block_sparse_moe.gate.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.input_layernorm.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.post_attention_layernorm.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.self_attn.k_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.self_attn.o_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.self_attn.q_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.16.self_attn.v_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.0.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.0.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.0.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.2.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.2.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.2.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.3.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.3.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.3.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.4.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.4.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.4.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.5.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.5.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.5.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.6.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.6.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.6.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.7.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.7.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.experts.7.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.block_sparse_moe.gate.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.input_layernorm.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.post_attention_layernorm.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.self_attn.k_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.self_attn.o_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.self_attn.q_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.17.self_attn.v_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.0.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.0.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.0.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.2.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.2.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.2.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.3.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.3.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.3.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.4.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.4.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.4.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.5.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.5.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.5.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.6.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.6.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.6.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.7.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.7.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.experts.7.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.block_sparse_moe.gate.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.input_layernorm.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.post_attention_layernorm.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.self_attn.k_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.self_attn.o_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.self_attn.q_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.18.self_attn.v_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.0.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.0.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.0.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.2.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.2.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.2.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.3.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.3.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.3.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.4.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.4.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.4.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.5.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.5.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.5.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.6.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.6.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.6.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.7.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.7.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.experts.7.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.block_sparse_moe.gate.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.input_layernorm.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.post_attention_layernorm.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.self_attn.k_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.self_attn.o_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.self_attn.q_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.19.self_attn.v_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.0.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.0.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.0.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.2.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.2.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.2.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.3.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.3.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.3.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.4.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.4.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.4.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.5.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.5.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.5.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.6.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.6.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.6.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.7.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.7.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.experts.7.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.block_sparse_moe.gate.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.input_layernorm.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.0.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.0.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.0.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.2.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.2.w2.weight": "model-00004-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.2.w3.weight": "model-00004-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.3.w1.weight": "model-00004-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.3.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.3.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.4.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.4.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.4.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.5.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.5.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.5.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.6.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.6.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.6.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.7.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.7.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.experts.7.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.20.block_sparse_moe.gate.weight": "model-00004-of-00008.safetensors",
+        "model.layers.20.input_layernorm.weight": "model-00005-of-00008.safetensors",
+        "model.layers.20.post_attention_layernorm.weight": "model-00005-of-00008.safetensors",
+        "model.layers.20.self_attn.k_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.20.self_attn.o_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.20.self_attn.q_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.20.self_attn.v_proj.weight": "model-00004-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.0.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.0.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.0.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.2.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.2.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.2.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.3.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.3.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.3.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.4.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.4.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.4.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.5.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.5.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.5.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.6.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.6.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.6.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.7.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.7.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.experts.7.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.block_sparse_moe.gate.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.input_layernorm.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.post_attention_layernorm.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.self_attn.k_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.self_attn.o_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.self_attn.q_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.21.self_attn.v_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.0.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.0.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.0.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.2.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.2.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.2.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.3.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.3.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.3.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.4.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.4.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.4.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.5.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.5.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.5.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.6.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.6.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.6.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.7.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.7.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.experts.7.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.block_sparse_moe.gate.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.input_layernorm.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.post_attention_layernorm.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.self_attn.k_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.self_attn.o_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.self_attn.q_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.22.self_attn.v_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.0.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.0.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.0.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.2.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.2.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.2.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.3.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.3.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.3.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.4.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.4.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.4.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.5.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.5.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.5.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.6.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.6.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.6.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.7.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.7.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.experts.7.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.block_sparse_moe.gate.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.input_layernorm.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.post_attention_layernorm.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.self_attn.k_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.self_attn.o_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.self_attn.q_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.23.self_attn.v_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.0.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.0.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.0.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.2.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.2.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.2.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.3.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.3.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.3.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.4.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.4.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.4.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.5.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.5.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.5.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.6.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.6.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.6.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.7.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.7.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.experts.7.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.block_sparse_moe.gate.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.input_layernorm.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.post_attention_layernorm.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.self_attn.k_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.self_attn.o_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.self_attn.q_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.24.self_attn.v_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.0.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.0.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.0.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.2.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.2.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.2.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.3.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.3.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.3.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.4.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.4.w2.weight": "model-00005-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.4.w3.weight": "model-00005-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.5.w1.weight": "model-00005-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.5.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.5.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.6.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.6.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.6.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.7.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.7.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.experts.7.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.25.block_sparse_moe.gate.weight": "model-00005-of-00008.safetensors",
+        "model.layers.25.input_layernorm.weight": "model-00006-of-00008.safetensors",
+        "model.layers.25.post_attention_layernorm.weight": "model-00006-of-00008.safetensors",
+        "model.layers.25.self_attn.k_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.25.self_attn.o_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.25.self_attn.q_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.25.self_attn.v_proj.weight": "model-00005-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.0.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.0.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.0.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.2.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.2.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.2.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.3.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.3.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.3.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.4.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.4.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.4.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.5.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.5.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.5.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.6.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.6.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.6.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.7.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.7.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.experts.7.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.block_sparse_moe.gate.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.input_layernorm.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.post_attention_layernorm.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.self_attn.k_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.self_attn.o_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.self_attn.q_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.26.self_attn.v_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.0.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.0.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.0.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.2.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.2.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.2.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.3.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.3.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.3.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.4.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.4.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.4.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.5.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.5.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.5.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.6.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.6.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.6.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.7.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.7.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.experts.7.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.block_sparse_moe.gate.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.input_layernorm.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.post_attention_layernorm.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.self_attn.k_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.self_attn.o_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.self_attn.q_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.27.self_attn.v_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.0.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.0.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.0.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.2.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.2.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.2.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.3.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.3.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.3.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.4.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.4.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.4.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.5.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.5.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.5.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.6.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.6.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.6.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.7.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.7.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.experts.7.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.block_sparse_moe.gate.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.input_layernorm.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.post_attention_layernorm.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.self_attn.k_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.self_attn.o_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.self_attn.q_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.28.self_attn.v_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.0.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.0.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.0.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.2.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.2.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.2.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.3.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.3.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.3.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.4.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.4.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.4.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.5.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.5.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.5.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.6.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.6.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.6.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.7.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.7.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.experts.7.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.block_sparse_moe.gate.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.input_layernorm.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.post_attention_layernorm.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.self_attn.k_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.self_attn.o_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.self_attn.q_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.29.self_attn.v_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.0.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.0.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.0.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.2.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.2.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.2.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.3.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.3.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.3.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.4.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.4.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.4.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.5.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.5.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.5.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.6.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.6.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.6.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.7.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.7.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.experts.7.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.block_sparse_moe.gate.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.input_layernorm.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.0.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.0.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.0.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.2.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.2.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.2.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.3.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.3.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.3.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.4.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.4.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.4.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.5.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.5.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.5.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.6.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.6.w2.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.6.w3.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.7.w1.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.7.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.experts.7.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.30.block_sparse_moe.gate.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.input_layernorm.weight": "model-00007-of-00008.safetensors",
+        "model.layers.30.post_attention_layernorm.weight": "model-00007-of-00008.safetensors",
+        "model.layers.30.self_attn.k_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.self_attn.o_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.self_attn.q_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.30.self_attn.v_proj.weight": "model-00006-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.0.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.0.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.0.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.2.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.2.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.2.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.3.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.3.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.3.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.4.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.4.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.4.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.5.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.5.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.5.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.6.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.6.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.6.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.7.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.7.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.experts.7.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.block_sparse_moe.gate.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.input_layernorm.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.post_attention_layernorm.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.self_attn.k_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.self_attn.o_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.self_attn.q_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.31.self_attn.v_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.0.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.0.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.0.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.2.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.2.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.2.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.3.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.3.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.3.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.4.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.4.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.4.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.5.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.5.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.5.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.6.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.6.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.6.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.7.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.7.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.experts.7.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.block_sparse_moe.gate.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.input_layernorm.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.post_attention_layernorm.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.self_attn.k_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.self_attn.o_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.self_attn.q_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.32.self_attn.v_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.0.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.0.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.0.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.2.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.2.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.2.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.3.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.3.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.3.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.4.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.4.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.4.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.5.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.5.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.5.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.6.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.6.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.6.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.7.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.7.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.experts.7.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.block_sparse_moe.gate.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.input_layernorm.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.post_attention_layernorm.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.self_attn.k_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.self_attn.o_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.self_attn.q_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.33.self_attn.v_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.0.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.0.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.0.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.2.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.2.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.2.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.3.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.3.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.3.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.4.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.4.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.4.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.5.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.5.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.5.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.6.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.6.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.6.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.7.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.7.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.experts.7.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.block_sparse_moe.gate.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.input_layernorm.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.post_attention_layernorm.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.self_attn.k_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.self_attn.o_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.self_attn.q_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.34.self_attn.v_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.0.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.0.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.0.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.2.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.2.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.2.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.3.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.3.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.3.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.4.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.4.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.4.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.5.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.5.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.5.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.6.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.6.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.6.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.7.w1.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.7.w2.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.experts.7.w3.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.block_sparse_moe.gate.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.input_layernorm.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.post_attention_layernorm.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.self_attn.k_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.self_attn.o_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.self_attn.q_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.35.self_attn.v_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.0.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.0.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.0.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.2.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.2.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.2.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.3.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.3.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.3.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.4.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.4.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.4.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.5.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.5.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.5.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.6.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.6.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.6.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.7.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.7.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.experts.7.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.block_sparse_moe.gate.weight": "model-00007-of-00008.safetensors",
+        "model.layers.36.input_layernorm.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.post_attention_layernorm.weight": "model-00008-of-00008.safetensors",
+        "model.layers.36.self_attn.k_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.36.self_attn.o_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.36.self_attn.q_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.36.self_attn.v_proj.weight": "model-00007-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.0.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.0.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.0.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.2.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.2.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.2.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.3.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.3.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.3.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.4.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.4.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.4.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.5.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.5.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.5.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.6.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.6.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.6.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.7.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.7.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.experts.7.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.block_sparse_moe.gate.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.input_layernorm.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.post_attention_layernorm.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.self_attn.k_proj.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.self_attn.o_proj.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.self_attn.q_proj.weight": "model-00008-of-00008.safetensors",
+        "model.layers.37.self_attn.v_proj.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.0.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.0.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.0.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.2.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.2.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.2.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.3.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.3.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.3.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.4.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.4.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.4.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.5.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.5.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.5.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.6.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.6.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.6.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.7.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.7.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.experts.7.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.block_sparse_moe.gate.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.input_layernorm.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.post_attention_layernorm.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.self_attn.k_proj.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.self_attn.o_proj.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.self_attn.q_proj.weight": "model-00008-of-00008.safetensors",
+        "model.layers.38.self_attn.v_proj.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.0.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.0.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.0.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.2.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.2.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.2.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.3.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.3.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.3.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.4.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.4.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.4.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.5.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.5.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.5.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.6.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.6.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.6.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.7.w1.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.7.w2.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.experts.7.w3.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.block_sparse_moe.gate.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.input_layernorm.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.post_attention_layernorm.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.self_attn.k_proj.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.self_attn.o_proj.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.self_attn.q_proj.weight": "model-00008-of-00008.safetensors",
+        "model.layers.39.self_attn.v_proj.weight": "model-00008-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.0.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.0.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.0.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.2.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.2.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.2.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.3.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.3.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.3.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.4.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.4.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.4.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.5.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.5.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.5.w3.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.6.w1.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.6.w2.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.6.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.7.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.7.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.experts.7.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.4.block_sparse_moe.gate.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.input_layernorm.weight": "model-00002-of-00008.safetensors",
+        "model.layers.4.post_attention_layernorm.weight": "model-00002-of-00008.safetensors",
+        "model.layers.4.self_attn.k_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.self_attn.o_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.self_attn.q_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.4.self_attn.v_proj.weight": "model-00001-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.0.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.0.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.0.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.2.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.2.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.2.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.3.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.3.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.3.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.4.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.4.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.4.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.5.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.5.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.5.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.6.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.6.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.6.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.7.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.7.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.experts.7.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.block_sparse_moe.gate.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.input_layernorm.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.post_attention_layernorm.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.self_attn.k_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.self_attn.o_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.self_attn.q_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.5.self_attn.v_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.0.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.0.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.0.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.2.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.2.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.2.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.3.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.3.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.3.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.4.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.4.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.4.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.5.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.5.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.5.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.6.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.6.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.6.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.7.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.7.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.experts.7.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.block_sparse_moe.gate.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.input_layernorm.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.post_attention_layernorm.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.self_attn.k_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.self_attn.o_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.self_attn.q_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.6.self_attn.v_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.0.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.0.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.0.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.2.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.2.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.2.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.3.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.3.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.3.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.4.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.4.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.4.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.5.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.5.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.5.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.6.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.6.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.6.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.7.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.7.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.experts.7.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.block_sparse_moe.gate.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.input_layernorm.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.post_attention_layernorm.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.self_attn.k_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.self_attn.o_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.self_attn.q_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.7.self_attn.v_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.0.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.0.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.0.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.2.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.2.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.2.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.3.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.3.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.3.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.4.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.4.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.4.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.5.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.5.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.5.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.6.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.6.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.6.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.7.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.7.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.experts.7.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.block_sparse_moe.gate.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.input_layernorm.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.post_attention_layernorm.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.self_attn.k_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.self_attn.o_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.self_attn.q_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.8.self_attn.v_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.0.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.0.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.0.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.2.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.2.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.2.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.3.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.3.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.3.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.4.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.4.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.4.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.5.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.5.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.5.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.6.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.6.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.6.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.7.w1.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.7.w2.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.experts.7.w3.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.block_sparse_moe.gate.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.input_layernorm.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.post_attention_layernorm.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.self_attn.k_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.self_attn.o_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.self_attn.q_proj.weight": "model-00002-of-00008.safetensors",
+        "model.layers.9.self_attn.v_proj.weight": "model-00002-of-00008.safetensors",
+        "model.norm.weight": "model-00008-of-00008.safetensors"
+    }
+}

modeling_jiutian.py

+# coding=utf-8
+# Copyright 2023 Mistral AI and the HuggingFace Inc. team. All rights reserved.
+#
+# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
+# and OPT implementations in this library. It has been modified from its
+# original forms to accommodate minor architectural differences compared
+# to GPT-NeoX and OPT used by the Meta AI team that trained the model.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" PyTorch Jiutian-MoE model."""
+import inspect
+import math
+import warnings
+from typing import List, Optional, Tuple, Union
+
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+from torch import nn
+from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
+
+from transformers.activations import ACT2FN
+from transformers.cache_utils import Cache, DynamicCache
+from transformers.modeling_attn_mask_utils import (
+    _prepare_4d_causal_attention_mask,
+)
+from transformers.modeling_outputs import (
+    MoeCausalLMOutputWithPast,
+    MoeModelOutputWithPast,
+    SequenceClassifierOutputWithPast,
+)
+from transformers.modeling_utils import PreTrainedModel
+from transformers.pytorch_utils import is_torch_greater_or_equal_than_1_13
+from transformers.utils import (
+    is_flash_attn_2_available,
+    is_flash_attn_greater_or_equal_2_10,
+    logging,
+)
+from transformers.utils.import_utils import is_torch_fx_available
+from .configuration_jiutian import JiutianConfig
+
+
+if is_flash_attn_2_available():
+    from flash_attn import flash_attn_func, flash_attn_varlen_func
+    from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input  # noqa
+
+    _flash_supports_window_size = "window_size" in list(inspect.signature(flash_attn_func).parameters)
+
+# This makes `_prepare_4d_causal_attention_mask` a leaf function in the FX graph.
+# It means that the function will not be traced through and simply appear as a node in the graph.
+if is_torch_fx_available():
+    if not is_torch_greater_or_equal_than_1_13:
+        import torch.fx
+
+    _prepare_4d_causal_attention_mask = torch.fx.wrap(_prepare_4d_causal_attention_mask)
+
+
+logger = logging.get_logger(__name__)
+
+_CONFIG_FOR_DOC = "JiutianConfig"
+
+
+def load_balancing_loss_func(gate_logits: torch.Tensor, num_experts: torch.Tensor = None, top_k=2) -> float:
+    r"""
+    Computes auxiliary load balancing loss as in Switch Transformer - implemented in Pytorch.
+
+    See Switch Transformer (https://arxiv.org/abs/2101.03961) for more details. This function implements the loss
+    function presented in equations (4) - (6) of the paper. It aims at penalizing cases where the routing between
+    experts is too unbalanced.
+
+    Args:
+        gate_logits (Union[`torch.Tensor`, Tuple[torch.Tensor]):
+            Logits from the `gate`, should be a tuple of tensors. Shape: [batch_size, seqeunce_length, num_experts].
+        num_experts (`int`, *optional*):
+            Number of experts
+
+    Returns:
+        The auxiliary loss.
+    """
+    if gate_logits is None:
+        return 0
+
+    if isinstance(gate_logits, tuple):
+        # cat along the layers?
+        compute_device = gate_logits[0].device
+        gate_logits = torch.cat([gate.to(compute_device) for gate in gate_logits], dim=0)
+
+    routing_weights, selected_experts = torch.topk(gate_logits, top_k, dim=-1)
+    routing_weights = routing_weights.softmax(dim=-1)
+
+    # cast the expert indices to int64, otherwise one-hot encoding will fail
+    if selected_experts.dtype != torch.int64:
+        selected_experts = selected_experts.to(torch.int64)
+
+    if len(selected_experts.shape) == 2:
+        selected_experts = selected_experts.unsqueeze(2)
+
+    expert_mask = torch.nn.functional.one_hot(selected_experts, num_experts)
+
+    # For a given token, determine if it was routed to a given expert.
+    expert_mask = torch.max(expert_mask, axis=-2).values
+
+    # cast to float32 otherwise mean will fail
+    expert_mask = expert_mask.to(torch.float32)
+    tokens_per_group_and_expert = torch.mean(expert_mask, axis=-2)
+
+    router_prob_per_group_and_expert = torch.mean(routing_weights, axis=-1)
+    return torch.mean(tokens_per_group_and_expert * router_prob_per_group_and_expert.unsqueeze(-1)) * (num_experts**2)
+
+
+def _get_unpad_data(attention_mask):
+    seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
+    indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
+    max_seqlen_in_batch = seqlens_in_batch.max().item()
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
+    return (
+        indices,
+        cu_seqlens,
+        max_seqlen_in_batch,
+    )
+
+
+class JiutianRMSNorm(nn.Module):
+    def __init__(self, hidden_size, eps=1e-6):
+        super().__init__()
+        self.weight = nn.Parameter(torch.ones(hidden_size))
+        self.variance_epsilon = eps
+
+    def forward(self, hidden_states):
+        input_dtype = hidden_states.dtype
+        hidden_states = hidden_states.to(torch.float32)
+        variance = hidden_states.pow(2).mean(-1, keepdim=True)
+        hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
+        return self.weight * hidden_states.to(input_dtype)
+
+
+class JiutianContextualRotaryEmbedding(nn.Module):
+
+    def __init__(
+            self,
+            head_num,
+            head_dim,
+            npos_max
+    ):
+        super().__init__()
+
+        self.npos_max = npos_max
+        # 每个头都有独立的emb
+        self.pos_emb = nn.parameter.Parameter(
+            torch.zeros(head_num, head_dim, npos_max)
+        )
+        logger.info(f"ContextualRotaryEmbedding shape: {head_num}, {head_dim}, {npos_max}")
+
+    def forward(self, query, attn_logits) -> torch.Tensor:
+        gates = torch.sigmoid(attn_logits)
+        pos = gates.flip(-1).cumsum(dim=-1).flip(-1)
+        pos = pos.clamp(max=self.npos_max - 1)
+
+        pos_ceil = pos.ceil().long()
+        pos_floor = pos.floor().long()
+        # 仅最后两个维度会进行matmul
+        # [b, nq, s, nh] * [1(b), nq, nh, e] -> [b, nq, s, e]
+        logits_int = torch.matmul(query, self.pos_emb)
+        logits_ceil = logits_int.gather(-1, pos_ceil)
+        logits_floor = logits_int.gather(-1, pos_floor)
+        w = pos - pos_floor
+        return logits_ceil * w + logits_floor * (1 - w)
+
+
+class JiutianRotaryEmbedding(nn.Module):
+    def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None):
+        super().__init__()
+
+        self.dim = dim
+        self.max_position_embeddings = max_position_embeddings
+        self.base = base
+        inv_freq = 1.0 / (self.base ** (torch.arange(0, self.dim, 2).float().to(device) / self.dim))
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+
+        # Build here to make `torch.jit.trace` work.
+        self._set_cos_sin_cache(
+            seq_len=max_position_embeddings, device=self.inv_freq.device, dtype=torch.get_default_dtype()
+        )
+
+    def _set_cos_sin_cache(self, seq_len, device, dtype):
+        self.max_seq_len_cached = seq_len
+        t = torch.arange(self.max_seq_len_cached, device=device, dtype=self.inv_freq.dtype)
+
+        freqs = torch.outer(t, self.inv_freq)
+        # Different from paper, but it uses a different permutation in order to obtain the same calculation
+        emb = torch.cat((freqs, freqs), dim=-1)
+        self.register_buffer("cos_cached", emb.cos().to(dtype), persistent=False)
+        self.register_buffer("sin_cached", emb.sin().to(dtype), persistent=False)
+
+    def forward(self, x, seq_len=None):
+        # x: [bs, num_attention_heads, seq_len, head_size]
+        if seq_len > self.max_seq_len_cached:
+            self._set_cos_sin_cache(seq_len=seq_len, device=x.device, dtype=x.dtype)
+
+        return (
+            self.cos_cached[:seq_len].to(dtype=x.dtype),
+            self.sin_cached[:seq_len].to(dtype=x.dtype),
+        )
+
+
+def rotate_half(x):
+    """Rotates half the hidden dims of the input."""
+    x1 = x[..., : x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2 :]
+    return torch.cat((-x2, x1), dim=-1)
+
+
+def apply_rotary_pos_emb(q, k, cos, sin, position_ids, unsqueeze_dim=1):
+    """Applies Rotary Position Embedding to the query and key tensors.
+
+    Args:
+        q (`torch.Tensor`): The query tensor.
+        k (`torch.Tensor`): The key tensor.
+        cos (`torch.Tensor`): The cosine part of the rotary embedding.
+        sin (`torch.Tensor`): The sine part of the rotary embedding.
+        position_ids (`torch.Tensor`):
+            The position indices of the tokens corresponding to the query and key tensors. For example, this can be
+            used to pass offsetted position ids when working with a KV-cache.
+        unsqueeze_dim (`int`, *optional*, defaults to 1):
+            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
+            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
+            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
+            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
+            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
+            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
+    Returns:
+        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
+    """
+    cos = cos[position_ids].unsqueeze(unsqueeze_dim)
+    sin = sin[position_ids].unsqueeze(unsqueeze_dim)
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    return q_embed, k_embed
+
+
+def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
+    """
+    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
+    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
+    """
+    batch, num_key_value_heads, slen, head_dim = hidden_states.shape
+    if n_rep == 1:
+        return hidden_states
+    hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
+    return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)
+
+
+class JiutianAttention(nn.Module):
+    """
+    Multi-headed attention from 'Attention Is All You Need' paper. Modified to use sliding window attention: Longformer
+    and "Generating Long Sequences with Sparse Transformers".
+    """
+
+    def __init__(self, config: JiutianConfig, layer_idx: Optional[int] = None):
+        super().__init__()
+        self.config = config
+        self.layer_idx = layer_idx
+        if layer_idx is None:
+            logger.warning_once(
+                f"Instantiating {self.__class__.__name__} without passing `layer_idx` is not recommended and will "
+                "to errors during the forward call, if caching is used. Please make sure to provide a `layer_idx` "
+                "when creating this class."
+            )
+
+        self.hidden_size = config.hidden_size
+        self.num_heads = config.num_attention_heads
+        self.head_dim = self.hidden_size // self.num_heads
+        self.num_key_value_heads = config.num_key_value_heads
+        self.num_key_value_groups = self.num_heads // self.num_key_value_heads
+        self.max_position_embeddings = config.max_position_embeddings
+        self.rope_theta = config.rope_theta
+        self.is_causal = True
+        self.attention_dropout = config.attention_dropout
+
+        if (self.head_dim * self.num_heads) != self.hidden_size:
+            raise ValueError(
+                f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}"
+                f" and `num_heads`: {self.num_heads})."
+            )
+        self.q_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=False)
+        self.k_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False)
+        self.v_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False)
+        self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=False)
+
+        if config.use_cope:
+            self.cope = JiutianContextualRotaryEmbedding(
+                head_num=self.num_heads,
+                head_dim=self.head_dim,
+                npos_max=self.max_position_embeddings,
+            )
+        else:
+            self.rotary_emb = JiutianRotaryEmbedding(
+                self.head_dim,
+                max_position_embeddings=self.max_position_embeddings,
+                base=self.rope_theta,
+            )
+
+    def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
+        return tensor.view(bsz, seq_len, self.num_heads, self.head_dim).transpose(1, 2).contiguous()
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+        **kwargs,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        if "padding_mask" in kwargs:
+            warnings.warn(
+                "Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
+            )
+        bsz, q_len, _ = hidden_states.size()
+
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+
+        kv_seq_len = key_states.shape[-2]
+        if past_key_value is not None:
+            if self.layer_idx is None:
+                raise ValueError(
+                    f"The cache structure has changed since version v4.36. If you are using {self.__class__.__name__} "
+                    "for auto-regressive decoding with k/v caching, please make sure to initialize the attention class "
+                    "with a layer index."
+                )
+            kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
+        cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
+
+        if past_key_value is not None:
+            cache_kwargs = {"sin": sin, "cos": cos}  # Specific to RoPE models
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        # repeat k/v heads if n_kv_heads < n_heads
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+        attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
+
+        if attn_weights.size() != (bsz, self.num_heads, q_len, kv_seq_len):
+            raise ValueError(
+                f"Attention weights should be of size {(bsz, self.num_heads, q_len, kv_seq_len)}, but is"
+                f" {attn_weights.size()}"
+            )
+
+        if attention_mask is not None:
+            if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
+                raise ValueError(
+                    f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.size()}"
+                )
+
+            attn_weights = attn_weights + attention_mask
+
+        # upcast attention to fp32
+        attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype)
+        attn_weights = nn.functional.dropout(attn_weights, p=self.attention_dropout, training=self.training)
+        attn_output = torch.matmul(attn_weights, value_states)
+
+        if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
+            raise ValueError(
+                f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is"
+                f" {attn_output.size()}"
+            )
+
+        attn_output = attn_output.transpose(1, 2).contiguous()
+        attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
+
+        attn_output = self.o_proj(attn_output)
+
+        if not output_attentions:
+            attn_weights = None
+
+        return attn_output, attn_weights, past_key_value
+
+
+class JiutianFlashAttention2(JiutianAttention):
+    
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+
+        # TODO: Should be removed once Flash Attention for RoCm is bumped to 2.1.
+        # flash_attn<2.1 generates top-left aligned causal mask, while what is needed here is bottom-right alignement, that was made default for flash_attn>=2.1. This attribute is used to handle this difference. Reference: https://github.com/Dao-AILab/flash-attention/releases/tag/v2.1.0.
+        # Beware that with flash_attn<2.1, using q_seqlen != k_seqlen (except for the case q_seqlen == 1) produces a wrong mask (top-left).
+        self._flash_attn_uses_top_left_mask = not is_flash_attn_greater_or_equal_2_10()
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+        **kwargs,
+    ):
+        if "padding_mask" in kwargs:
+            warnings.warn(
+                "Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
+            )
+
+            # overwrite attention_mask with padding_mask
+            attention_mask = kwargs.pop("padding_mask")
+        bsz, q_len, _ = hidden_states.size()
+
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+
+        kv_seq_len = key_states.shape[-2]
+        if past_key_value is not None:
+            if self.layer_idx is None:
+                raise ValueError(
+                    f"The cache structure has changed since version v4.36. If you are using {self.__class__.__name__} "
+                    "for auto-regressive decoding with k/v caching, please make sure to initialize the attention class "
+                    "with a layer index."
+                )
+            kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
+
+        # Because the input can be padded, the absolute sequence length depends on the max position id.
+        rotary_seq_len = max(kv_seq_len, position_ids[:, -1].max().item()) + 1
+        cos, sin = self.rotary_emb(value_states, seq_len=rotary_seq_len)
+
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
+
+        use_sliding_windows = (
+            _flash_supports_window_size
+            and getattr(self.config, "sliding_window", None) is not None
+            and kv_seq_len > self.config.sliding_window
+        )
+
+        if not _flash_supports_window_size:
+            logger.warning_once(
+                "The current flash attention version does not support sliding window attention, for a more memory efficient implementation"
+                " make sure to upgrade flash-attn library."
+            )
+
+        if past_key_value is not None:
+            # Activate slicing cache only if the config has a value `sliding_windows` attribute
+            cache_has_contents = past_key_value.get_seq_length(self.layer_idx) > 0
+            if (
+                getattr(self.config, "sliding_window", None) is not None
+                and kv_seq_len > self.config.sliding_window
+                and cache_has_contents
+            ):
+                slicing_tokens = 1 - self.config.sliding_window
+
+                past_key = past_key_value[self.layer_idx][0]
+                past_value = past_key_value[self.layer_idx][1]
+
+                past_key = past_key[:, :, slicing_tokens:, :].contiguous()
+                past_value = past_value[:, :, slicing_tokens:, :].contiguous()
+
+                if past_key.shape[-2] != self.config.sliding_window - 1:
+                    raise ValueError(
+                        f"past key must have a shape of (`batch_size, num_heads, self.config.sliding_window-1, head_dim`), got"
+                        f" {past_key.shape}"
+                    )
+
+                if attention_mask is not None:
+                    attention_mask = attention_mask[:, slicing_tokens:]
+                    attention_mask = torch.cat([attention_mask, torch.ones_like(attention_mask[:, -1:])], dim=-1)
+
+            cache_kwargs = {"sin": sin, "cos": cos}  # Specific to RoPE models
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        # repeat k/v heads if n_kv_heads < n_heads
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+        dropout_rate = 0.0 if not self.training else self.attention_dropout
+
+        # In PEFT, usually we cast the layer norms in float32 for training stability reasons
+        # therefore the input hidden states gets silently casted in float32. Hence, we need
+        # cast them back in float16 just to be sure everything works as expected.
+        input_dtype = query_states.dtype
+        if input_dtype == torch.float32:
+            # Handle the case where the model is quantized
+            if hasattr(self.config, "_pre_quantization_dtype"):
+                target_dtype = self.config._pre_quantization_dtype
+            else:
+                target_dtype = self.q_proj.weight.dtype
+
+            logger.warning_once(
+                f"The input hidden states seems to be silently casted in float32, this might be related to"
+                f" the fact you have upcasted embedding or layer norm layers in float32. We will cast back the input in"
+                f" {target_dtype}."
+            )
+
+            query_states = query_states.to(target_dtype)
+            key_states = key_states.to(target_dtype)
+            value_states = value_states.to(target_dtype)
+
+        # Reashape to the expected shape for Flash Attention
+        query_states = query_states.transpose(1, 2)
+        key_states = key_states.transpose(1, 2)
+        value_states = value_states.transpose(1, 2)
+
+        attn_output = self._flash_attention_forward(
+            query_states,
+            key_states,
+            value_states,
+            attention_mask,
+            q_len,
+            dropout=dropout_rate,
+            use_sliding_windows=use_sliding_windows,
+        )
+
+        attn_output = attn_output.reshape(bsz, q_len, self.hidden_size).contiguous()
+        attn_output = self.o_proj(attn_output)
+
+        if not output_attentions:
+            attn_weights = None
+
+        return attn_output, attn_weights, past_key_value
+
+    def _flash_attention_forward(
+        self,
+        query_states,
+        key_states,
+        value_states,
+        attention_mask,
+        query_length,
+        dropout=0.0,
+        softmax_scale=None,
+        use_sliding_windows=False,
+    ):
+        """
+        Calls the forward method of Flash Attention - if the input hidden states contain at least one padding token
+        first unpad the input, then computes the attention scores and pad the final attention scores.
+
+        Args:
+            query_states (`torch.Tensor`):
+                Input query states to be passed to Flash Attention API
+            key_states (`torch.Tensor`):
+                Input key states to be passed to Flash Attention API
+            value_states (`torch.Tensor`):
+                Input value states to be passed to Flash Attention API
+            attention_mask (`torch.Tensor`):
+                The padding mask - corresponds to a tensor of size `(batch_size, seq_len)` where 0 stands for the
+                position of padding tokens and 1 for the position of non-padding tokens.
+            dropout (`int`, *optional*):
+                Attention dropout
+            softmax_scale (`float`, *optional*):
+                The scaling of QK^T before applying softmax. Default to 1 / sqrt(head_dim)
+            use_sliding_windows (`bool`, *optional*):
+                Whether to activate sliding window attention.
+        """
+        if not self._flash_attn_uses_top_left_mask:
+            causal = self.is_causal
+        else:
+            causal = self.is_causal and query_length != 1
+
+        # Contains at least one padding token in the sequence
+        if attention_mask is not None:
+            batch_size = query_states.shape[0]
+            query_states, key_states, value_states, indices_q, cu_seq_lens, max_seq_lens = self._upad_input(
+                query_states, key_states, value_states, attention_mask, query_length
+            )
+
+            cu_seqlens_q, cu_seqlens_k = cu_seq_lens
+            max_seqlen_in_batch_q, max_seqlen_in_batch_k = max_seq_lens
+
+            if not use_sliding_windows:
+                attn_output_unpad = flash_attn_varlen_func(
+                    query_states,
+                    key_states,
+                    value_states,
+                    cu_seqlens_q=cu_seqlens_q,
+                    cu_seqlens_k=cu_seqlens_k,
+                    max_seqlen_q=max_seqlen_in_batch_q,
+                    max_seqlen_k=max_seqlen_in_batch_k,
+                    dropout_p=dropout,
+                    softmax_scale=softmax_scale,
+                    causal=causal,
+                )
+            else:
+                attn_output_unpad = flash_attn_varlen_func(
+                    query_states,
+                    key_states,
+                    value_states,
+                    cu_seqlens_q=cu_seqlens_q,
+                    cu_seqlens_k=cu_seqlens_k,
+                    max_seqlen_q=max_seqlen_in_batch_q,
+                    max_seqlen_k=max_seqlen_in_batch_k,
+                    dropout_p=dropout,
+                    softmax_scale=softmax_scale,
+                    causal=causal,
+                    window_size=(self.config.sliding_window, self.config.sliding_window),
+                )
+
+            attn_output = pad_input(attn_output_unpad, indices_q, batch_size, query_length)
+        else:
+            if not use_sliding_windows:
+                attn_output = flash_attn_func(
+                    query_states,
+                    key_states,
+                    value_states,
+                    dropout,
+                    softmax_scale=softmax_scale,
+                    causal=causal,
+                )
+            else:
+                attn_output = flash_attn_func(
+                    query_states,
+                    key_states,
+                    value_states,
+                    dropout,
+                    softmax_scale=softmax_scale,
+                    causal=causal,
+                    window_size=(self.config.sliding_window, self.config.sliding_window),
+                )
+
+        return attn_output
+
+    def _upad_input(self, query_layer, key_layer, value_layer, attention_mask, query_length):
+        batch_size, kv_seq_len, num_heads, head_dim = key_layer.shape
+
+        # On the first iteration we need to properly re-create the padding mask
+        # by slicing it on the proper place
+        if kv_seq_len != attention_mask.shape[-1]:
+            attention_mask_num_tokens = attention_mask.shape[-1]
+            attention_mask = attention_mask[:, attention_mask_num_tokens - kv_seq_len :]
+
+        indices_k, cu_seqlens_k, max_seqlen_in_batch_k = _get_unpad_data(attention_mask)
+
+        key_layer = index_first_axis(key_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k)
+        value_layer = index_first_axis(value_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k)
+
+        if query_length == kv_seq_len:
+            query_layer = index_first_axis(
+                query_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k
+            )
+            cu_seqlens_q = cu_seqlens_k
+            max_seqlen_in_batch_q = max_seqlen_in_batch_k
+            indices_q = indices_k
+        elif query_length == 1:
+            max_seqlen_in_batch_q = 1
+            cu_seqlens_q = torch.arange(
+                batch_size + 1, dtype=torch.int32, device=query_layer.device
+            )  # There is a memcpy here, that is very bad.
+            indices_q = cu_seqlens_q[:-1]
+            query_layer = query_layer.squeeze(1)
+        else:
+            # The -q_len: slice assumes left padding.
+            attention_mask = attention_mask[:, -query_length:]
+            query_layer, indices_q, cu_seqlens_q, max_seqlen_in_batch_q = unpad_input(query_layer, attention_mask)
+
+        return (
+            query_layer,
+            key_layer,
+            value_layer,
+            indices_q,
+            (cu_seqlens_q, cu_seqlens_k),
+            (max_seqlen_in_batch_q, max_seqlen_in_batch_k),
+        )
+
+
+class JiutianBlockSparseTop2MLP(nn.Module):
+    def __init__(self, config: JiutianConfig, is_sub_expert: bool = False):
+        super().__init__()
+        self.ffn_dim = config.intermediate_size
+        if is_sub_expert:
+            assert self.ffn_dim % 2 == 0
+            self.ffn_dim //= 2
+        self.hidden_dim = config.hidden_size
+
+        self.w1 = nn.Linear(self.hidden_dim, self.ffn_dim, bias=False)
+        self.w2 = nn.Linear(self.ffn_dim, self.hidden_dim, bias=False)
+        self.w3 = nn.Linear(self.hidden_dim, self.ffn_dim, bias=False)
+
+        self.act_fn = ACT2FN[config.hidden_act]
+
+    def forward(self, hidden_states):
+        current_hidden_states = self.act_fn(self.w1(hidden_states)) * self.w3(hidden_states)
+        current_hidden_states = self.w2(current_hidden_states)
+        return current_hidden_states
+
+
+class JiutianMoEDroplessTokenDispatcher:
+    """
+    []
+    """
+
+    def __init__(
+            self, num_local_experts: int, num_experts_per_tok: int
+    ) -> None:
+        """
+        Initialize the zero token dropping router.
+        """
+        self.hidden_shape = None
+        self.num_local_experts = num_local_experts
+        assert self.num_local_experts > 0, "Expected at least one expert"
+        self.router_topk = num_experts_per_tok
+        self.expert_mask = None
+
+    def get_expert_mask(self, device, dtype):
+        mask = torch.eye(self.num_local_experts,
+                         device=device).to(dtype)
+        mask[2:, :2] = -1 * self.router_topk
+        mask[1, 1] = 0
+        mask[0, 1] = 1
+        logger.info(f"expert_mask: {mask}")
+        return mask
+
+    def token_permutation(
+            self, hidden_states: torch.Tensor, max_prob: torch.Tensor, max_ind: torch.Tensor
+    ):
+        if self.expert_mask is None:
+            self.expert_mask = self.get_expert_mask(device=hidden_states.device, dtype=hidden_states.dtype)
+
+        self.hidden_shape = hidden_states.shape  # [N, h]
+
+        # [N, E_2] -> [N, E_2, E]
+        mask = torch.nn.functional.one_hot(max_ind, num_classes=self.num_local_experts).to(hidden_states.dtype)
+        token_dispatcher = torch.matmul(self.expert_mask, mask.sum(dim=1).transpose(0, 1))  # [E, N]
+        token_dispatcher[token_dispatcher > 0.5] = 1
+        token_dispatcher[token_dispatcher < 0.5] = 0
+
+        # [N, 1, E_2] * [N, E_2, E] -> [N, 1, E] -> [N, E] -> [E, N]
+        score_dispather = torch.matmul(max_prob.unsqueeze(dim=1), mask).squeeze(dim=1).transpose(0, 1)
+        score_dispather *= token_dispatcher
+        score_dispather[0, :] = token_dispatcher[0, :].to(score_dispather.dtype)
+
+        # 获取各个expert被分配token数
+        # counter = token_dispatcher.sum(dim=1)
+        # tokens_per_expert[0] = counter[0]
+        # tokens_per_expert[2:] = counter[1:]
+        # tokens_per_expert = tokens_per_expert.cpu().to(torch.long)
+        tokens_per_expert = token_dispatcher.sum(dim=1).cpu().to(torch.long)
+
+        global_local_map = token_dispatcher.nonzero()[:, 1]
+        global_local_map = global_local_map.view(-1, 1).expand(-1, hidden_states.shape[-1])
+        permuted_local_hidden_states = torch.gather(hidden_states, 0, global_local_map)
+        local_probs = score_dispather.masked_select(token_dispatcher.bool()).view(-1, 1)
+
+        num_token_need = token_dispatcher.sum()
+        assert num_token_need == permuted_local_hidden_states.shape[0], f"num_token_need {num_token_need}, hidden_shape {self.hidden_shape}"
+        assert local_probs.shape[0] == permuted_local_hidden_states.shape[0]
+
+        return (
+            permuted_local_hidden_states,
+            tokens_per_expert,
+            local_probs,
+            global_local_map,
+        )
+
+    def token_unpermutation(
+            self,
+            hidden_states: torch.Tensor,
+            scores: torch.Tensor,
+            global_local_map: torch.Tensor = None
+    ):
+        scores = scores.to(dtype=hidden_states.dtype)
+        unpermuted_local_hidden = hidden_states * scores
+
+        unpermuted_global_hidden = torch.zeros(
+            self.hidden_shape,
+            dtype=hidden_states.dtype,
+            device=hidden_states.device,
+        )
+        
+        output_total = unpermuted_global_hidden.scatter_add(
+            0, global_local_map, unpermuted_local_hidden
+        )
+
+        return output_total
+
+
+Jiutian_ATTENTION_CLASSES = {
+    "eager": JiutianAttention,
+    "flash_attention_2": JiutianFlashAttention2,
+}
+
+
+class JiutianSparseMoeBlock(nn.Module):
+    """
+    This implementation is
+    strictly equivalent to standard MoE with full capacity (no
+    dropped tokens). It's faster since it formulates MoE operations
+    in terms of block-sparse operations to accomodate imbalanced
+    assignments of tokens to experts, whereas standard MoE either
+    (1) drop tokens at the cost of reduced performance or (2) set
+    capacity factor to number of experts and thus waste computation
+    and memory on padding.
+    """
+
+    def __init__(self, config):
+        super().__init__()
+        self.hidden_dim = config.hidden_size
+        self.ffn_dim = config.intermediate_size
+        self.num_experts = config.num_local_experts
+        self.top_k = config.num_experts_per_tok
+
+        # gating
+        self.gate = nn.Linear(self.hidden_dim, self.num_experts, bias=False)
+        self.token_dispatcher = JiutianMoEDroplessTokenDispatcher(
+            self.num_experts, self.top_k
+        )
+        self.experts = nn.ModuleList([JiutianBlockSparseTop2MLP(config), nn.Module()])  # expert 0 & 1
+        self.experts.extend([JiutianBlockSparseTop2MLP(config, is_sub_expert=True) for _ in range(2, self.num_experts)])
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        """ [YMJ] """
+        batch_size, sequence_length, hidden_dim = hidden_states.shape  # [b, s, h]
+        hidden_states = hidden_states.view(-1, hidden_dim)  # [b * s -> N, h]
+        # router_logits: (batch * sequence_length, n_experts)
+        router_logits = self.gate(hidden_states)  # [N, E]
+
+        routing_weights = F.softmax(router_logits, dim=1, dtype=torch.float)
+        routing_weights, selected_experts = torch.topk(routing_weights, self.top_k, dim=-1)  # [N, E_k]
+        # 重新将top的数值归一化
+        routing_weights = torch.softmax(routing_weights, dim=-1, dtype=torch.float).type_as(router_logits)
+
+        # 置换操作
+        (
+            dispatched_input,
+            tokens_per_expert,
+            routing_weights,
+            global_local_map,
+        ) = self.token_dispatcher.token_permutation(hidden_states, routing_weights, selected_experts)
+
+        output_local = torch.zeros_like(dispatched_input)
+        cumsum_num_tokens = torch.cumsum(tokens_per_expert, dim=0)
+        # Insert zero at the begining for offset index's convenience
+        zero_tensor = torch.zeros(1, dtype=torch.long)
+        cumsum_num_tokens = torch.cat((zero_tensor, cumsum_num_tokens))
+        for expert_num, expert in enumerate(self.experts):
+            start = cumsum_num_tokens[expert_num]
+            end = cumsum_num_tokens[expert_num + 1]
+            if expert_num == 1:
+                continue
+            hidden = dispatched_input[start:end]
+            output = expert(hidden)
+
+            output_local[start:end] = output
+
+        final_hidden_states = self.token_dispatcher.token_unpermutation(
+            output_local, routing_weights, global_local_map
+        )
+        final_hidden_states = final_hidden_states.reshape(batch_size, sequence_length, hidden_dim)
+        return final_hidden_states, router_logits
+
+
+class JiutianDecoderLayer(nn.Module):
+    def __init__(self, config: JiutianConfig, layer_idx: int):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+
+        self.self_attn = Jiutian_ATTENTION_CLASSES[config._attn_implementation](config, layer_idx)
+
+        self.block_sparse_moe = JiutianSparseMoeBlock(config)
+        self.input_layernorm = JiutianRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.post_attention_layernorm = JiutianRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Tuple[torch.Tensor]] = None,
+        output_attentions: Optional[bool] = False,
+        output_router_logits: Optional[bool] = False,
+        use_cache: Optional[bool] = False,
+        **kwargs,
+    ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
+        if "padding_mask" in kwargs:
+            warnings.warn(
+                "Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
+            )
+        """
+        Args:
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
+            attention_mask (`torch.FloatTensor`, *optional*): attention mask of size
+                `(batch, sequence_length)` where padding elements are indicated by 0.
+            past_key_value (`Tuple(torch.FloatTensor)`, *optional*): cached past key and value projection states
+            output_attentions (`bool`, *optional*):
+                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
+                returned tensors for more detail.
+            output_router_logits (`bool`, *optional*):
+                Whether or not to return the logits of all the routers. They are useful for computing the router loss, and
+                should not be returned during inference.
+            use_cache (`bool`, *optional*):
+                If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding
+                (see `past_key_values`).
+        """
+
+        residual = hidden_states
+
+        hidden_states = self.input_layernorm(hidden_states)
+
+        # Self Attention
+        hidden_states, self_attn_weights, present_key_value = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_value=past_key_value,
+            output_attentions=output_attentions,
+            use_cache=use_cache,
+        )
+        hidden_states = residual + hidden_states
+
+        # Fully Connected
+        residual = hidden_states
+        hidden_states = self.post_attention_layernorm(hidden_states)
+        hidden_states, router_logits = self.block_sparse_moe(hidden_states)
+        hidden_states = residual + hidden_states
+
+        outputs = (hidden_states,)
+
+        if output_attentions:
+            outputs += (self_attn_weights,)
+
+        if use_cache:
+            outputs += (present_key_value,)
+
+        if output_router_logits:
+            outputs += (router_logits,)
+
+        return outputs
+
+
+class JiutianPreTrainedModel(PreTrainedModel):
+    config_class = JiutianConfig
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["JiutianDecoderLayer"]
+    _skip_keys_device_placement = "past_key_values"
+    _supports_flash_attn_2 = True
+    _supports_cache_class = True
+
+    def _init_weights(self, module):
+        std = self.config.initializer_range
+        if isinstance(module, nn.Linear):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.padding_idx is not None:
+                module.weight.data[module.padding_idx].zero_()
+
+
+class JiutianModel(JiutianPreTrainedModel):
+
+    def __init__(self, config: JiutianConfig):
+        super().__init__(config)
+        self.padding_idx = config.pad_token_id
+        self.vocab_size = config.vocab_size
+
+        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
+        self.layers = nn.ModuleList(
+            [JiutianDecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
+        )
+        self._use_flash_attention_2 = config._attn_implementation == "flash_attention_2"
+        self.norm = JiutianRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+        self.gradient_checkpointing = False
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.embed_tokens = value
+
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        output_router_logits: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, MoeModelOutputWithPast]:
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_router_logits = (
+            output_router_logits if output_router_logits is not None else self.config.output_router_logits
+        )
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        use_cache = use_cache if use_cache is not None else self.config.use_cache
+
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        # retrieve input_ids and inputs_embeds
+        if input_ids is not None and inputs_embeds is not None:
+            raise ValueError("You cannot specify both decoder_input_ids and decoder_inputs_embeds at the same time")
+        elif input_ids is not None:
+            batch_size, seq_length = input_ids.shape
+        elif inputs_embeds is not None:
+            batch_size, seq_length, _ = inputs_embeds.shape
+        else:
+            raise ValueError("You have to specify either decoder_input_ids or decoder_inputs_embeds")
+
+        past_key_values_length = 0
+
+        if self.gradient_checkpointing and self.training:
+            if use_cache:
+                logger.warning_once(
+                    "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
+                )
+                use_cache = False
+
+        if use_cache:
+            use_legacy_cache = not isinstance(past_key_values, Cache)
+            if use_legacy_cache:
+                past_key_values = DynamicCache.from_legacy_cache(past_key_values)
+            past_key_values_length = past_key_values.get_usable_length(seq_length)
+
+        if position_ids is None:
+            device = input_ids.device if input_ids is not None else inputs_embeds.device
+            position_ids = torch.arange(
+                past_key_values_length, seq_length + past_key_values_length, dtype=torch.long, device=device
+            )
+            position_ids = position_ids.unsqueeze(0).view(-1, seq_length)
+        else:
+            position_ids = position_ids.view(-1, seq_length).long()
+
+        if inputs_embeds is None:
+            inputs_embeds = self.embed_tokens(input_ids)
+
+        if attention_mask is not None and self._use_flash_attention_2 and use_cache:
+            is_padding_right = attention_mask[:, -1].sum().item() != batch_size
+            if is_padding_right:
+                raise ValueError(
+                    "You are attempting to perform batched generation with padding_side='right'"
+                    " this may lead to unexpected behaviour for Flash Attention version of Jiutian-MoE. Make sure to "
+                    " call `tokenizer.padding_side  = 'left'` before tokenizing the input. "
+                )
+
+        if self._use_flash_attention_2:
+            # 2d mask is passed through the layers
+            attention_mask = attention_mask if (attention_mask is not None and 0 in attention_mask) else None
+        else:
+            # 4d mask is passed through the layers
+            attention_mask = _prepare_4d_causal_attention_mask(
+                attention_mask,
+                (batch_size, seq_length),
+                inputs_embeds,
+                past_key_values_length,
+                sliding_window=self.config.sliding_window,
+            )
+
+        hidden_states = inputs_embeds
+
+        # decoder layers
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attns = () if output_attentions else None
+        all_router_logits = () if output_router_logits else None
+        next_decoder_cache = None
+
+        for decoder_layer in self.layers:
+            if output_hidden_states:
+                all_hidden_states += (hidden_states,)
+
+            if self.gradient_checkpointing and self.training:
+                layer_outputs = self._gradient_checkpointing_func(
+                    decoder_layer.__call__,
+                    hidden_states,
+                    attention_mask,
+                    position_ids,
+                    past_key_values,
+                    output_attentions,
+                    output_router_logits,
+                    use_cache,
+                )
+            else:
+                layer_outputs = decoder_layer(
+                    hidden_states,
+                    attention_mask=attention_mask,
+                    position_ids=position_ids,
+                    past_key_value=past_key_values,
+                    output_attentions=output_attentions,
+                    output_router_logits=output_router_logits,
+                    use_cache=use_cache,
+                )
+
+            hidden_states = layer_outputs[0]
+
+            if use_cache:
+                next_decoder_cache = layer_outputs[2 if output_attentions else 1]
+
+            if output_attentions:
+                all_self_attns += (layer_outputs[1],)
+
+            if output_router_logits:
+                all_router_logits += (layer_outputs[-1],)
+
+        hidden_states = self.norm(hidden_states)
+
+        # add hidden states from the last decoder layer
+        if output_hidden_states:
+            all_hidden_states += (hidden_states,)
+
+        next_cache = None
+        if use_cache:
+            next_cache = next_decoder_cache.to_legacy_cache() if use_legacy_cache else next_decoder_cache
+
+        if not return_dict:
+            return tuple(
+                v
+                for v in [hidden_states, next_cache, all_hidden_states, all_self_attns, all_router_logits]
+                if v is not None
+            )
+        return MoeModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=next_cache,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attns,
+            router_logits=all_router_logits,
+        )
+
+
+class JiutianForCausalLM(JiutianPreTrainedModel):
+    _tied_weights_keys = ["lm_head.weight"]
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.model = JiutianModel(config)
+        self.vocab_size = config.vocab_size
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+        self.router_aux_loss_coef = config.router_aux_loss_coef
+        self.num_experts = config.num_local_experts
+        self.num_experts_per_tok = config.num_experts_per_tok
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.model.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.model.embed_tokens = value
+
+    def get_output_embeddings(self):
+        return self.lm_head
+
+    def set_output_embeddings(self, new_embeddings):
+        self.lm_head = new_embeddings
+
+    def set_decoder(self, decoder):
+        self.model = decoder
+
+    def get_decoder(self):
+        return self.model
+
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        output_router_logits: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, MoeCausalLMOutputWithPast]:
+
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_router_logits = (
+            output_router_logits if output_router_logits is not None else self.config.output_router_logits
+        )
+
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
+        outputs = self.model(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            output_router_logits=output_router_logits,
+            return_dict=return_dict,
+        )
+
+        hidden_states = outputs[0]
+        logits = self.lm_head(hidden_states)
+        logits = logits.float()
+
+        loss = None
+        if labels is not None:
+            # Shift so that tokens < n predict n
+            shift_logits = logits[..., :-1, :].contiguous()
+            shift_labels = labels[..., 1:].contiguous()
+            # Flatten the tokens
+            loss_fct = CrossEntropyLoss()
+            shift_logits = shift_logits.view(-1, self.config.vocab_size)
+            shift_labels = shift_labels.view(-1)
+            # Enable model parallelism
+            shift_labels = shift_labels.to(shift_logits.device)
+            loss = loss_fct(shift_logits, shift_labels)
+
+        aux_loss = None
+        if output_router_logits:
+            aux_loss = load_balancing_loss_func(
+                outputs.router_logits if return_dict else outputs[-1], self.num_experts, self.num_experts_per_tok
+            )
+            if labels is not None:
+                loss += self.router_aux_loss_coef * aux_loss
+
+        if not return_dict:
+            output = (logits,) + outputs[1:]
+            if output_router_logits:
+                output = (aux_loss,) + output
+            return (loss,) + output if loss is not None else output
+
+        return MoeCausalLMOutputWithPast(
+            loss=loss,
+            aux_loss=aux_loss,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+            router_logits=outputs.router_logits,
+        )
+
+    def prepare_inputs_for_generation(
+        self, input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, **kwargs
+    ):
+        # Omit tokens covered by past_key_values
+        if past_key_values is not None:
+            if isinstance(past_key_values, Cache):
+                cache_length = past_key_values.get_seq_length()
+                past_length = past_key_values.seen_tokens
+                max_cache_length = past_key_values.get_max_length()
+            else:
+                cache_length = past_length = past_key_values[0][0].shape[2]
+                max_cache_length = None
+
+            # Keep only the unprocessed tokens:
+            # 1 - If the length of the attention_mask exceeds the length of input_ids, then we are in a setting where
+            # some of the inputs are exclusivelly passed as part of the cache (e.g. when passing input_embeds as
+            # input)
+            if attention_mask is not None and attention_mask.shape[1] > input_ids.shape[1]:
+                input_ids = input_ids[:, -(attention_mask.shape[1] - past_length) :]
+            # 2 - If the past_length is smaller than input_ids', then input_ids holds all input tokens. We can discard
+            # input_ids based on the past_length.
+            elif past_length < input_ids.shape[1]:
+                input_ids = input_ids[:, past_length:]
+            # 3 - Otherwise (past_length >= input_ids.shape[1]), let's assume input_ids only has unprocessed tokens.
+
+            # If we are about to go beyond the maximum cache length, we need to crop the input attention mask.
+            if (
+                max_cache_length is not None
+                and attention_mask is not None
+                and cache_length + input_ids.shape[1] > max_cache_length
+            ):
+                attention_mask = attention_mask[:, -max_cache_length:]
+
+        position_ids = kwargs.get("position_ids", None)
+        if attention_mask is not None and position_ids is None:
+            # create position_ids on the fly for batch generation
+            position_ids = attention_mask.long().cumsum(-1) - 1
+            position_ids.masked_fill_(attention_mask == 0, 1)
+            if past_key_values:
+                position_ids = position_ids[:, -input_ids.shape[1] :]
+
+        # if `inputs_embeds` are passed, we only want to use them in the 1st generation step
+        if inputs_embeds is not None and past_key_values is None:
+            model_inputs = {"inputs_embeds": inputs_embeds}
+        else:
+            model_inputs = {"input_ids": input_ids}
+
+        model_inputs.update(
+            {
+                "position_ids": position_ids,
+                "past_key_values": past_key_values,
+                "use_cache": kwargs.get("use_cache"),
+                "attention_mask": attention_mask,
+            }
+        )
+        return model_inputs
+
+    @staticmethod
+    def _reorder_cache(past_key_values, beam_idx):
+        reordered_past = ()
+        for layer_past in past_key_values:
+            reordered_past += (
+                tuple(past_state.index_select(0, beam_idx.to(past_state.device)) for past_state in layer_past),
+            )
+        return reordered_past
+
+
+class JiutianForSequenceClassification(JiutianPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.model = JiutianModel(config)
+        self.score = nn.Linear(config.hidden_size, self.num_labels, bias=False)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.model.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.model.embed_tokens = value
+
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, SequenceClassifierOutputWithPast]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
+            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        transformer_outputs = self.model(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+        hidden_states = transformer_outputs[0]
+        logits = self.score(hidden_states)
+
+        if input_ids is not None:
+            batch_size = input_ids.shape[0]
+        else:
+            batch_size = inputs_embeds.shape[0]
+
+        if self.config.pad_token_id is None and batch_size != 1:
+            raise ValueError("Cannot handle batch sizes > 1 if no padding token is defined.")
+        if self.config.pad_token_id is None:
+            sequence_lengths = -1
+        else:
+            if input_ids is not None:
+                sequence_lengths = (torch.eq(input_ids, self.config.pad_token_id).int().argmax(-1) - 1).to(
+                    logits.device
+                )
+            else:
+                sequence_lengths = -1
+
+        pooled_logits = logits[torch.arange(batch_size, device=logits.device), sequence_lengths]
+
+        loss = None
+        if labels is not None:
+            labels = labels.to(logits.device)
+            if self.config.problem_type is None:
+                if self.num_labels == 1:
+                    self.config.problem_type = "regression"
+                elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
+                    self.config.problem_type = "single_label_classification"
+                else:
+                    self.config.problem_type = "multi_label_classification"
+
+            if self.config.problem_type == "regression":
+                loss_fct = MSELoss()
+                if self.num_labels == 1:
+                    loss = loss_fct(pooled_logits.squeeze(), labels.squeeze())
+                else:
+                    loss = loss_fct(pooled_logits, labels)
+            elif self.config.problem_type == "single_label_classification":
+                loss_fct = CrossEntropyLoss()
+                loss = loss_fct(pooled_logits.view(-1, self.num_labels), labels.view(-1))
+            elif self.config.problem_type == "multi_label_classification":
+                loss_fct = BCEWithLogitsLoss()
+                loss = loss_fct(pooled_logits, labels)
+        if not return_dict:
+            output = (pooled_logits,) + transformer_outputs[1:]
+            return ((loss,) + output) if loss is not None else output
+
+        return SequenceClassifierOutputWithPast(
+            loss=loss,
+            logits=pooled_logits,
+            past_key_values=transformer_outputs.past_key_values,
+            hidden_states=transformer_outputs.hidden_states,
+            attentions=transformer_outputs.attentions,
+        )
+

special_tokens_map.json

+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": "<unk>",
+  "pad_token": "<unk>",
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenization_jiutian.py

+# coding=utf-8
+# Copyright 2022 EleutherAI and the HuggingFace Inc. team. All rights reserved.
+#
+# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
+# and OPT implementations in this library. It has been modified from its
+# original forms to accommodate minor architectural differences compared
+# to GPT-NeoX and OPT used by the Meta AI team that trained the model.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""Tokenization classes for jiutian."""
+import os
+from shutil import copyfile
+from typing import Any, Dict, List, Optional, Tuple
+
+import sentencepiece as spm
+
+from transformers.tokenization_utils import AddedToken, PreTrainedTokenizer
+from transformers.utils import logging
+
+logger = logging.get_logger(__name__)
+
+VOCAB_FILES_NAMES = {"vocab_file": "tokenizer.model"}
+
+SPIECE_UNDERLINE = "▁"
+
+
+class JiutianTokenizer(PreTrainedTokenizer):
+    vocab_files_names = VOCAB_FILES_NAMES
+    model_input_names = ["input_ids", "attention_mask"]
+
+    def __init__(
+        self,
+        vocab_file,
+        unk_token="<unk>",
+        bos_token="<s>",
+        eos_token="</s>",
+        pad_token=None,
+        sp_model_kwargs: Optional[Dict[str, Any]] = None,
+        add_bos_token=True,
+        add_eos_token=False,
+        clean_up_tokenization_spaces=False,
+        spaces_between_special_tokens=False,
+        legacy=None,
+        add_prefix_space=True,
+        **kwargs,
+    ):
+        self.sp_model_kwargs = {} if sp_model_kwargs is None else sp_model_kwargs
+        bos_token = AddedToken(bos_token, normalized=False, special=True) if isinstance(bos_token, str) else bos_token
+        eos_token = AddedToken(eos_token, normalized=False, special=True) if isinstance(eos_token, str) else eos_token
+        unk_token = AddedToken(unk_token, normalized=False, special=True) if isinstance(unk_token, str) else unk_token
+        pad_token = AddedToken(pad_token, normalized=False, special=True) if isinstance(pad_token, str) else pad_token
+
+        self.legacy = legacy
+        self.vocab_file = vocab_file
+        self.add_bos_token = add_bos_token
+        self.add_eos_token = add_eos_token
+        self.add_prefix_space = add_prefix_space
+
+        self.sp_model = spm.SentencePieceProcessor(**self.sp_model_kwargs)
+        self.sp_model.Load(self.vocab_file)
+
+        super().__init__(
+            bos_token=bos_token,
+            eos_token=eos_token,
+            unk_token=unk_token,
+            pad_token=pad_token,
+            add_bos_token=add_bos_token,
+            add_eos_token=add_eos_token,
+            sp_model_kwargs=self.sp_model_kwargs,
+            clean_up_tokenization_spaces=clean_up_tokenization_spaces,
+            spaces_between_special_tokens=spaces_between_special_tokens,
+            legacy=legacy,
+            add_prefix_space=add_prefix_space,
+            **kwargs,
+        )
+
+    @property
+    def unk_token_length(self):
+        return len(self.sp_model.encode(str(self.unk_token)))
+
+    def __getstate__(self):
+        state = self.__dict__.copy()
+        state["sp_model"] = None
+        return state
+
+    def __setstate__(self, d):
+        self.__dict__ = d
+        self.sp_model = spm.SentencePieceProcessor(**self.sp_model_kwargs)
+
+    @property
+    def vocab_size(self):
+        """Returns vocab size"""
+        return self.sp_model.get_piece_size()
+
+    def get_vocab(self):
+        """Returns vocab as a dict"""
+        vocab = {self.convert_ids_to_tokens(i): i for i in range(self.vocab_size)}
+        vocab.update(self.added_tokens_encoder)
+        return vocab
+
+    # Copied from transformers.models.t5.tokenization_t5.T5Tokenizer._tokenize
+    def _tokenize(self, text, **kwargs):
+        """
+        Returns a tokenized string.
+
+        We de-activated the `add_dummy_prefix` option, thus the sentencepiece internals will always strip any
+        SPIECE_UNDERLINE. For example: `self.sp_model.encode(f"{SPIECE_UNDERLINE}Hey", out_type = str)` will give
+        `['H', 'e', 'y']` instead of `['▁He', 'y']`. Thus we always encode `f"{unk_token}text"` and strip the
+        `unk_token`. Here is an example with `unk_token = "<unk>"` and `unk_token_length = 4`.
+        `self.tokenizer.sp_model.encode("<unk> Hey", out_type = str)[4:]`.
+        """
+        return self.sp_model.encode(text, out_type=str)
+
+    def _convert_token_to_id(self, token):
+        """Converts a token (str) in an id using the vocab."""
+        return self.sp_model.piece_to_id(token)
+
+    def _convert_id_to_token(self, index):
+        """Converts an index (integer) in a token (str) using the vocab."""
+        token = self.sp_model.IdToPiece(index)
+        return token
+
+    def convert_tokens_to_string(self, tokens):
+        """Converts a sequence of tokens (string) in a single string."""
+        # since we manually add the prefix space, we have to remove it when decoding
+        if tokens[0].startswith(SPIECE_UNDERLINE) and self.add_prefix_space:
+            tokens[0] = tokens[0][1:]
+
+        current_sub_tokens = []
+        out_string = ""
+        prev_is_special = False
+        for i, token in enumerate(tokens):
+            # make sure that special tokens are not decoded using sentencepiece model
+            if token in self.all_special_tokens:
+                if not prev_is_special and i != 0 and self.legacy:
+                    out_string += " "
+                out_string += self.sp_model.decode(current_sub_tokens) + token
+                prev_is_special = True
+                current_sub_tokens = []
+            else:
+                if prev_is_special and i == 1 and self.add_prefix_space and not token.startswith(SPIECE_UNDERLINE):
+                    out_string += " "
+                current_sub_tokens.append(token)
+                prev_is_special = False
+        out_string += self.sp_model.decode(current_sub_tokens)
+        return out_string
+
+    def save_vocabulary(self, save_directory, filename_prefix: Optional[str] = None) -> Tuple[str]:
+        """
+        Save the vocabulary and special tokens file to a directory.
+
+        Args:
+            save_directory (`str`):
+                The directory in which to save the vocabulary.
+
+        Returns:
+            `Tuple(str)`: Paths to the files saved.
+        """
+        if not os.path.isdir(save_directory):
+            logger.error(f"Vocabulary path ({save_directory}) should be a directory")
+            return
+        out_vocab_file = os.path.join(
+            save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
+        )
+
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
+            copyfile(self.vocab_file, out_vocab_file)
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
+
+        return (out_vocab_file,)
+
+    def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
+        bos_token_id = [self.bos_token_id] if self.add_bos_token else []
+        eos_token_id = [self.eos_token_id] if self.add_eos_token else []
+
+        output = bos_token_id + token_ids_0 + eos_token_id
+
+        if token_ids_1 is not None:
+            output = output + bos_token_id + token_ids_1 + eos_token_id
+
+        return output
+
+    def get_special_tokens_mask(
+        self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None, already_has_special_tokens: bool = False
+    ) -> List[int]:
+        """
+        Retrieve sequence ids from a token list that has no special tokens added. This method is called when adding
+        special tokens using the tokenizer `prepare_for_model` method.
+
+        Args:
+            token_ids_0 (`List[int]`):
+                List of IDs.
+            token_ids_1 (`List[int]`, *optional*):
+                Optional second list of IDs for sequence pairs.
+            already_has_special_tokens (`bool`, *optional*, defaults to `False`):
+                Whether or not the token list is already formatted with special tokens for the model.
+
+        Returns:
+            `List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
+        """
+        if already_has_special_tokens:
+            return super().get_special_tokens_mask(
+                token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True
+            )
+
+        bos_token_id = [1] if self.add_bos_token else []
+        eos_token_id = [1] if self.add_eos_token else []
+
+        if token_ids_1 is None:
+            return bos_token_id + ([0] * len(token_ids_0)) + eos_token_id
+        return (
+            bos_token_id
+            + ([0] * len(token_ids_0))
+            + eos_token_id
+            + bos_token_id
+            + ([0] * len(token_ids_1))
+            + eos_token_id
+        )
+
+    def create_token_type_ids_from_sequences(
+        self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None
+    ) -> List[int]:
+        """
+        Creates a mask from the two sequences passed to be used in a sequence-pair classification task. An ALBERT
+        sequence pair mask has the following format:
+
+        ```
+        0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1
+        | first sequence    | second sequence |
+        ```
+
+        if token_ids_1 is None, only returns the first portion of the mask (0s).
+
+        Args:
+            token_ids_0 (`List[int]`):
+                List of ids.
+            token_ids_1 (`List[int]`, *optional*):
+                Optional second list of IDs for sequence pairs.
+
+        Returns:
+            `List[int]`: List of [token type IDs](../glossary#token-type-ids) according to the given sequence(s).
+        """
+        bos_token_id = [self.bos_token_id] if self.add_bos_token else []
+        eos_token_id = [self.eos_token_id] if self.add_eos_token else []
+
+        output = [0] * len(bos_token_id + token_ids_0 + eos_token_id)
+
+        if token_ids_1 is not None:
+            output += [1] * len(bos_token_id + token_ids_1 + eos_token_id)
+
+        return output

tokenizer_config.json

+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "add_prefix_space": true,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "auto_map": {
+    "AutoTokenizer": [
+      "tokenization_jiutian.JiutianTokenizer",
+      null
+    ]
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<unk>",
+  "legacy": true,
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<unk>",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "JiutianTokenizer",
+  "trust_remote_code": true,
+  "unk_token": "<unk>"
+}