Initial commit

LICENSE

+MIT License
+
+Copyright (c) 2023 DeepSeek
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
\ No newline at end of file

README.md

+## DeepSeek-V4
+## 论文
+[DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence](https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro/blob/main/DeepSeek_V4.pdf)
+
+## 模型简介
+DeepSeek-V4 系列的预览版本，包含两款强大的混合专家（Mixture-of-Experts, MoE）语言模型：DeepSeek-V4-Pro（总参数量 1.6T，激活参数量 49B）和 DeepSeek-V4-Flash（总参数量 284B，激活参数量 13B），两者均支持 百万 token 的上下文长度。
+
+DeepSeek-V4 系列在架构与优化方面引入了多项关键升级：
+
+- 混合注意力架构：设计了一种混合注意力机制，结合压缩稀疏注意力（Compressed Sparse Attention, CSA）与重度压缩注意力（Heavily Compressed Attention, HCA），显著提升长上下文处理效率。在百万 token 上下文场景下，DeepSeek-V4-Pro 相比 DeepSeek-V3.2 仅需 27% 的单 token 推理 FLOPs 和 10% 的 KV 缓存。
+- 流形约束超连接（Manifold-Constrained Hyper-Connections, mHC）：在传统残差连接基础上引入 mHC，增强跨层信号传播的稳定性，同时保留模型的表达能力。
+- Muon 优化器：采用 Muon 优化器，以实现更快的收敛速度和更高的训练稳定性。
+
+## 环境依赖
+| 软件 |   版本    |
+| :------: |:-------:|
+|     DTK      |  26.04  |
+|    python    | 3.10.12 |
+|    torch     |  2.9.0+das.opt1.dtk2604.20260331.g4e3c1e7  |
+|    tilelang     |  0.1.7.post3+cpu.git52700923  |
+
+
+推荐使用镜像:harbor.sourcefind.cn:5443/dcu/admin/base/custom:torch-2.9.0-ubuntu22.04-dtk26.04-deepseek-v4-0425
+
+- 挂载地址`-v`根据实际模型情况修改
+
+```bash
+docker run -it \
+    --shm-size 200g \
+    --network=host \
+    --name deepseek-v4 \
+    --privileged \
+    --device=/dev/kfd \
+    --device=/dev/dri \
+    --device=/dev/mkfd \
+    --group-add video \
+    --cap-add=SYS_PTRACE \
+    --security-opt seccomp=unconfined \
+    -u root \
+    -v /opt/hyhal/:/opt/hyhal/:ro \
+    -v /path/your_code_data/:/path/your_code_data/ \
+    harbor.sourcefind.cn:5443/dcu/admin/base/custom:torch-2.9.0-ubuntu22.04-dtk26.04-deepseek-v4-0425 bash
+```
+
+更多镜像可前往[光源](https://sourcefind.cn/#/service-list)下载使用。
+
+关于本项目DCU显卡所需的特殊深度学习库可从[光合](https://developer.sourcefind.cn/tool/)开发者社区下载安装。
+
+## 数据集
+`暂无`
+
+## 训练
+`暂无`
+
+## 推理
+### pytorch
+#### 单机推理
+1. 模型转换与切分
+
+```bash
+#注意将脚本中对应的路径及参数设置成用户实际值
+#其中：INPUT_FP8_HF_PATH为模型下载路径；OUTPUT_BF16_HF_PATH为bf16模型存放路径；SAVE_PATH为切分好的模型路径；mp根据实际卡数调整
+cd convert_weight
+bash convert_weight.sh
+```
+
+2. 启动对话推理
+```bash
+cd ../inference
+#注意将脚本中对应的路径及参数设置成用户实际值
+sh start_torch.sh
+```
+
+## 效果展示
+<div align=center>
+    <img src="./doc/result_dcu.png"/>
+</div>
+
+### 精度
+`DCU与GPU精度一致，推理框架：pytorch。`
+
+## 预训练权重
+| 模型名称  | 权重大小 | DCU型号  | 最低卡数需求 |下载地址|
+|:-----:|:----:|:------:|:------:|:----------:|
+| DeepSeek-V4-Flash | 158B | BW1100 |   8    | [Hugging Face](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash) |
+
+## 源码仓库及问题反馈
+- https://developer.sourcefind.cn/codes/modelzoo/deepseek-v4
+
+## 参考资料
+- https://github.com/deepseek-ai

convert_weight/convert.py

+import os
+import shutil
+from argparse import ArgumentParser
+from glob import glob
+from tqdm import tqdm, trange
+
+import torch
+from safetensors.torch import safe_open, save_file
+
+
+FP4_TABLE = torch.tensor([
+    0.0, 0.5, 1.0, 1.5, 2.0, 3.0, 4.0, 6.0,
+    0.0, -0.5, -1.0, -1.5, -2.0, -3.0, -4.0, -6.0
+], dtype=torch.float32)
+
+
+def cast_e2m1fn_to_e4m3fn(x: torch.Tensor, scale: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]:
+    """
+    Casts a tensor from e2m1fn to e4m3fn losslessly.
+    """
+    assert x.dtype == torch.int8
+    assert x.ndim == 2
+    out_dim, in_dim = x.size()
+    in_dim *= 2
+    fp8_block_size = 128
+    fp4_block_size = 32
+    assert in_dim % fp8_block_size == 0 and out_dim % fp8_block_size == 0
+    assert scale.size(0) == out_dim and scale.size(1) == in_dim // fp4_block_size
+
+    x = x.view(torch.uint8)
+    low  = x & 0x0F
+    high = (x >> 4) & 0x0F
+    x = torch.stack([FP4_TABLE[low.long()], FP4_TABLE[high.long()]], dim=-1).flatten(2)
+
+    # max_fp4 (6.0) * MAX_OFFSET must fit in e4m3fn (max 448)
+    # 6.0 * 2^6 = 384 < 448; 6.0 * 2^7 = 768 > 448; so MAX_OFFSET_BITS = 6
+    MAX_OFFSET_BITS = 6
+
+    bOut = out_dim // fp8_block_size
+    bIn = in_dim // fp8_block_size
+    # bOut, bIn, 128, 128
+    x = x.view(bOut, fp8_block_size, bIn, fp8_block_size).transpose(1, 2)
+    # bOut, bIn, 128*4
+    scale = scale.float().view(bOut, fp8_block_size, bIn, -1).transpose(1, 2).flatten(2)
+    ## bOut, bIn, 1
+    scale_max_offset_bits = scale.amax(dim=-1, keepdim=True) / (2**MAX_OFFSET_BITS)
+    # bOut, bIn, 128*4
+    offset = scale / scale_max_offset_bits
+    # bOut, bIn, 128, 128
+    offset = offset.unflatten(-1, (fp8_block_size, -1)).repeat_interleave(fp4_block_size, dim=-1)
+    x = (x * offset).transpose(1, 2).reshape(out_dim, in_dim)
+    return x.to(torch.float8_e4m3fn), scale_max_offset_bits.squeeze(-1).to(torch.float8_e8m0fnu)
+
+
+mapping = {
+    "embed_tokens": ("embed", 0),
+    "input_layernorm": ("attn_norm", None),
+    "post_attention_layernorm": ("ffn_norm", None),
+    "q_proj": ("wq", 0),
+    "q_a_proj": ("wq_a", None),
+    "q_a_layernorm": ("q_norm", None),
+    "q_b_proj": ("wq_b", 0),
+    "kv_a_proj_with_mqa": ("wkv_a", None),
+    "kv_a_layernorm": ("kv_norm", None),
+    "kv_b_proj": ("wkv_b", 0),
+    "o_proj": ("wo", 1),
+    "gate_proj": ("w1", 0),
+    "down_proj": ("w2", 1),
+    "up_proj": ("w3", 0),
+    "lm_head": ("head", 0),
+
+    "embed": ("embed", 0),
+    "wq_b": ("wq_b", 0),
+    "wo_a": ("wo_a", 0),
+    "wo_b": ("wo_b", 1),
+    "head": ("head", 0),
+    "attn_sink": ("attn_sink", 0),
+    "weights_proj": ("weights_proj", 0),
+}
+
+
+def main(hf_ckpt_path, save_path, n_experts, mp, expert_dtype):
+    """
+    Converts and saves model checkpoint files into a specified format.
+
+    Args:
+        hf_ckpt_path (str): Path to the directory containing the input checkpoint files.
+        save_path (str): Path to the directory where the converted checkpoint files will be saved.
+        n_experts (int): Total number of experts in the model.
+        mp (int): Model parallelism factor.
+        
+    Returns:
+        None
+    """
+    torch.set_num_threads(8)
+    n_local_experts = n_experts // mp
+    state_dicts = [{} for _ in range(mp)]
+
+    for file_path in tqdm(glob(os.path.join(hf_ckpt_path, "*.safetensors"))):
+        with safe_open(file_path, framework="pt", device="cpu") as f:
+            for name in f.keys():
+                param: torch.Tensor = f.get_tensor(name)
+                if name.startswith("model."):
+                    name = name[len("model."):]
+                if name.startswith("mtp.") and ("emb" in name or name.endswith("head.weight")):
+                    continue
+                name = name.replace("self_attn", "attn")
+                name = name.replace("mlp", "ffn")
+                name = name.replace("weight_scale_inv", "scale")
+                name = name.replace("e_score_correction_bias", "bias")
+                if any(x in name for x in ["hc", "attn_sink", "tie2eid", "ape"]):    # without .weight
+                    key = name.split(".")[-1]
+                else:
+                    key = name.split(".")[-2]
+                if key in mapping:
+                    new_key, dim = mapping[key]
+                else:
+                    new_key, dim = key, None
+                name = name.replace(key, new_key)
+                for i in range(mp):
+                    new_param = param
+                    if "experts" in name and "shared_experts" not in name:
+                        idx = int(name.split(".")[-3])
+                        if idx < i * n_local_experts or idx >= (i + 1) * n_local_experts:
+                            continue
+                    elif dim is not None:
+                        changed=True
+                        if not changed:
+                            assert param.size(dim) % mp == 0, f"Dimension {dim} must be divisible by {mp}"
+                            shard_size = param.size(dim) // mp
+                            new_param = param.narrow(dim, i * shard_size, shard_size).contiguous()
+                        else:
+                            print(f"Processing parameter {name} with shape {param.shape} for model parallel shard {i}")
+                            if "wo_a" not in name and "wo_b" not in name:
+                                assert param.size(dim) % mp == 0, f"Dimension {dim} must be divisible by {mp}"
+                                shard_size = param.size(dim) // mp
+                                new_param = param.narrow(dim, i * shard_size, shard_size).contiguous()
+                            else:
+                                num_projection_groups = mp//8
+                                new_mp = mp // num_projection_groups
+                                new_i = i // num_projection_groups
+                                shard_size = param.size(dim) // new_mp
+                                assert(shard_size==1024)
+                                new_param = param.narrow(dim, new_i * shard_size, shard_size).contiguous()
+                    state_dicts[i][name] = new_param
+
+    os.makedirs(save_path, exist_ok=True)
+
+    for i in trange(mp):
+        names = list(state_dicts[i].keys())
+        save_file(state_dicts[i], os.path.join(save_path, f"model{i}-mp{mp}.safetensors"))
+
+    for file in ["tokenizer.json", "tokenizer_config.json"]:
+        old_file_path = os.path.join(hf_ckpt_path, file)
+        new_file_path = os.path.join(save_path, file)
+        shutil.copyfile(old_file_path, new_file_path)
+
+
+if __name__ == "__main__":
+    parser = ArgumentParser()
+    parser.add_argument("--hf-ckpt-path", type=str, required=True)
+    parser.add_argument("--save-path", type=str, required=True)
+    parser.add_argument("--n-experts", type=int, required=True)
+    parser.add_argument("--model-parallel", type=int, required=True)
+    parser.add_argument("--expert-dtype", type=str, choices=["fp8", "fp4"], required=False, default=None)
+    args = parser.parse_args()
+    assert args.n_experts % args.model_parallel == 0, "Number of experts must be divisible by model parallelism"
+    main(args.hf_ckpt_path, args.save_path, args.n_experts, args.model_parallel, args.expert_dtype)

convert_weight/convert_weight.py

+"""
+DeepSeek-V3.2 FP8/FP4 -> BF16 Converter
+
+Key differences from V3:
+  - Scale format: ue8m0 (unsigned E8M0, power-of-2 scales, may be stored as uint8)
+  - weight_dequant uses block-reshape approach (from V3.2 model.py:490)
+  - DSA (DeepSeek Attention): each layer has `indexer` submodule
+      - indexer.wq_b, indexer.wk: FP8 (have weight_scale_inv)
+      - indexer.k_norm (weight+bias), indexer.weights_proj: BF16 (no scale)
+  - MTP layer (layer 61): enorm, hnorm, eh_proj, shared_head.norm/head, embed_tokens
+  - No dependency on kernel.py (pure PyTorch)
+  - 62 layers (0-61), 163 shards, ~92k keys
+  - Experts weights use FP4 (MXFP4 E2M1, packed 2 per uint8, group_size=32, E8M0 scale)
+"""
+
+import os
+import json
+import shutil
+from argparse import ArgumentParser
+from glob import glob
+from tqdm import tqdm
+
+import torch
+from safetensors.torch import load_file, save_file
+
+
+BLOCK_SIZE = 128
+FP4_GROUP_SIZE = 32
+
+# E2M1 FP4 lookup table: 4-bit index -> float value
+# Bit layout: sign(1) | exponent(2) | mantissa(1), bias=1
+_FP4_E2M1_LUT = torch.tensor([
+    0.0, 0.5, 1.0, 1.5, 2.0, 3.0, 4.0, 6.0,
+    -0.0, -0.5, -1.0, -1.5, -2.0, -3.0, -4.0, -6.0,
+], dtype=torch.bfloat16)
+
+
+def dequant_fp4_weight(weight_packed: torch.Tensor, scale_fp8: torch.Tensor) -> torch.Tensor:
+    """
+    Dequantize MXFP4 weight to bf16 by unpacking FP4 E2M1 nibbles via LUT.
+
+    Each int8 byte stores 2 FP4 values (low nibble + high nibble).
+    float4_e2m1fn_x2 does NOT support .to(bfloat16), so we manually unpack.
+    Scale is E8M0, one per group of 32 elements along the last dimension.
+
+    Args:
+        weight_packed: [out_features, in_features/2], int8, each byte = 2 FP4 values
+        scale_fp8:     [out_features, in_features/32], float8_e8m0fnu, E8M0 scale per group of 32
+    Returns:
+        bf16 tensor [out_features, in_features]
+    """
+    out_features, packed_in = weight_packed.shape
+    in_features = packed_in * 2
+
+    # Unpack two FP4 values per byte via nibble extraction + LUT
+    raw = weight_packed.to(torch.uint8)
+    low_nibble = (raw & 0x0F).to(torch.long)
+    high_nibble = (raw >> 4).to(torch.long)
+    lut = _FP4_E2M1_LUT.to(weight_packed.device)
+    low_vals = lut[low_nibble]
+    high_vals = lut[high_nibble]
+    # Interleave: [low_0, high_0, low_1, high_1, ...]
+    fp4_values = torch.stack([low_vals, high_vals], dim=-1).reshape(out_features, in_features)
+
+    # Decode E8M0 scale and expand to match fp4_values
+    scale = decode_e8m0_scale(scale_fp8)
+    if scale.dim() == 2 and scale.shape[0] == out_features:
+        # Scale already shaped [out_features, num_groups_per_row]
+        num_groups_per_row = scale.shape[1]
+    else:
+        # Flat scale: compute groups per row from total count
+        total_scales = scale.numel()
+        num_groups_per_row = total_scales // out_features
+        scale = scale.reshape(out_features, num_groups_per_row)
+    actual_group_size = in_features // num_groups_per_row
+    scale = scale.unsqueeze(-1).expand(-1, -1, actual_group_size).reshape(out_features, in_features)
+
+    return (fp4_values * scale).to(torch.bfloat16)
+
+
+def is_expert_weight(name: str) -> bool:
+    """Check if a weight belongs to an expert (MoE) layer, excluding shared_experts."""
+    return "experts" in name and "shared_experts" not in name
+
+
+def decode_e8m0_scale(scale: torch.Tensor) -> torch.Tensor:
+    """
+    Decode E8M0 (unsigned 8-bit exponent-only) scale to float32.
+    E8M0 stores only the exponent: value = 2^(exp - 127), same as IEEE 754 exponent encoding.
+    If scale is already float32, return as-is.
+    """
+    if scale.dtype == torch.float32:
+        return scale
+    if scale.dtype in (torch.bfloat16, torch.float16):
+        return scale.float()
+    # float8_e8m0fnu: .to(int32) does value conversion not bit reinterpret,
+    # so use native .float() which handles E8M0 correctly
+    if scale.dtype == torch.float8_e8m0fnu:
+        return scale.float()
+    # uint8 / int8 raw E8M0 bytes: interpret as IEEE 754 exponent
+    if scale.element_size() == 1:
+        # Reconstruct float32 from exponent: set exponent bits in IEEE 754 float32
+        # float32 = sign(1) + exponent(8) + mantissa(23)
+        # E8M0 value stored as raw exponent byte -> float = 2^(byte - 127)
+        exp_bits = scale.to(torch.int32) << 23
+        return exp_bits.view(torch.float32)
+    return scale.float()
+
+
+def weight_dequant(weight: torch.Tensor, scale: torch.Tensor, block_size: int = BLOCK_SIZE) -> torch.Tensor:
+    """
+    Dequantize FP8 weight to BF16 using block-wise scale.
+    Based on V3.2 model.py:490-495, works on both CPU and CUDA.
+
+    Each (block_size x block_size) block of the weight is multiplied by one scale value.
+    Handles non-aligned dimensions (e.g. kv_a_proj_with_mqa shape 576x7168, 576 % 128 != 0)
+    by padding to the nearest multiple of block_size, dequantizing, then trimming.
+    """
+    shape = weight.shape
+    assert weight.dim() == 2, f"Expected 2D weight, got {weight.dim()}D"
+    M, N = shape
+
+    # Decode E8M0 scale if needed
+    scale = decode_e8m0_scale(scale)
+
+    # Pad to nearest multiple of block_size if needed
+    pad_m = (block_size - M % block_size) % block_size
+    pad_n = (block_size - N % block_size) % block_size
+    if pad_m or pad_n:
+        weight = torch.nn.functional.pad(weight, (0, pad_n, 0, pad_m))
+    Mp, Np = weight.shape
+
+    # V3.2 dequant: reshape into blocks, scale, reshape back
+    weight = weight.view(
+        Mp // block_size, block_size,
+        Np // block_size, block_size
+    ).transpose(1, 2).contiguous().view(-1, block_size * block_size)
+
+    weight = (weight.float() * scale.reshape(-1, 1)).to(torch.bfloat16)
+
+    weight = weight.view(
+        Mp // block_size, Np // block_size,
+        block_size, block_size
+    ).transpose(1, 2).contiguous().view(Mp, Np)
+
+    # Trim padding
+    if pad_m or pad_n:
+        weight = weight[:M, :N]
+
+    return weight
+
+
+def main(fp8_path, bf16_path, device="cuda"):
+    torch.set_default_dtype(torch.bfloat16)
+
+    if device == "cuda" and not torch.cuda.is_available():
+        print("CUDA not available, falling back to CPU")
+        device = "cpu"
+
+    os.makedirs(bf16_path, exist_ok=True)
+
+    # 1. Copy non-safetensor files (config.json, tokenizer, etc.)
+    print("Copying auxiliary files...")
+    for file_path in glob(os.path.join(fp8_path, "*")):
+        fname = os.path.basename(file_path)
+        if fname.endswith(".safetensors") or fname == "model.safetensors.index.json":
+            continue
+        dst = os.path.join(bf16_path, fname)
+        if os.path.isfile(file_path):
+            shutil.copy2(file_path, dst)
+            print(f"  Copied {fname}")
+
+    # 2. Load model index
+    model_index_file = os.path.join(fp8_path, "model.safetensors.index.json")
+    with open(model_index_file, "r") as f:
+        model_index = json.load(f)
+    weight_map = model_index["weight_map"]
+
+    # 3. Pre-build scale_inv lookup: weight_name -> scale_inv_name
+    # V3.2 naming: "xxx.weight" -> "xxx.weight_scale_inv"
+    scale_inv_map = {}                                                                                                                                                                                    
+    all_scale_names = set()                                                                                                                                                                               
+    for name in weight_map:                                                                                                                                                                               
+        if name.endswith("scale"):                                                                                                                                                                        
+            weight_name = name[:-len("scale")] + "weight"                                                                                                                                                 
+            if weight_name in weight_map:                                                                                                                                                                 
+                all_scale_names.add(name)                                                                                                                                                                 
+                scale_inv_map[weight_name] = name    
+
+    # Separate expert (FP4) vs non-expert (FP8) scale mappings
+    fp4_scale_map = {k: v for k, v in scale_inv_map.items() if is_expert_weight(k)}
+    fp8_scale_map = {k: v for k, v in scale_inv_map.items() if not is_expert_weight(k)}
+
+    print(f"Model: DeepSeek (DeepseekForCausalLM)")
+    print(f"Device: {device}")
+    print(f"Total keys in index: {len(weight_map)}")
+    print(f"FP4 expert weights with scale: {len(fp4_scale_map)}")
+    print(f"FP8 non-expert weights with scale: {len(fp8_scale_map)}")
+    print(f"Scale entries: {len(all_scale_names)}")
+
+    # Cache for loaded safetensor files on CPU (handles cross-shard scale lookups)
+    # All shards are loaded to CPU to avoid GPU OOM; individual tensors are moved to
+    # GPU for dequant one at a time.
+    loaded_files = {}
+    use_cuda = (device == "cuda")
+
+    def get_tensor(tensor_name):
+        """Load tensor from the correct shard file (CPU), with caching."""
+        file_name = weight_map.get(tensor_name)
+        if file_name is None:
+            raise KeyError(tensor_name)
+        if file_name not in loaded_files:
+            file_path = os.path.join(fp8_path, file_name)
+            loaded_files[file_name] = load_file(file_path, device="cpu")
+        return loaded_files[file_name][tensor_name]
+
+    # 4. Process safetensor files one by one
+    safetensor_files = sorted(glob(os.path.join(fp8_path, "*.safetensors")))
+    converted_count = 0
+    kept_count = 0
+
+    for safetensor_file in tqdm(safetensor_files, desc="Converting FP8 -> BF16"):
+        file_name = os.path.basename(safetensor_file)
+        current_state_dict = load_file(safetensor_file, device="cpu")
+        loaded_files[file_name] = current_state_dict
+
+        new_state_dict = {}
+        for weight_name, weight in current_state_dict.items():
+            # Skip scale_inv tensors (will be removed from output)
+            if weight_name in all_scale_names:
+                continue
+
+            if weight.element_size() == 1 and weight_name in scale_inv_map:
+                scale_inv_name = scale_inv_map[weight_name]
+                try:
+                    scale_inv = get_tensor(scale_inv_name)
+                    if is_expert_weight(weight_name):
+                        # FP4 expert weight -> dequantize using MXFP4 logic
+                        # print(f"  FP4 dequant: {weight_name}, weight={weight.shape}, scale={scale_inv.shape}, dtype={scale_inv.dtype}")
+                        if use_cuda:
+                            result = dequant_fp4_weight(weight.cuda(), scale_inv.cuda())
+                            new_state_dict[weight_name] = result.cpu()
+                        else:
+                            new_state_dict[weight_name] = dequant_fp4_weight(weight, scale_inv)
+                    else:
+                        # FP8 non-expert weight -> dequantize using block-wise scale
+                        if use_cuda:
+                            result = weight_dequant(weight.cuda(), scale_inv.cuda())
+                            new_state_dict[weight_name] = result.cpu()
+                        else:
+                            new_state_dict[weight_name] = weight_dequant(weight, scale_inv)
+                    converted_count += 1
+                except KeyError:
+                    print(f"  Warning: scale '{scale_inv_name}' not loadable for {weight_name}, keeping raw")
+                    new_state_dict[weight_name] = weight
+            else:
+                # BF16/FP32 weights: norms, biases, gate.weight, indexer.weights_proj,
+                # indexer.k_norm, MTP layers (enorm, hnorm, eh_proj, shared_head), embed, lm_head, etc.
+                new_state_dict[weight_name] = weight
+                kept_count += 1
+
+        # Save converted shard
+        save_file(new_state_dict, os.path.join(bf16_path, file_name))
+
+        # Memory management: keep at most 2 cached shard files on CPU
+        # (needed for cross-shard scale lookups, e.g. weight in shard N, scale in shard N+1)
+        while len(loaded_files) > 2:
+            oldest = next(iter(loaded_files))
+            del loaded_files[oldest]
+
+    # 5. Update model index: remove all scale_inv entries
+    new_weight_map = {k: v for k, v in weight_map.items() if k not in all_scale_names}
+    new_index = {
+        "metadata": model_index.get("metadata", {}),
+        "weight_map": new_weight_map,
+    }
+    with open(os.path.join(bf16_path, "model.safetensors.index.json"), "w") as f:
+        json.dump(new_index, f, indent=2)
+
+    print(f"\nDone!")
+    print(f"  FP8/FP4 -> BF16 converted: {converted_count}")
+    print(f"  Already BF16/FP32 (kept as-is): {kept_count}")
+    print(f"  Scale entries removed: {len(all_scale_names)}")
+    print(f"  Output keys: {len(new_weight_map)} (was {len(weight_map)})")
+    print(f"  Output saved to: {bf16_path}")
+
+
+if __name__ == "__main__":
+    parser = ArgumentParser(description="Convert DeepSeek-V3.2 FP8/FP4 checkpoint to BF16")
+    parser.add_argument("--input-fp8-hf-path", type=str, required=True,
+                        help="Path to the FP8 HuggingFace model directory (DeepSeek-V3.2)")
+    parser.add_argument("--output-bf16-hf-path", type=str, required=True,
+                        help="Path to the output BF16 model directory")
+    parser.add_argument("--device", type=str, default="cuda", choices=["cuda", "cpu"],
+                        help="Device for dequantization (default: cuda)")
+    args = parser.parse_args()
+    main(args.input_fp8_hf_path, args.output_bf16_hf_path, args.device)

convert_weight/convert_weight.sh

+#!/usr/bin/env bash
+
+set -euo pipefail
+
+# Step 1: fp4/fp8 -> bf16
+INPUT_FP8_HF_PATH="deepseek-ai/DeepSeek-V4-Flash"
+OUTPUT_BF16_HF_PATH="deepseek-ai/DeepSeek-V4-Flash-bf16"
+
+python3 convert_weight.py \
+  --input-fp8-hf-path "${INPUT_FP8_HF_PATH}" \
+  --output-bf16-hf-path "${OUTPUT_BF16_HF_PATH}"
+
+# Step 2: bf16 -> bf16-mp16
+MP=8
+HF_CKPT_BF16_PATH="${OUTPUT_BF16_HF_PATH}"
+SAVE_PATH="deepseek-ai/DeepSeek-V4-Flash-bf16-mp16"
+EXPERTS=256
+
+python3 convert.py \
+  --hf-ckpt-path "${HF_CKPT_BF16_PATH}" \
+  --save-path "${SAVE_PATH}" \
+  --n-experts "${EXPERTS}" \
+  --model-parallel "${MP}"

inference/README.md

+# Inference code for DeepSeek models
+
+First convert huggingface model weight files to the format of this project.
+```bash
+export EXPERTS=256
+export MP=4
+export CONFIG=config.json
+python convert.py --hf-ckpt-path ${HF_CKPT_PATH} --save-path ${SAVE_PATH} --n-experts ${EXPERTS} --model-parallel ${MP}
+```
+
+Then chat with DeepSeek model at will!
+```bash
+torchrun --nproc-per-node ${MP} generate.py --ckpt-path ${SAVE_PATH} --config ${CONFIG} --interactive
+```
+
+Or batch inference from file.
+```bash
+torchrun --nproc-per-node ${MP} generate.py --ckpt-path ${SAVE_PATH} --config ${CONFIG} --input-file ${FILE}
+```
+
+Or multi nodes inference.
+```bash
+torchrun --nnodes ${NODES} --nproc-per-node $((MP / NODES)) --node-rank $RANK --master-addr $ADDR generate.py --ckpt-path ${SAVE_PATH} --config ${CONFIG} --input-file ${FILE}
+```
+
+If you want to use fp8, just remove `"expert_dtype": "fp4"` in `config.json` and specify `--expert-dtype fp8` in `convert.py`.

inference/config.json

+{
+    "vocab_size": 129280,
+    "dim": 4096,
+    "moe_inter_dim": 2048,
+    "n_layers": 43,
+    "n_hash_layers": 3,
+    "n_heads": 64,
+    "n_routed_experts": 256,
+    "n_shared_experts": 1,
+    "n_activated_experts": 6,
+    "score_func": "sqrtsoftplus",
+    "route_scale": 1.5,
+    "swiglu_limit": 10.0,
+    "q_lora_rank": 1024,
+    "head_dim": 512,
+    "rope_head_dim": 64,
+    "o_groups": 8,
+    "o_lora_rank": 1024,
+    "window_size": 128,
+    "original_seq_len": 65536,
+    "rope_theta": 10000,
+    "rope_factor": 16,
+    "beta_fast": 32,
+    "beta_slow": 1,
+    "index_n_heads": 64,
+    "index_head_dim": 128,
+    "index_topk": 512,
+    "hc_mult": 4,
+    "hc_sinkhorn_iters": 20,
+    "dtype": "fp8",
+    "scale_fmt": "ue8m0",
+    "compress_rope_theta": 160000,
+    "compress_ratios": [0, 0, 4, 128, 4, 128, 4, 128, 4, 128, 4, 128, 4, 128, 4, 128, 4, 128, 4, 128, 4, 128, 4, 128, 4, 128, 4, 128, 4, 128, 4, 128, 4, 128, 4, 128, 4, 128, 4, 128, 4, 128, 4, 0]
+}
\ No newline at end of file

inference/encoding_dsv4.py

+"""
+DeepSeek-V4 Encoding
+
+A self-contained implementation for encoding/decoding DeepSeek-V4 chat messages
+with tool calling, thinking mode, and quick instruction task support.
+"""
+
+from typing import Any, Dict, List, Union, Optional, Tuple
+import copy
+import json
+import re
+
+# ============================================================
+# Special Tokens
+# ============================================================
+
+bos_token: str = "<｜begin▁of▁sentence｜>"
+eos_token: str = "<｜end▁of▁sentence｜>"
+thinking_start_token: str = "<think>"
+thinking_end_token: str = "</think>"
+dsml_token: str = "｜DSML｜"
+
+USER_SP_TOKEN = "<｜User｜>"
+ASSISTANT_SP_TOKEN = "<｜Assistant｜>"
+LATEST_REMINDER_SP_TOKEN = "<｜latest_reminder｜>"
+
+# Task special tokens for internal classification tasks
+DS_TASK_SP_TOKENS = {
+    "action": "<｜action｜>",
+    "query": "<｜query｜>",
+    "authority": "<｜authority｜>",
+    "domain": "<｜domain｜>",
+    "title": "<｜title｜>",
+    "read_url": "<｜read_url｜>",
+}
+VALID_TASKS = set(DS_TASK_SP_TOKENS.keys())
+
+# ============================================================
+# Templates
+# ============================================================
+
+system_msg_template: str = "{content}"
+user_msg_template: str = "{content}"
+latest_reminder_msg_template: str = "{content}"
+assistant_msg_template: str = "{reasoning}{content}{tool_calls}" + eos_token
+assistant_msg_wo_eos_template: str = "{reasoning}{content}{tool_calls}"
+thinking_template: str = "{reasoning_content}"
+
+response_format_template: str = (
+    "## Response Format:\n\nYou MUST strictly adhere to the following schema to reply:\n{schema}"
+)
+tool_call_template: str = (
+    "<{dsml_token}invoke name=\"{name}\">\n{arguments}\n</{dsml_token}invoke>"
+)
+tool_calls_template = (
+    "<{dsml_token}{tc_block_name}>\n{tool_calls}\n</{dsml_token}{tc_block_name}>"
+)
+tool_calls_block_name: str = "tool_calls"
+
+tool_output_template: str = (
+    "<tool_result>{content}</tool_result>"
+)
+
+REASONING_EFFORT_MAX = (
+    "Reasoning Effort: Absolute maximum with no shortcuts permitted.\n"
+    "You MUST be very thorough in your thinking and comprehensively decompose the problem to resolve the root cause, rigorously stress-testing your logic against all potential paths, edge cases, and adversarial scenarios.\n"
+    "Explicitly write out your entire deliberation process, documenting every intermediate step, considered alternative, and rejected hypothesis to ensure absolutely no assumption is left unchecked.\n\n"
+)
+
+TOOLS_TEMPLATE = """## Tools
+
+You have access to a set of tools to help answer the user's question. You can invoke tools by writing a "<{dsml_token}tool_calls>" block like the following:
+
+<{dsml_token}tool_calls>
+<{dsml_token}invoke name="$TOOL_NAME">
+<{dsml_token}parameter name="$PARAMETER_NAME" string="true|false">$PARAMETER_VALUE</{dsml_token}parameter>
+...
+</{dsml_token}invoke>
+<{dsml_token}invoke name="$TOOL_NAME2">
+...
+</{dsml_token}invoke>
+</{dsml_token}tool_calls>
+
+String parameters should be specified as is and set `string="true"`. For all other types (numbers, booleans, arrays, objects), pass the value in JSON format and set `string="false"`.
+
+If thinking_mode is enabled (triggered by {thinking_start_token}), you MUST output your complete reasoning inside {thinking_start_token}...{thinking_end_token} BEFORE any tool calls or final response.
+
+Otherwise, output directly after {thinking_end_token} with tool calls or final response.
+
+### Available Tool Schemas
+
+{tool_schemas}
+
+You MUST strictly follow the above defined tool name and parameter schemas to invoke tool calls.
+"""
+
+# ============================================================
+# Utility Functions
+# ============================================================
+
+def to_json(value: Any) -> str:
+    """Serialize a value to JSON string."""
+    try:
+        return json.dumps(value, ensure_ascii=False)
+    except:
+        return json.dumps(value, ensure_ascii=True)
+
+
+def tools_from_openai_format(tools):
+    """Extract function definitions from OpenAI-format tool list."""
+    return [tool["function"] for tool in tools]
+
+
+def tool_calls_from_openai_format(tool_calls):
+    """Convert OpenAI-format tool calls to internal format."""
+    return [
+        {
+            "name": tool_call["function"]["name"],
+            "arguments": tool_call["function"]["arguments"],
+        }
+        for tool_call in tool_calls
+    ]
+
+
+def tool_calls_to_openai_format(tool_calls):
+    """Convert internal tool calls to OpenAI format."""
+    return [
+        {
+            "type": "function",
+            "function": {
+                "name": tool_call["name"],
+                "arguments": tool_call["arguments"],
+            }
+        }
+        for tool_call in tool_calls
+    ]
+
+
+def encode_arguments_to_dsml(tool_call: Dict[str, str]) -> str:
+    """
+    Encode tool call arguments into DSML parameter format.
+
+    Args:
+        tool_call: Dict with "name" and "arguments" (JSON string) keys.
+
+    Returns:
+        DSML-formatted parameter string.
+    """
+    p_dsml_template = '<{dsml_token}parameter name="{key}" string="{is_str}">{value}</{dsml_token}parameter>'
+    P_dsml_strs = []
+
+    try:
+        arguments = json.loads(tool_call["arguments"])
+    except Exception as err:
+        arguments = {"arguments": tool_call["arguments"]}
+
+    for k, v in arguments.items():
+        p_dsml_str = p_dsml_template.format(
+            dsml_token=dsml_token,
+            key=k,
+            is_str="true" if isinstance(v, str) else "false",
+            value=v if isinstance(v, str) else to_json(v),
+        )
+        P_dsml_strs.append(p_dsml_str)
+
+    return "\n".join(P_dsml_strs)
+
+
+def decode_dsml_to_arguments(tool_name: str, tool_args: Dict[str, Tuple[str, str]]) -> Dict[str, str]:
+    """
+    Decode DSML parameters back to a tool call dict.
+
+    Args:
+        tool_name: Name of the tool.
+        tool_args: Dict mapping param_name -> (value, is_string_flag).
+
+    Returns:
+        Dict with "name" and "arguments" (JSON string) keys.
+    """
+    def _decode_value(key: str, value: str, string: str):
+        if string == "true":
+            value = to_json(value)
+        return f"{to_json(key)}: {value}"
+
+    tool_args_json = "{" + ", ".join([_decode_value(k, v, string=is_str) for k, (v, is_str) in tool_args.items()]) + "}"
+    return dict(name=tool_name, arguments=tool_args_json)
+
+
+def render_tools(tools: List[Dict[str, Union[str, Dict[str, Any]]]]) -> str:
+    """
+    Render tool schemas into the system prompt format.
+
+    Args:
+        tools: List of tool schema dicts (each with name, description, parameters).
+
+    Returns:
+        Formatted tools section string.
+    """
+    tools_json = [to_json(t) for t in tools]
+
+    return TOOLS_TEMPLATE.format(
+        tool_schemas="\n".join(tools_json),
+        dsml_token=dsml_token,
+        thinking_start_token=thinking_start_token,
+        thinking_end_token=thinking_end_token,
+    )
+
+
+def find_last_user_index(messages: List[Dict[str, Any]]) -> int:
+    """Find the index of the last user/developer message."""
+    last_user_index = -1
+    for idx in range(len(messages) - 1, -1, -1):
+        if messages[idx].get("role") in ["user", "developer"]:
+            last_user_index = idx
+            break
+    return last_user_index
+
+
+# ============================================================
+# Message Rendering
+# ============================================================
+
+def render_message(index: int, messages: List[Dict[str, Any]], thinking_mode: str, drop_thinking: bool = True, reasoning_effort: Optional[str] = None) -> str:
+    """
+    Render a single message at the given index into its encoded string form.
+
+    This is the core function that converts each message in the conversation
+    into the DeepSeek-V4 format.
+
+    Args:
+        index: Index of the message to render.
+        messages: Full list of messages in the conversation.
+        thinking_mode: Either "chat" or "thinking".
+        drop_thinking: Whether to drop reasoning content from earlier turns.
+        reasoning_effort: Optional reasoning effort level ("max", "high", or None).
+
+    Returns:
+        Encoded string for this message.
+    """
+    assert 0 <= index < len(messages)
+    assert thinking_mode in ["chat", "thinking"], f"Invalid thinking_mode `{thinking_mode}`"
+
+    prompt = ""
+    msg = messages[index]
+    last_user_idx = find_last_user_index(messages)
+
+    role = msg.get("role")
+    content = msg.get("content")
+    tools = msg.get("tools")
+    response_format = msg.get("response_format")
+    tool_calls = msg.get("tool_calls")
+    reasoning_content = msg.get("reasoning_content")
+    wo_eos = msg.get("wo_eos", False)
+
+    if tools:
+        tools = tools_from_openai_format(tools)
+    if tool_calls:
+        tool_calls = tool_calls_from_openai_format(tool_calls)
+
+    # Reasoning effort prefix (only at index 0 in thinking mode with max effort)
+    assert reasoning_effort in ['max', None, 'high'], f"Invalid reasoning effort: {reasoning_effort}"
+    if index == 0 and thinking_mode == "thinking" and reasoning_effort == 'max':
+        prompt += REASONING_EFFORT_MAX
+
+    if role == "system":
+        prompt += system_msg_template.format(content=content or "")
+        if tools:
+            prompt += "\n\n" + render_tools(tools)
+        if response_format:
+            prompt += "\n\n" + response_format_template.format(schema=to_json(response_format))
+
+    elif role == "developer":
+        assert content, f"Invalid message for role `{role}`: {msg}"
+
+        content_developer = USER_SP_TOKEN
+        content_developer += content
+
+        if tools:
+            content_developer += "\n\n" + render_tools(tools)
+        if response_format:
+            content_developer += "\n\n" + response_format_template.format(schema=to_json(response_format))
+
+        prompt += user_msg_template.format(content=content_developer)
+
+    elif role == "user":
+        prompt += USER_SP_TOKEN
+
+        # Handle content blocks (tool results mixed with text)
+        content_blocks = msg.get("content_blocks")
+        if content_blocks:
+            parts = []
+            for block in content_blocks:
+                block_type = block.get("type")
+                if block_type == "text":
+                    parts.append(block.get("text", ""))
+                elif block_type == "tool_result":
+                    tool_content = block.get("content", "")
+                    if isinstance(tool_content, list):
+                        text_parts = []
+                        for b in tool_content:
+                            if b.get("type") == "text":
+                                text_parts.append(b.get("text", ""))
+                            else:
+                                text_parts.append(f"[Unsupported {b.get('type')}]")
+                        tool_content = "\n\n".join(text_parts)
+                    parts.append(tool_output_template.format(content=tool_content))
+                else:
+                    parts.append(f"[Unsupported {block_type}]")
+            prompt += "\n\n".join(parts)
+        else:
+            prompt += content or ""
+
+    elif role == "latest_reminder":
+        prompt += LATEST_REMINDER_SP_TOKEN + latest_reminder_msg_template.format(content=content)
+
+    elif role == "tool":
+        raise NotImplementedError("deepseek_v4 merges tool messages into user; please preprocess with merge_tool_messages()")
+
+    elif role == "assistant":
+        thinking_part = ""
+        tc_content = ""
+
+        if tool_calls:
+            tc_list = [
+                tool_call_template.format(
+                    dsml_token=dsml_token,
+                    name=tc.get("name"),
+                    arguments=encode_arguments_to_dsml(tc)
+                )
+                for tc in tool_calls
+            ]
+            tc_content += '\n\n' + tool_calls_template.format(
+                dsml_token=dsml_token,
+                tool_calls="\n".join(tc_list),
+                tc_block_name=tool_calls_block_name,
+            )
+
+        summary_content = content or ""
+        rc = reasoning_content or ""
+
+        # Check if previous message has a task - if so, this is a task output (no thinking)
+        prev_has_task = index - 1 >= 0 and messages[index - 1].get("task") is not None
+
+        if thinking_mode == "thinking" and not prev_has_task:
+            if not drop_thinking or index > last_user_idx:
+                thinking_part = thinking_template.format(reasoning_content=rc) + thinking_end_token
+            else:
+                thinking_part = ""
+
+        if wo_eos:
+            prompt += assistant_msg_wo_eos_template.format(
+                reasoning=thinking_part,
+                content=summary_content,
+                tool_calls=tc_content,
+            )
+        else:
+            prompt += assistant_msg_template.format(
+                reasoning=thinking_part,
+                content=summary_content,
+                tool_calls=tc_content,
+            )
+    else:
+        raise NotImplementedError(f"Unknown role: {role}")
+
+    # Append transition tokens based on what follows
+    if index + 1 < len(messages) and messages[index + 1].get("role") not in ["assistant", "latest_reminder"]:
+        return prompt
+
+    task = messages[index].get("task")
+    if task is not None:
+        # Task special token for internal classification tasks
+        assert task in VALID_TASKS, f"Invalid task: '{task}'. Valid tasks are: {list(VALID_TASKS)}"
+        task_sp_token = DS_TASK_SP_TOKENS[task]
+
+        if task != "action":
+            # Non-action tasks: append task sp token directly after the message
+            prompt += task_sp_token
+        else:
+            # Action task: append Assistant + thinking token + action sp token
+            prompt += ASSISTANT_SP_TOKEN
+            prompt += thinking_end_token if thinking_mode != "thinking" else thinking_start_token
+            prompt += task_sp_token
+
+    elif messages[index].get("role") in ["user", "developer"]:
+        # Normal generation: append Assistant + thinking token
+        prompt += ASSISTANT_SP_TOKEN
+        if not drop_thinking and thinking_mode == "thinking":
+            prompt += thinking_start_token
+        elif drop_thinking and thinking_mode == "thinking" and index >= last_user_idx:
+            prompt += thinking_start_token
+        else:
+            prompt += thinking_end_token
+
+    return prompt
+
+
+# ============================================================
+# Preprocessing
+# ============================================================
+
+def merge_tool_messages(messages: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
+    """
+    Merge tool messages into the preceding user message using content_blocks format.
+
+    DeepSeek-V4 does not have a standalone "tool" role; instead, tool results
+    are encoded as <tool_result> blocks within user messages.
+
+    This function converts a standard OpenAI-format conversation (with separate
+    "tool" role messages) into V4 format where tool results are merged into
+    user messages.
+
+    Args:
+        messages: List of message dicts in OpenAI format.
+
+    Returns:
+        Processed message list with tool messages merged into user messages.
+    """
+    merged: List[Dict[str, Any]] = []
+
+    for msg in messages:
+        msg = copy.deepcopy(msg)
+        role = msg.get("role")
+
+        if role == "tool":
+            # Convert tool message to a user message with tool_result block
+            tool_block = {
+                "type": "tool_result",
+                "tool_use_id": msg.get("tool_call_id", ""),
+                "content": msg.get("content", ""),
+            }
+            # Merge into previous message if it's already a user (merged tool)
+            if merged and merged[-1].get("role") == "user" and "content_blocks" in merged[-1]:
+                merged[-1]["content_blocks"].append(tool_block)
+            else:
+                merged.append({
+                    "role": "user",
+                    "content_blocks": [tool_block],
+                })
+        elif role == "user":
+            text_block = {"type": "text", "text": msg.get("content", "")}
+            if merged and merged[-1].get("role") == "user" and "content_blocks" in merged[-1] and merged[-1].get("task") is None:
+                merged[-1]["content_blocks"].append(text_block)
+            else:
+                new_msg = {
+                    "role": "user",
+                    "content": msg.get("content", ""),
+                    "content_blocks": [text_block],
+                }
+                # Preserve extra fields (task, wo_eos, mask, etc.)
+                for key in ("task", "wo_eos", "mask"):
+                    if key in msg:
+                        new_msg[key] = msg[key]
+                merged.append(new_msg)
+        else:
+            merged.append(msg)
+
+    return merged
+
+
+def sort_tool_results_by_call_order(messages: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
+    """
+    Sort tool_result blocks within user messages by the order of tool_calls
+    in the preceding assistant message.
+
+    Args:
+        messages: Preprocessed message list (after merge_tool_messages).
+
+    Returns:
+        Message list with sorted tool result blocks.
+    """
+    last_tool_call_order: Dict[str, int] = {}
+
+    for msg in messages:
+        role = msg.get("role")
+        if role == "assistant" and msg.get("tool_calls"):
+            last_tool_call_order = {}
+            for idx, tc in enumerate(msg["tool_calls"]):
+                tc_id = tc.get("id") or tc.get("function", {}).get("id", "")
+                if tc_id:
+                    last_tool_call_order[tc_id] = idx
+
+        elif role == "user" and msg.get("content_blocks"):
+            tool_blocks = [b for b in msg["content_blocks"] if b.get("type") == "tool_result"]
+            if len(tool_blocks) > 1 and last_tool_call_order:
+                sorted_blocks = sorted(
+                    tool_blocks,
+                    key=lambda b: last_tool_call_order.get(b.get("tool_use_id", ""), 0)
+                )
+                sorted_idx = 0
+                new_blocks = []
+                for block in msg["content_blocks"]:
+                    if block.get("type") == "tool_result":
+                        new_blocks.append(sorted_blocks[sorted_idx])
+                        sorted_idx += 1
+                    else:
+                        new_blocks.append(block)
+                msg["content_blocks"] = new_blocks
+
+    return messages
+
+
+# ============================================================
+# Main Encoding Function
+# ============================================================
+
+def encode_messages(
+    messages: List[Dict[str, Any]],
+    thinking_mode: str,
+    context: Optional[List[Dict[str, Any]]] = None,
+    drop_thinking: bool = True,
+    add_default_bos_token: bool = True,
+    reasoning_effort: Optional[str] = None,
+) -> str:
+    """
+    Encode a list of messages into the DeepSeek-V4 prompt format.
+
+    This is the main entry point for encoding conversations. It handles:
+    - BOS token insertion
+    - Thinking mode with optional reasoning content dropping
+    - Tool message merging into user messages
+    - Multi-turn conversation context
+
+    Args:
+        messages: List of message dicts to encode.
+        thinking_mode: Either "chat" or "thinking".
+        context: Optional preceding context messages (already encoded prefix).
+        drop_thinking: If True, drop reasoning_content from earlier assistant turns
+                      (only keep reasoning for messages after the last user message).
+        add_default_bos_token: Whether to prepend BOS token at conversation start.
+        reasoning_effort: Optional reasoning effort level ("max", "high", or None).
+
+    Returns:
+        The encoded prompt string.
+    """
+    context = context if context else []
+
+    # Preprocess: merge tool messages and sort tool results
+    messages = merge_tool_messages(messages)
+    messages = sort_tool_results_by_call_order(context + messages)[len(context):]
+    if context:
+        context = merge_tool_messages(context)
+        context = sort_tool_results_by_call_order(context)
+
+    full_messages = context + messages
+
+    prompt = bos_token if add_default_bos_token and len(context) == 0 else ""
+
+    # Resolve drop_thinking: if any message has tools defined, don't drop thinking
+    effective_drop_thinking = drop_thinking
+    if any(m.get("tools") for m in full_messages):
+        effective_drop_thinking = False
+
+    if thinking_mode == "thinking" and effective_drop_thinking:
+        full_messages = _drop_thinking_messages(full_messages)
+        # After dropping, recalculate how many messages to render
+        # (context may have shrunk too)
+        num_to_render = len(full_messages) - len(_drop_thinking_messages(context))
+        context_len = len(full_messages) - num_to_render
+    else:
+        num_to_render = len(messages)
+        context_len = len(context)
+
+    for idx in range(num_to_render):
+        prompt += render_message(
+            idx + context_len,
+            full_messages,
+            thinking_mode=thinking_mode,
+            drop_thinking=effective_drop_thinking,
+            reasoning_effort=reasoning_effort,
+        )
+
+    return prompt
+
+
+def _drop_thinking_messages(messages: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
+    """
+    Drop reasoning_content and non-essential messages before the last user message.
+
+    Behavior:
+    - Messages with role in ["user", "system", "tool", "latest_reminder"] are always kept.
+    - Messages at or after the last user index are always kept.
+    - Assistant messages before the last user get reasoning_content removed.
+    - Developer messages before the last user are dropped entirely.
+    """
+    last_user_idx = find_last_user_index(messages)
+    result = []
+    keep_roles = {"user", "system", "tool", "latest_reminder", "direct_search_results"}
+
+    for idx, msg in enumerate(messages):
+        role = msg.get("role")
+        if role in keep_roles or idx >= last_user_idx:
+            result.append(msg)
+        elif role == "assistant":
+            msg = copy.copy(msg)
+            msg.pop("reasoning_content", None)
+            result.append(msg)
+        # developer and other roles before last_user_idx are dropped
+
+    return result
+
+
+# ============================================================
+# Parsing (Decoding model output)
+# ============================================================
+
+def _read_until_stop(index: int, text: str, stop: List[str]) -> Tuple[int, str, Optional[str]]:
+    """
+    Read text from index until one of the stop strings is found.
+
+    Returns:
+        Tuple of (new_index, content_before_stop, matched_stop_string_or_None).
+    """
+    min_pos = len(text)
+    matched_stop = None
+
+    for s in stop:
+        pos = text.find(s, index)
+        if pos != -1 and pos < min_pos:
+            min_pos = pos
+            matched_stop = s
+
+    if matched_stop:
+        content = text[index:min_pos]
+        return min_pos + len(matched_stop), content, matched_stop
+    else:
+        content = text[index:]
+        return len(text), content, None
+
+
+def parse_tool_calls(index: int, text: str) -> Tuple[int, Optional[str], List[Dict[str, str]]]:
+    """
+    Parse DSML tool calls from text starting at the given index.
+
+    Args:
+        index: Starting position in text.
+        text: The full text to parse.
+
+    Returns:
+        Tuple of (new_index, last_stop_token, list_of_tool_call_dicts).
+        Each tool call dict has "name" and "arguments" keys.
+    """
+    tool_calls: List[Dict[str, Any]] = []
+    stop_token = None
+    tool_calls_end_token = f"</{dsml_token}{tool_calls_block_name}>"
+
+    while index < len(text):
+        index, _, stop_token = _read_until_stop(index, text, [f"<{dsml_token}invoke", tool_calls_end_token])
+        if _ != ">\n":
+            raise ValueError(f"Tool call format error: expected '>\\n' but got '{_}'")
+
+        if stop_token == tool_calls_end_token:
+            break
+
+        if stop_token is None:
+            raise ValueError("Missing special token in tool calls")
+
+        index, tool_name_content, stop_token = _read_until_stop(index, text, [f"<{dsml_token}parameter", f"</{dsml_token}invoke"])
+
+        p_tool_name = re.findall(r'^\s*name="(.*?)">\n$', tool_name_content, flags=re.DOTALL)
+        if len(p_tool_name) != 1:
+            raise ValueError(f"Tool name format error: '{tool_name_content}'")
+        tool_name = p_tool_name[0]
+
+        tool_args: Dict[str, Tuple[str, str]] = {}
+        while stop_token == f"<{dsml_token}parameter":
+            index, param_content, stop_token = _read_until_stop(index, text, [f"/{dsml_token}parameter"])
+
+            param_kv = re.findall(r'^ name="(.*?)" string="(true|false)">(.*?)<$', param_content, flags=re.DOTALL)
+            if len(param_kv) != 1:
+                raise ValueError(f"Parameter format error: '{param_content}'")
+            param_name, string, param_value = param_kv[0]
+
+            if param_name in tool_args:
+                raise ValueError(f"Duplicate parameter name: '{param_name}'")
+            tool_args[param_name] = (param_value, string)
+
+            index, content, stop_token = _read_until_stop(index, text, [f"<{dsml_token}parameter", f"</{dsml_token}invoke"])
+            if content != ">\n":
+                raise ValueError(f"Parameter format error: expected '>\\n' but got '{content}'")
+
+        tool_call = decode_dsml_to_arguments(tool_name=tool_name, tool_args=tool_args)
+        tool_calls.append(tool_call)
+
+    return index, stop_token, tool_calls
+
+
+def parse_message_from_completion_text(text: str, thinking_mode: str) -> Dict[str, Any]:
+    """
+    Parse a model completion text into a structured assistant message.
+
+    This function takes the raw text output from the model (a single assistant turn)
+    and extracts:
+    - reasoning_content (thinking block)
+    - content (summary/response)
+    - tool_calls (if any)
+
+    NOTE: This function is designed to parse only correctly formatted strings and
+    will raise ValueError for malformed output.
+
+    Args:
+        text: The raw completion text (including EOS token).
+        thinking_mode: Either "chat" or "thinking".
+
+    Returns:
+        Dict with keys: "role", "content", "reasoning_content", "tool_calls".
+        tool_calls are in OpenAI format.
+    """
+    summary_content, reasoning_content, tool_calls = "", "", []
+    index, stop_token = 0, None
+    tool_calls_start_token = f"\n\n<{dsml_token}{tool_calls_block_name}"
+
+    is_thinking = thinking_mode == "thinking"
+    is_tool_calling = False
+
+    if is_thinking:
+        index, content_delta, stop_token = _read_until_stop(index, text, [thinking_end_token, tool_calls_start_token])
+        reasoning_content = content_delta
+        assert stop_token == thinking_end_token, "Invalid thinking format: missing </think>"
+
+    index, content_delta, stop_token = _read_until_stop(index, text, [eos_token, tool_calls_start_token])
+    summary_content = content_delta
+    if stop_token == tool_calls_start_token:
+        is_tool_calling = True
+    else:
+        assert stop_token == eos_token, "Invalid format: missing EOS token"
+
+    if is_tool_calling:
+        index, stop_token, tool_calls = parse_tool_calls(index, text)
+
+        index, tool_ends_text, stop_token = _read_until_stop(index, text, [eos_token])
+        assert not tool_ends_text, "Unexpected content after tool calls"
+
+    assert len(text) == index and stop_token in [eos_token, None], "Unexpected content at end"
+
+    for sp_token in [bos_token, eos_token, thinking_start_token, thinking_end_token, dsml_token]:
+        assert sp_token not in summary_content and sp_token not in reasoning_content, \
+            f"Unexpected special token '{sp_token}' in content"
+
+    return {
+        "role": "assistant",
+        "content": summary_content,
+        "reasoning_content": reasoning_content,
+        "tool_calls": tool_calls_to_openai_format(tool_calls)
+    }

inference/generate.py

+import os
+import json
+import sys
+from argparse import ArgumentParser
+from typing import List
+
+import torch
+import torch.distributed as dist
+from transformers import AutoTokenizer
+from safetensors.torch import load_model
+
+from model import Transformer, ModelArgs
+current_dir = os.path.dirname(os.path.abspath(__file__))
+encoding_dir = os.path.join(current_dir, '../encoding')
+sys.path.insert(0, os.path.abspath(encoding_dir))
+from encoding_dsv4 import encode_messages, parse_message_from_completion_text
+
+
+def sample(logits, temperature: float = 1.0):
+    """Gumbel-max trick: equivalent to multinomial sampling but faster on GPU,
+    since it avoids the GPU-to-CPU sync in torch.multinomial."""
+    logits = logits / max(temperature, 1e-5)
+    probs = torch.softmax(logits, dim=-1, dtype=torch.float32)
+    return probs.div_(torch.empty_like(probs).exponential_(1)).argmax(dim=-1)
+
+
+@torch.inference_mode()
+def generate(
+    model: Transformer,
+    prompt_tokens: List[List[int]],
+    max_new_tokens: int,
+    eos_id: int,
+    temperature: float = 1.0
+) -> List[List[int]]:
+    """Batch generation with left-padded prompts.
+
+    The first forward pass processes [min_prompt_len:] tokens (prefill phase).
+    Subsequent passes generate one token at a time (decode phase). For positions
+    still within a prompt, the ground-truth token overrides the model's prediction.
+    """
+    prompt_lens = [len(t) for t in prompt_tokens]
+    assert max(prompt_lens) <= model.max_seq_len, f"Prompt length exceeds model maximum sequence length (max_seq_len={model.max_seq_len})"
+    total_len = min(model.max_seq_len, max_new_tokens + max(prompt_lens))
+    tokens = torch.full((len(prompt_tokens), total_len), -1, dtype=torch.long)
+    for i, t in enumerate(prompt_tokens):
+        tokens[i, :len(t)] = torch.tensor(t, dtype=torch.long)
+    prev_pos = 0
+    finished = torch.tensor([False] * len(prompt_tokens))
+    prompt_mask = tokens != -1
+    for cur_pos in range(min(prompt_lens), total_len):
+        logits = model.forward(tokens[:, prev_pos:cur_pos], prev_pos)
+        if temperature > 0:
+            next_token = sample(logits, temperature)
+        else:
+            next_token = logits.argmax(dim=-1)
+        next_token = torch.where(prompt_mask[:, cur_pos], tokens[:, cur_pos], next_token)
+        tokens[:, cur_pos] = next_token
+        finished |= torch.logical_and(~prompt_mask[:, cur_pos], next_token == eos_id)
+        prev_pos = cur_pos
+        if finished.all():
+            break
+    completion_tokens = []
+    for i, toks in enumerate(tokens.tolist()):
+        toks = toks[prompt_lens[i]:prompt_lens[i]+max_new_tokens]
+        if eos_id in toks:
+            toks = toks[:toks.index(eos_id)]
+        toks.append(eos_id)
+        completion_tokens.append(toks)
+    return completion_tokens
+
+
+def main(
+    ckpt_path: str,
+    config: str,
+    input_file: str = "",
+    interactive: bool = True,
+    max_new_tokens: int = 100,
+    temperature: float = 1.0,
+) -> None:
+    world_size = int(os.getenv("WORLD_SIZE", "1"))
+    rank = int(os.getenv("RANK", "0"))
+    local_rank = int(os.getenv("LOCAL_RANK", "0"))
+    if world_size > 1:
+        dist.init_process_group("nccl")
+    global print
+    if rank != 0:
+        print = lambda *_, **__: None
+    torch.cuda.set_device(local_rank)
+    torch.cuda.memory._set_allocator_settings("expandable_segments:True")
+    torch.set_default_dtype(torch.bfloat16)
+    torch.set_num_threads(8)
+    torch.manual_seed(33377335)
+    with open(config) as f:
+        args = ModelArgs(**json.load(f))
+    if interactive:
+        args.max_batch_size = 1
+    print(args)
+    with torch.device("cuda"):
+        model = Transformer(args)
+    tokenizer = AutoTokenizer.from_pretrained(ckpt_path)
+    print("load model")
+    load_model(model, os.path.join(ckpt_path, f"model{rank}-mp{world_size}.safetensors"), strict=False)
+    torch.set_default_device("cuda")
+    print("I'm DeepSeek 👋")
+
+    if interactive:
+        messages = []
+        while True:
+            if world_size == 1:
+                prompt = input(">>> ")
+            elif rank == 0:
+                prompt = input(">>> ")
+                objects = [prompt]
+                dist.broadcast_object_list(objects, 0)
+            else:
+                objects = [None]
+                dist.broadcast_object_list(objects, 0)
+                prompt = objects[0]
+            if prompt == "/exit":
+                break
+            elif prompt == "/clear":
+                messages.clear()
+                continue
+            messages.append({"role": "user", "content": prompt})
+            prompt_tokens = tokenizer.encode(encode_messages(messages, thinking_mode="chat"))
+            completion_tokens = generate(model, [prompt_tokens], max_new_tokens, tokenizer.eos_token_id, temperature)
+            completion = tokenizer.decode(completion_tokens[0])
+            print(completion)
+            messages.append(parse_message_from_completion_text(completion, thinking_mode="chat"))
+    else:
+        with open(input_file) as f:
+            prompts = f.read().split("\n\n")
+        prompt_tokens = [tokenizer.encode(encode_messages([{"role": "user", "content": prompt}], thinking_mode="chat")) for prompt in prompts]
+        completion_tokens = generate(model, prompt_tokens, max_new_tokens, tokenizer.eos_token_id, temperature)
+        completions = tokenizer.batch_decode(completion_tokens)
+        for prompt, completion in zip(prompts, completions):
+            print("Prompt:", prompt)
+            print("Completion:", completion)
+            print()
+
+    if world_size > 1:
+        dist.destroy_process_group()
+
+
+if __name__ == "__main__":
+    parser = ArgumentParser()
+    parser.add_argument("--ckpt-path", type=str, required=True)
+    parser.add_argument("--config", type=str, required=True)
+    parser.add_argument("--input-file", type=str, default="")
+    parser.add_argument("--interactive", action="store_true")
+    parser.add_argument("--max-new-tokens", type=int, default=300)
+    parser.add_argument("--temperature", type=float, default=0.6)
+    args = parser.parse_args()
+    assert args.input_file or args.interactive, "Either input-file or interactive mode must be specified"
+    main(args.ckpt_path, args.config, args.input_file, args.interactive, args.max_new_tokens, args.temperature)

inference/kernel.py

+import torch
+import tilelang
+import tilelang.language as T
+from typing import Tuple, Optional
+
+
+tilelang.set_log_level("WARNING")
+
+pass_configs = {
+    tilelang.PassConfigKey.TL_DISABLE_WARP_SPECIALIZED: True,
+    tilelang.PassConfigKey.TL_DISABLE_TMA_LOWER: True,
+    tilelang.PassConfigKey.TL_ENABLE_AGGRESSIVE_SHARED_MEMORY_MERGE: True,
+}
+
+FP8 = "float8_e4m3"
+FP4 = "float4_e2m1fn"
+# FE8M0 = "float8_e8m0fnu"
+BF16 = "bfloat16"
+FP32 = "float32"
+INT32 = "int32"
+
+
+def fast_log2_ceil(x):
+    """Compute ceil(log2(x)) via IEEE 754 bit manipulation. Avoids slow log/ceil intrinsics."""
+    bits_x = T.reinterpret("uint32", x)
+    exp_x = (bits_x >> 23) & 0xFF
+    man_bits = bits_x & ((1 << 23) - 1)
+    return T.Cast("int32", exp_x - 127 + T.if_then_else(man_bits != 0, 1, 0))
+
+
+def fast_pow2(x):
+    """Compute 2^x for integer x via IEEE 754 bit manipulation."""
+    bits_x = (x + 127) << 23
+    return T.reinterpret("float32", bits_x)
+
+
+def fast_round_scale(amax, fp8_max_inv):
+    return fast_pow2(fast_log2_ceil(amax * fp8_max_inv))
+
+
+@tilelang.jit(pass_configs=pass_configs)
+def act_quant_kernel(
+    N, block_size=128, in_dtype=BF16, out_dtype=FP8, scale_dtype=FP32,
+    round_scale=False, inplace=False
+):
+    """Block-wise FP8 quantization. inplace=True does fused quant+dequant back to BF16."""
+    M = T.symbolic("M")
+    fp8_min = -448.0
+    fp8_max = 448.0
+    fp8_max_inv = 1 / fp8_max
+    num_stages = 0 if round_scale or inplace else 2
+    blk_m = 32
+    group_size = block_size
+    # Internal computation in FP32; scale_dtype controls output storage format.
+    compute_dtype = FP32
+    out_dtype = in_dtype if inplace else out_dtype
+
+    @T.prim_func
+    def act_quant_kernel_(
+        X: T.Tensor[(M, N), in_dtype],
+        Y: T.Tensor[(M, N), out_dtype],
+        S: T.Tensor[(M, T.ceildiv(N, group_size)), scale_dtype],
+    ):
+        with T.Kernel(T.ceildiv(M, blk_m), T.ceildiv(N, group_size), threads=128) as (
+            pid_m,
+            pid_n,
+        ):
+            x_shared = T.alloc_shared((blk_m, group_size), in_dtype)
+            x_local = T.alloc_fragment((blk_m, group_size), in_dtype)
+            amax_local = T.alloc_fragment((blk_m,), compute_dtype)
+            s_local = T.alloc_fragment((blk_m,), compute_dtype)
+            y_local = T.alloc_fragment((blk_m, group_size), out_dtype)
+            y_shared = T.alloc_shared((blk_m, group_size), out_dtype)
+
+            for _ in T.Pipelined(1, num_stages=num_stages):
+                T.copy(X[pid_m * blk_m, pid_n * group_size], x_shared)
+                T.copy(x_shared, x_local)
+                T.reduce_absmax(x_local, amax_local, dim=1)
+                for i in T.Parallel(blk_m):
+                    amax_local[i] = T.max(amax_local[i], 1e-4)
+                    if round_scale:
+                        s_local[i] = fast_round_scale(amax_local[i], fp8_max_inv)
+                    else:
+                        s_local[i] = amax_local[i] * fp8_max_inv
+                if inplace:
+                    for i, j in T.Parallel(blk_m, group_size):
+                        y_local[i, j] = T.Cast(
+                            out_dtype,
+                            T.Cast(compute_dtype, T.Cast(out_dtype, T.clamp(
+                                x_local[i, j] / s_local[i], fp8_min, fp8_max
+                            ))) * s_local[i],
+                        )
+                else:
+                    for i, j in T.Parallel(blk_m, group_size):
+                        y_local[i, j] = T.clamp(
+                            x_local[i, j] / s_local[i], fp8_min, fp8_max
+                        )
+                for i in T.Parallel(blk_m):
+                    S[pid_m * blk_m + i, pid_n] = T.Cast(scale_dtype, s_local[i])
+                T.copy(y_local, y_shared)
+                T.copy(y_shared, Y[pid_m * blk_m, pid_n * group_size])
+
+    return act_quant_kernel_
+
+
+def act_quant(
+    x: torch.Tensor, block_size: int = 128, scale_fmt: Optional[str] = None,
+    scale_dtype: torch.dtype = torch.float32, inplace: bool = False,
+) -> torch.Tensor:
+    """Block-wise FP8 quantization. inplace=True does fused quant+dequant back to BF16.
+    When scale_fmt is set, scales are rounded to power-of-2 (MXFP)."""
+    N = x.size(-1)
+    assert N % block_size == 0
+    # tl_dtype = FE8M0 if scale_dtype == torch.float8_e8m0fnu else FP32
+    tl_dtype = FP32
+    z = x.contiguous()
+    y = torch.empty_like(z) if inplace else torch.empty_like(z, dtype=torch.float8_e4m3fn)
+    s = z.new_empty(*z.size()[:-1], N // block_size, dtype=scale_dtype)
+    kernel = act_quant_kernel(
+        N, block_size, scale_dtype=tl_dtype,
+        round_scale=scale_fmt is not None, inplace=inplace,
+    )
+    kernel(z.view(-1, N), y.view(-1, N), s.view(-1, N // block_size))
+    if inplace:
+        x.copy_(y)
+        return x
+    return y, s
+
+
+@tilelang.jit(pass_configs=pass_configs)
+def fp8_gemm_kernel(N, K, out_dtype=BF16, accum_dtype=FP32, scale_dtype=FP32):
+    assert out_dtype in [BF16, FP32]
+
+    M = T.symbolic("M")
+    group_size = 128
+    block_M = 32
+    block_N = 128
+    block_K = 128
+
+    @T.prim_func
+    def fp8_gemm_kernel_(
+        A: T.Tensor[(M, K), FP8],
+        B: T.Tensor[(N, K), FP8],
+        C: T.Tensor[(M, N), out_dtype],
+        scales_a: T.Tensor[(M, T.ceildiv(K, group_size)), scale_dtype],
+        scales_b: T.Tensor[(T.ceildiv(N, group_size), T.ceildiv(K, group_size)), scale_dtype],
+    ):
+        with T.Kernel(T.ceildiv(N, block_N), T.ceildiv(M, block_M), threads=128) as (
+            bx,
+            by,
+        ):
+            A_shared = T.alloc_shared((block_M, block_K), FP8)
+            B_shared = T.alloc_shared((block_N, block_K), FP8)
+            C_shared = T.alloc_shared((block_M, block_N), out_dtype)
+            Scale_C_shared = T.alloc_shared((block_M), FP32)
+            C_local = T.alloc_fragment((block_M, block_N), accum_dtype)
+            C_local_accum = T.alloc_fragment((block_M, block_N), accum_dtype)
+
+            # Improve L2 Cache
+            T.use_swizzle(panel_size=10)
+            T.clear(C_local)
+            T.clear(C_local_accum)
+
+            K_iters = T.ceildiv(K, block_K)
+            for k in T.Pipelined(K_iters, num_stages=2):
+                T.copy(A[by * block_M, k * block_K], A_shared)
+                T.copy(B[bx * block_N, k * block_K], B_shared)
+                # Cast scales to FP32 for computation; scales_b has one value per block_N group
+                Scale_B = T.Cast(FP32, scales_b[bx * block_N // group_size, k])
+                for i in T.Parallel(block_M):
+                    Scale_C_shared[i] = T.Cast(FP32, scales_a[by * block_M + i, k]) * Scale_B
+
+                T.gemm(A_shared, B_shared, C_local, transpose_B=True)
+                # Separate accumulator for scale-corrected results (2x accumulation precision)
+                for i, j in T.Parallel(block_M, block_N):
+                    C_local_accum[i, j] += C_local[i, j] * Scale_C_shared[i]
+                T.clear(C_local)
+            T.copy(C_local_accum, C_shared)
+            T.copy(C_shared, C[by * block_M, bx * block_N])
+
+    return fp8_gemm_kernel_
+
+
+def fp8_gemm(
+    a: torch.Tensor, a_s: torch.Tensor, b: torch.Tensor, b_s: torch.Tensor,
+    scale_dtype: torch.dtype = torch.float32,
+) -> torch.Tensor:
+    """C[M,N] = A[M,K] @ B[N,K]^T with per-128 block FP8 scaling on both A and B."""
+    assert a.is_contiguous() and b.is_contiguous(), "Input tensors must be contiguous"
+    assert a_s.is_contiguous() and b_s.is_contiguous(), (
+        "Scaling factor tensors must be contiguous"
+    )
+    # tl_dtype = FP32
+    tl_dtype = FE8M0 if scale_dtype == torch.float8_e8m0fnu else FP32
+    K = a.size(-1)
+    M = a.numel() // K
+    N = b.size(0)
+    c = a.new_empty(*a.size()[:-1], N, dtype=torch.get_default_dtype())
+    kernel = fp8_gemm_kernel(N, K, scale_dtype=tl_dtype)
+    kernel(a.view(M, K), b, c.view(M, N), a_s.view(M, -1), b_s)
+    return c
+
+
+@tilelang.jit(pass_configs=pass_configs)
+def sparse_attn_kernel(h_orig: int, d: int, scale=None):
+    """Sparse multi-head attention via index gathering + online softmax (FlashAttention-style).
+    For each (batch, seq_pos), gathers top-k KV positions by index, computes attention
+    with numerically stable running max/sum, and includes a learnable attn_sink bias."""
+    b = T.symbolic("b")
+    m = T.symbolic("m")
+    n = T.symbolic("n")
+    topk = T.symbolic("topk")
+    if scale is None:
+        scale = (1.0 / d) ** 0.5
+
+    num_stages = 0
+    threads = 256
+    block = 32
+    num_blocks = tilelang.cdiv(topk, block)
+
+    padded_H = max(tilelang.math.next_power_of_2(h_orig), 16)
+
+    max_block_m = 16
+    if h_orig > max_block_m:
+        assert h_orig % max_block_m == 0, f"h should be a multiple of {max_block_m}"
+        REPLICATE_H = h_orig // max_block_m
+    else:
+        REPLICATE_H = 1
+
+    h = padded_H if REPLICATE_H == 1 else max_block_m
+
+    @T.prim_func
+    def sparse_attn_kernel_(
+        q: T.Tensor[(b, m, h_orig, d), BF16],
+        kv: T.Tensor[(b, n, d), BF16],
+        o: T.Tensor[(b, m, h_orig, d), BF16],
+        attn_sink: T.Tensor[(h_orig,), FP32],
+        topk_idxs: T.Tensor[(b, m, topk), INT32],
+    ):
+        with T.Kernel(m * REPLICATE_H, b, threads=threads) as (bx, by):
+            q_shared = T.alloc_fragment((h, d), BF16)
+            kv_shared = T.alloc_shared((block, d), BF16)
+            # o_shared = T.alloc_shared((h, d), BF16)
+            acc_s_cast = T.alloc_shared((h, block), BF16)
+
+            idxs = T.alloc_fragment(block, INT32)
+            acc_s = T.alloc_fragment((h, block), FP32)
+            acc_o = T.alloc_fragment((h, d), FP32)
+            scores_max = T.alloc_fragment(h, FP32)
+            scores_max_prev = T.alloc_fragment(h, FP32)
+            scores_scale = T.alloc_fragment(h, FP32)
+            scores_sum = T.alloc_fragment(h, FP32)
+            sum_exp = T.alloc_fragment(h, FP32)
+
+            T.clear(acc_o)
+            T.clear(sum_exp)
+            T.fill(scores_max, -T.infinity(FP32))
+
+            s_i = bx if REPLICATE_H == 1 else (bx // REPLICATE_H)
+
+            H0 = (0 if REPLICATE_H == 1 else (bx % REPLICATE_H) * h)
+            H1 = H0 + h
+
+            T.copy(q[by, s_i, H0:H1, :], q_shared)
+
+            for t in T.Pipelined(num_blocks, num_stages=num_stages):
+                for i in T.Parallel(block):
+                    idxs[i] = T.if_then_else(t * block + i < topk, topk_idxs[by, s_i, t * block + i], -1)
+                for i, j in T.Parallel(block, d):
+                    kv_shared[i, j] = T.if_then_else(idxs[i] != -1, kv[by, idxs[i], j], 0)
+                for i, j in T.Parallel(h, block):
+                    acc_s[i, j] = T.if_then_else(idxs[j] != -1, 0, -T.infinity(FP32))
+                T.gemm(q_shared, kv_shared, acc_s, transpose_B=True, policy=T.GemmWarpPolicy.FullRow)
+                for i, j in T.Parallel(h, block):
+                    acc_s[i, j] *= scale
+                T.copy(scores_max, scores_max_prev)
+                T.reduce_max(acc_s, scores_max, dim=1, clear=False)
+                for i in T.Parallel(h):
+                    scores_scale[i] = T.exp(scores_max_prev[i] - scores_max[i])
+                for i, j in T.Parallel(h, block):
+                    acc_s[i, j] = T.exp(acc_s[i, j] - scores_max[i])
+                T.reduce_sum(acc_s, scores_sum, dim=1)
+                for i in T.Parallel(h):
+                    sum_exp[i] = sum_exp[i] * scores_scale[i] + scores_sum[i]
+                T.copy(acc_s, acc_s_cast)
+                for i, j in T.Parallel(h, d):
+                    acc_o[i, j] *= scores_scale[i]
+                T.gemm(acc_s_cast, kv_shared, acc_o, policy=T.GemmWarpPolicy.FullRow)
+
+            for i in T.Parallel(h):
+                sum_exp[i] += T.exp(attn_sink[i] - scores_max[i])
+            for i, j in T.Parallel(h, d):
+                acc_o[i, j] /= sum_exp[i]
+
+            o_shared = T.alloc_shared((h, d), BF16)
+            T.copy(acc_o, o_shared)
+            T.copy(o_shared, o[by, s_i, H0:H1, :])
+
+    return sparse_attn_kernel_
+
+
+def sparse_attn(
+    q: torch.Tensor, kv: torch.Tensor, attn_sink: torch.Tensor, topk_idxs: torch.Tensor, softmax_scale: float
+) -> torch.Tensor:
+    b, s, h, d = q.size()
+    # print(f"Teng {q.size()=}")
+    # Pad heads to 16 for kernel efficiency (stripped after)
+    if h < 16:
+        q = torch.cat([q, q.new_zeros(b, s, 16 - h, d)], dim=2)
+        attn_sink = torch.cat([attn_sink, attn_sink.new_zeros(16 - h)])
+    o = torch.empty_like(q)
+    kernel = sparse_attn_kernel(q.size(2), d, softmax_scale)
+    kernel(q, kv, o, attn_sink, topk_idxs)
+    if h < 16:
+        o = o.narrow(2, 0, h).contiguous()
+    return o
+
+
+@tilelang.jit(pass_configs=pass_configs)
+def hc_split_sinkhorn_kernel(hc: int, sinkhorn_iters: int, eps: float):
+    n = T.symbolic("n")
+    mix_hc = (2 + hc) * hc
+    threads = 64
+
+    @T.prim_func
+    def hc_split_sinkhorn_kernel_(
+        mixes: T.Tensor[(n, mix_hc), FP32],
+        hc_scale: T.Tensor[(3,), FP32],
+        hc_base: T.Tensor[(mix_hc,), FP32],
+        pre: T.Tensor[(n, hc), FP32],
+        post: T.Tensor[(n, hc), FP32],
+        comb: T.Tensor[(n, hc, hc), FP32],
+    ):
+        with T.Kernel(n, threads=threads) as i:
+            mixes_shared = T.alloc_shared(mix_hc, FP32)
+            comb_frag = T.alloc_fragment((hc, hc), FP32)
+            T.copy(mixes[i, :], mixes_shared)
+
+            for j in T.Parallel(hc):
+                pre[i, j] = T.sigmoid(mixes_shared[j] * hc_scale[0] + hc_base[j]) + eps
+            for j in T.Parallel(hc):
+                post[i, j] = 2 * T.sigmoid(mixes_shared[j + hc] * hc_scale[1] + hc_base[j + hc])
+            for j, k in T.Parallel(hc, hc):
+                comb_frag[j, k] = mixes_shared[j * hc + k + hc * 2] * hc_scale[2] + hc_base[j * hc + k + hc * 2]
+
+            row_sum = T.alloc_fragment(hc, FP32)
+            col_sum = T.alloc_fragment(hc, FP32)
+
+            # comb = comb.softmax(-1) + eps
+            row_max = T.alloc_fragment(hc, FP32)
+            T.reduce_max(comb_frag, row_max, dim=1)
+            for j, k in T.Parallel(hc, hc):
+                comb_frag[j, k] = T.exp(comb_frag[j, k] - row_max[j])
+            T.reduce_sum(comb_frag, row_sum, dim=1)
+            for j, k in T.Parallel(hc, hc):
+                comb_frag[j, k] = comb_frag[j, k] / row_sum[j] + eps
+
+            # comb = comb / (comb.sum(-2) + eps)
+            T.reduce_sum(comb_frag, col_sum, dim=0)
+            for j, k in T.Parallel(hc, hc):
+                comb_frag[j, k] = comb_frag[j, k] / (col_sum[k] + eps)
+
+            for _ in T.serial(sinkhorn_iters - 1):
+                # comb = comb / (comb.sum(-1) + eps)
+                T.reduce_sum(comb_frag, row_sum, dim=1)
+                for j, k in T.Parallel(hc, hc):
+                    comb_frag[j, k] = comb_frag[j, k] / (row_sum[j] + eps)
+                # comb = comb / (comb.sum(-2) + eps)
+                T.reduce_sum(comb_frag, col_sum, dim=0)
+                for j, k in T.Parallel(hc, hc):
+                    comb_frag[j, k] = comb_frag[j, k] / (col_sum[k] + eps)
+
+            T.copy(comb_frag, comb[i, :, :])
+
+    return hc_split_sinkhorn_kernel_
+
+
+def hc_split_sinkhorn(mixes: torch.Tensor, hc_scale: torch.Tensor, hc_base: torch.Tensor, hc_mult: int = 4, sinkhorn_iters: int = 20, eps: float = 1e-6):
+    b, s, _ = mixes.size()
+    pre = mixes.new_empty(b, s, hc_mult)
+    post = mixes.new_empty(b, s, hc_mult)
+    comb = mixes.new_empty(b, s, hc_mult, hc_mult)
+    kernel = hc_split_sinkhorn_kernel(hc_mult, sinkhorn_iters, eps)
+    kernel(mixes.view(-1, (2 + hc_mult) * hc_mult), hc_scale, hc_base,
+           pre.view(-1, hc_mult), post.view(-1, hc_mult), comb.view(-1, hc_mult, hc_mult))
+    return pre, post, comb

inference/model.py

+import math
+from dataclasses import dataclass
+from typing import Tuple, Optional, Literal
+from functools import lru_cache
+from contextlib import contextmanager
+
+import torch
+from torch import nn
+import torch.nn.functional as F
+import torch.distributed as dist
+
+# from kernel import act_quant, fp4_act_quant, fp8_gemm, fp4_gemm, sparse_attn, hc_split_sinkhorn
+from kernel import sparse_attn, hc_split_sinkhorn
+
+try:
+    from scipy.linalg import hadamard
+except ImportError:
+    hadamard = None
+
+
+world_size = 1
+rank = 0
+block_size = 128
+fp4_block_size = 32
+default_dtype = torch.bfloat16
+scale_fmt = None
+scale_dtype = torch.float32
+
+
+@contextmanager
+def set_dtype(dtype):
+    """Temporarily override torch default dtype, restoring it on exit (even if an exception occurs)."""
+    prev = torch.get_default_dtype()
+    torch.set_default_dtype(dtype)
+    try:
+        yield
+    finally:
+        torch.set_default_dtype(prev)
+
+@dataclass
+class ModelArgs:
+    """Model hyperparameters. Field names match the config JSON keys."""
+    max_batch_size: int = 4
+    max_seq_len: int = 4096
+    dtype: Literal["bf16", "fp8"] = "fp8"
+    scale_fmt: Literal[None, "ue8m0"] = "ue8m0"
+    expert_dtype: Literal[None, "fp4", "fp8"] = None
+    scale_dtype: Literal["fp32", "fp8"] = "fp8"
+    vocab_size: int = 129280
+    dim: int = 4096
+    moe_inter_dim: int = 4096
+    n_layers: int = 7
+    n_hash_layers: int = 0
+    n_mtp_layers: int = 1
+    n_heads: int = 64
+    # moe
+    n_routed_experts: int = 8
+    n_shared_experts: int = 1
+    n_activated_experts: int = 2
+    score_func: Literal["softmax", "sigmoid", "sqrtsoftplus"] = "sqrtsoftplus"
+    route_scale: float = 1.
+    swiglu_limit: float = 0.
+    # mqa
+    q_lora_rank: int = 1024
+    head_dim: int = 512
+    rope_head_dim: int = 64
+    norm_eps: float = 1e-6
+    o_groups: int = 8
+    o_lora_rank: int = 1024
+    window_size: int = 128
+    compress_ratios: Tuple[int] = (0, 0, 4, 128, 4, 128, 4, 0)
+    # yarn
+    compress_rope_theta: float = 40000.0
+    original_seq_len: int = 0
+    rope_theta: float = 10000.0
+    rope_factor: float = 40
+    beta_fast: int = 32
+    beta_slow: int = 1
+    # index
+    index_n_heads: int = 64
+    index_head_dim: int = 128
+    index_topk: int = 512
+    # hc
+    hc_mult: int = 4
+    hc_sinkhorn_iters: int = 20
+    hc_eps: float = 1e-6
+
+
+class ParallelEmbedding(nn.Module):
+    """Embedding sharded along the vocab dimension. Each rank holds vocab_size // world_size rows.
+    Out-of-range indices are zero-masked before all_reduce to combine partial embeddings."""
+    def __init__(self, vocab_size: int, dim: int):
+        super().__init__()
+        self.vocab_size = vocab_size
+        self.dim = dim
+        assert vocab_size % world_size == 0, f"Vocabulary size must be divisible by world size (world_size={world_size})"
+        self.part_vocab_size = (vocab_size // world_size)
+        self.vocab_start_idx = rank * self.part_vocab_size
+        self.vocab_end_idx = self.vocab_start_idx + self.part_vocab_size
+        self.weight = nn.Parameter(torch.empty(self.part_vocab_size, self.dim))
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if world_size > 1:
+            mask = (x < self.vocab_start_idx) | (x >= self.vocab_end_idx)
+            x = x - self.vocab_start_idx
+            x[mask] = 0
+        y = F.embedding(x, self.weight)
+        if world_size > 1:
+            y[mask] = 0
+            dist.all_reduce(y)
+        return y
+
+
+def linear(x: torch.Tensor, weight: torch.Tensor, bias: Optional[torch.Tensor] = None) -> torch.Tensor:
+    """Dispatches to fp4_gemm / fp8_gemm / F.linear based on weight dtype.
+    For quantized weights, x is first quantized to FP8 via act_quant."""
+    assert bias is None
+    return F.linear(x, weight)
+
+
+class Linear(nn.Module):
+    """Linear layer supporting BF16, FP8, and FP4 weight formats with per-block scaling."""
+
+    def __init__(self, in_features: int, out_features: int, bias: bool = False, dtype = None):
+        super().__init__()
+        self.in_features = in_features
+        self.out_features = out_features
+        dtype = dtype or default_dtype
+        if dtype == torch.float4_e2m1fn_x2:
+            # FP4: weight is [out, in//2] in float4_e2m1fn_x2, logically [out, in] in fp4
+            # Scale is [out, in//32] in float8_e8m0fnu (1 scale per 32 fp4 elements along K)
+            self.weight = nn.Parameter(torch.empty(out_features, in_features // 2, dtype=torch.float4_e2m1fn_x2))
+            scale_out_features = out_features
+            scale_in_features = in_features // fp4_block_size
+            self.weight.scale = self.scale = nn.Parameter(torch.empty(scale_out_features, scale_in_features, dtype=torch.float8_e8m0fnu))
+        elif dtype == torch.float8_e4m3fn:
+            self.weight = nn.Parameter(torch.empty(out_features, in_features, dtype=dtype))
+            scale_out_features = (out_features + block_size - 1) // block_size
+            scale_in_features = (in_features + block_size - 1) // block_size
+            self.weight.scale = self.scale = nn.Parameter(torch.empty(scale_out_features, scale_in_features, dtype=torch.float8_e8m0fnu))
+        else:
+            self.weight = nn.Parameter(torch.empty(out_features, in_features, dtype=dtype))
+            self.register_parameter("scale", None)
+        if bias:
+            self.bias = nn.Parameter(torch.empty(out_features))
+        else:
+            self.register_parameter("bias", None)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        return linear(x, self.weight, self.bias)
+
+
+class ColumnParallelLinear(Linear):
+    """Shards output dim across TP ranks. No all-reduce needed on output."""
+    def __init__(self, in_features: int, out_features: int, bias: bool = False, dtype = None):
+        assert out_features % world_size == 0, f"Output features must be divisible by world size (world_size={world_size})"
+        self.part_out_features = out_features // world_size
+        super().__init__(in_features, self.part_out_features, bias, dtype)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        return linear(x, self.weight, self.bias)
+
+
+class RowParallelLinear(Linear):
+    """Shards input dim across TP ranks. All-reduce on output to sum partial results."""
+    def __init__(self, in_features: int, out_features: int, bias: bool = False, dtype = None):
+        assert in_features % world_size == 0, f"Input features must be divisible by world size (world_size={world_size})"
+        self.part_in_features = in_features // world_size
+        super().__init__(self.part_in_features, out_features, bias, dtype)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        y = linear(x, self.weight, None)
+        if world_size > 1:
+            y = y.float()
+            dist.all_reduce(y)
+        if self.bias is not None:
+            y += self.bias
+        return y.type_as(x)
+
+
+class RMSNorm(nn.Module):
+    def __init__(self, dim: int, eps: float = 1e-6):
+        super().__init__()
+        self.dim = dim
+        self.eps = eps
+        # rmsnorm in the checkpoint is stored in bf16, while the parameter here is stored in fp32 for convenient.
+        self.weight = nn.Parameter(torch.ones(dim, dtype=torch.float32))
+
+    def forward(self, x: torch.Tensor):
+        dtype = x.dtype
+        x = x.float()
+        var = x.square().mean(-1, keepdim=True)
+        x = x * torch.rsqrt(var + self.eps)
+        return (self.weight * x).to(dtype)
+
+
+@lru_cache(2)
+def precompute_freqs_cis(dim, seqlen, original_seq_len, base, factor, beta_fast, beta_slow) -> torch.Tensor:
+    """Precomputes complex exponentials for rotary embeddings with YaRN scaling.
+    When original_seq_len > 0, applies frequency interpolation with a smooth
+    linear ramp between beta_fast and beta_slow correction ranges."""
+
+    def find_correction_dim(num_rotations, dim, base, max_seq_len):
+        return dim * math.log(max_seq_len / (num_rotations * 2 * math.pi)) / (2 * math.log(base))
+
+    def find_correction_range(low_rot, high_rot, dim, base, max_seq_len):
+        low = math.floor(find_correction_dim(low_rot, dim, base, max_seq_len))
+        high = math.ceil(find_correction_dim(high_rot, dim, base, max_seq_len))
+        return max(low, 0), min(high, dim-1)
+
+    def linear_ramp_factor(min, max, dim):
+        if min == max:
+            max += 0.001
+        linear_func = (torch.arange(dim, dtype=torch.float32) - min) / (max - min)
+        ramp_func = torch.clamp(linear_func, 0, 1)
+        return ramp_func
+
+    freqs = 1.0 / (base ** (torch.arange(0, dim, 2, dtype=torch.float32) / dim))
+    if original_seq_len > 0:
+        low, high = find_correction_range(beta_fast, beta_slow, dim, base, original_seq_len)
+        smooth = 1 - linear_ramp_factor(low, high, dim // 2)
+        freqs = freqs / factor * (1 - smooth) + freqs * smooth
+
+    t = torch.arange(seqlen)
+    freqs = torch.outer(t, freqs)
+    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)
+    return freqs_cis
+
+
+def apply_rotary_emb(x: torch.Tensor, freqs_cis: torch.Tensor, inverse: bool = False) -> torch.Tensor:
+    """Applies rotary positional embeddings in-place. Uses conjugate for inverse (de-rotation)."""
+    y = x
+    x = torch.view_as_complex(x.float().unflatten(-1, (-1, 2)))
+    if inverse:
+        freqs_cis = freqs_cis.conj()
+    if x.ndim == 3:
+        freqs_cis = freqs_cis.view(1, x.size(1), x.size(-1))
+    else:
+        freqs_cis = freqs_cis.view(1, x.size(1), 1, x.size(-1))
+    x = torch.view_as_real(x * freqs_cis).flatten(-2)
+    y.copy_(x)
+    return y
+
+
+def hadamard_transform_ref(x, scale=1.0):
+    """
+    x: (..., dim)
+    out: (..., dim)
+    """
+    if hadamard is None:
+        raise ImportError("Please install scipy")
+    x_shape = x.shape
+    dim = x.shape[-1]
+    x = x.reshape(-1, dim)
+    log_dim = math.ceil(math.log2(dim))
+    dim_padded = 2 ** log_dim
+    if dim != dim_padded:
+        x = F.pad(x, (0, dim_padded - dim))
+    out = F.linear(x, torch.tensor(hadamard(dim_padded, dtype=float), dtype=x.dtype, device=x.device))
+    out = out * scale
+    return out[..., :dim].reshape(*x_shape)
+
+def rotate_activation(x: torch.Tensor) -> torch.Tensor:
+    """Applies randomized Hadamard rotation to spread information across dims before FP8 quant."""
+    assert x.dtype == torch.bfloat16
+    # from fast_hadamard_transform import hadamard_transform
+    return hadamard_transform_ref(x, scale=x.size(-1) ** -0.5)
+
+
+@lru_cache(1)
+def get_window_topk_idxs(window_size: int, bsz: int, seqlen: int, start_pos: int):
+    if start_pos >= window_size - 1:
+        start_pos %= window_size
+        matrix = torch.cat([torch.arange(start_pos + 1, window_size),  torch.arange(0, start_pos + 1)], dim=0)
+    elif start_pos > 0:
+        matrix = F.pad(torch.arange(start_pos + 1), (0, window_size - start_pos - 1), value=-1)
+    else:
+        base = torch.arange(seqlen).unsqueeze(1)
+        matrix = (base - window_size + 1).clamp(0) + torch.arange(min(seqlen, window_size))
+        matrix = torch.where(matrix > base, -1, matrix)
+    return matrix.unsqueeze(0).expand(bsz, -1, -1)
+
+
+@lru_cache(2)
+def get_compress_topk_idxs(ratio: int, bsz: int, seqlen: int, start_pos: int, offset: int):
+    if start_pos > 0:
+        matrix = torch.arange(0, (start_pos + 1) // ratio) + offset
+    else:
+        matrix = torch.arange(seqlen // ratio).repeat(seqlen, 1)
+        mask = matrix >= torch.arange(1, seqlen + 1).unsqueeze(1) // ratio
+        matrix = torch.where(mask, -1, matrix + offset)
+    return matrix.unsqueeze(0).expand(bsz, -1, -1)
+
+
+class Compressor(nn.Module):
+    """Compresses KV cache via learned gated pooling over `compress_ratio` consecutive tokens.
+    When overlap=True (ratio==4), uses overlapping windows for smoother compression boundaries."""
+
+    def __init__(self, args: ModelArgs, compress_ratio: int = 4, head_dim: int = 512, rotate: bool = False):
+        super().__init__()
+        self.dim = args.dim
+        self.head_dim = head_dim
+        self.rope_head_dim = args.rope_head_dim
+        self.nope_head_dim = head_dim - args.rope_head_dim
+        self.compress_ratio = compress_ratio
+        self.overlap = compress_ratio == 4
+        self.rotate = rotate
+        coff = 1 + self.overlap
+
+        self.ape = nn.Parameter(torch.empty(compress_ratio, coff * self.head_dim, dtype=torch.float32))
+        # wkv and wgate in the checkpoint is stored in bf16, while the parameter here is stored in fp32 for convenient.
+        # When overlap, the first half of dims is for overlapping compression, second half for normal.
+        self.wkv = Linear(self.dim, coff * self.head_dim, dtype=torch.float32)
+        self.wgate = Linear(self.dim, coff * self.head_dim, dtype=torch.float32)
+        self.norm = RMSNorm(self.head_dim, args.norm_eps)
+        self.kv_cache: torch.Tensor = None  # assigned lazily from Attention.kv_cache
+        # State buffers for decode-phase incremental compression.
+        # With overlap: state[:, :ratio] = overlapping window, state[:, ratio:] = current window.
+        self.register_buffer("kv_state", torch.zeros(args.max_batch_size, coff * compress_ratio, coff * self.head_dim, dtype=torch.float32), persistent=False)
+        self.register_buffer("score_state", torch.full((args.max_batch_size, coff * compress_ratio, coff * self.head_dim), float("-inf"), dtype=torch.float32), persistent=False)
+        self.freqs_cis: torch.Tensor = None
+
+    def overlap_transform(self, tensor: torch.Tensor, value=0):
+        # tensor: [b,s,r,2d]
+        b, s, _, _ = tensor.size()
+        ratio, d = self.compress_ratio, self.head_dim
+        new_tensor = tensor.new_full((b, s, 2 * ratio, d), value)
+        new_tensor[:, :, ratio:] = tensor[:, :, :, d:]
+        new_tensor[:, 1:, :ratio] = tensor[:, :-1, :, :d]
+        return new_tensor
+
+    def forward(self, x: torch.Tensor, start_pos: int):
+        assert self.kv_cache is not None
+        bsz, seqlen, _ = x.size()
+        ratio, overlap, d, rd = self.compress_ratio, self.overlap, self.head_dim, self.rope_head_dim
+        dtype = x.dtype
+        # compression need fp32
+        x = x.float()
+        kv = self.wkv(x)
+        score = self.wgate(x)
+        if start_pos == 0:
+            should_compress = seqlen >= ratio
+            remainder = seqlen % ratio
+            cutoff = seqlen - remainder
+            offset = ratio if overlap else 0
+            if overlap and cutoff >= ratio:
+                self.kv_state[:bsz, :ratio] = kv[:, cutoff-ratio : cutoff]
+                self.score_state[:bsz, :ratio] = score[:, cutoff-ratio : cutoff] + self.ape
+            if remainder > 0:
+                kv, self.kv_state[:bsz, offset : offset+remainder] = kv.split([cutoff, remainder], dim=1)
+                self.score_state[:bsz, offset : offset+remainder] = score[:, cutoff:] + self.ape[:remainder]
+                score = score[:, :cutoff]
+            kv = kv.unflatten(1, (-1, ratio))
+            score = score.unflatten(1, (-1, ratio)) + self.ape
+            if overlap:
+                kv = self.overlap_transform(kv, 0)
+                score = self.overlap_transform(score, float("-inf"))
+            kv = (kv * score.softmax(dim=2)).sum(dim=2)
+        else:
+            should_compress = (start_pos + 1) % self.compress_ratio == 0
+            score += self.ape[start_pos % ratio]
+            if overlap:
+                self.kv_state[:bsz, ratio + start_pos % ratio] = kv.squeeze(1)
+                self.score_state[:bsz, ratio + start_pos % ratio] = score.squeeze(1)
+                if should_compress:
+                    kv_state = torch.cat([self.kv_state[:bsz, :ratio, :d], self.kv_state[:bsz, ratio:, d:]], dim=1)
+                    score_state = torch.cat([self.score_state[:bsz, :ratio, :d], self.score_state[:bsz, ratio:, d:]], dim=1)
+                    kv = (kv_state * score_state.softmax(dim=1)).sum(dim=1, keepdim=True)
+                    self.kv_state[:bsz, :ratio] = self.kv_state[:bsz, ratio:]
+                    self.score_state[:bsz, :ratio] = self.score_state[:bsz, ratio:]
+            else:
+                self.kv_state[:bsz, start_pos % ratio] = kv.squeeze(1)
+                self.score_state[:bsz, start_pos % ratio] = score.squeeze(1)
+                if should_compress:
+                    kv = (self.kv_state[:bsz] * self.score_state[:bsz].softmax(dim=1)).sum(dim=1, keepdim=True)
+        if not should_compress:
+            return
+        kv = self.norm(kv.to(dtype))
+        if start_pos == 0:
+            freqs_cis = self.freqs_cis[:cutoff:ratio]
+        else:
+            freqs_cis = self.freqs_cis[start_pos + 1 - self.compress_ratio].unsqueeze(0)
+        apply_rotary_emb(kv[..., -rd:], freqs_cis)
+        if self.rotate:
+            kv = rotate_activation(kv)
+            # fp4_act_quant(kv, fp4_block_size, True)
+        # else:
+            # act_quant(kv[..., :-rd], 64, scale_fmt, scale_dtype, True)
+        if start_pos == 0:
+            self.kv_cache[:bsz, :seqlen // ratio] = kv
+        else:
+            self.kv_cache[:bsz, start_pos // ratio] = kv.squeeze(1)
+        return kv
+
+
+class Indexer(torch.nn.Module):
+    """Selects top-k compressed KV positions for sparse attention via learned scoring.
+    Has its own Compressor (with Hadamard rotation) to build compressed KV for scoring."""
+
+    def __init__(self, args: ModelArgs, compress_ratio: int = 4):
+        super().__init__()
+        self.dim = args.dim
+        self.n_heads = args.index_n_heads
+        self.n_local_heads = args.index_n_heads // world_size
+        self.head_dim = args.index_head_dim
+        self.rope_head_dim = args.rope_head_dim
+        self.index_topk = args.index_topk
+        self.q_lora_rank = args.q_lora_rank
+        self.wq_b = ColumnParallelLinear(self.q_lora_rank, self.n_heads * self.head_dim, dtype=torch.bfloat16)
+        self.weights_proj = ColumnParallelLinear(self.dim, self.n_heads, dtype=torch.bfloat16)
+        self.softmax_scale = self.head_dim ** -0.5
+        self.compress_ratio = compress_ratio
+
+        self.compressor = Compressor(args, compress_ratio, self.head_dim, True)
+        self.register_buffer("kv_cache", torch.zeros(args.max_batch_size, args.max_seq_len // compress_ratio, self.head_dim), persistent=False)
+        self.freqs_cis = None
+
+    def forward(self, x: torch.Tensor, qr: torch.Tensor, start_pos: int, offset: int):
+        bsz, seqlen, _ = x.size()
+        freqs_cis = self.freqs_cis[start_pos:start_pos+seqlen]
+        ratio = self.compress_ratio
+        rd = self.rope_head_dim
+        end_pos = start_pos + seqlen
+        if self.compressor.kv_cache is None:
+            self.compressor.kv_cache = self.kv_cache
+            self.compressor.freqs_cis = self.freqs_cis
+        q = self.wq_b(qr)
+        q = q.unflatten(-1, (self.n_local_heads, self.head_dim))
+        apply_rotary_emb(q[..., -rd:], freqs_cis)
+        q = rotate_activation(q)
+        # use fp4 simulation for q and kv in indexer
+        # fp4_act_quant(q, fp4_block_size, True)
+        self.compressor(x, start_pos)
+        weights = self.weights_proj(x) * (self.softmax_scale * self.n_heads ** -0.5)
+        # We performed QAT here, kv could also use fp8 format, though current implementation uses bf16
+        index_score = torch.einsum("bshd,btd->bsht", q, self.kv_cache[:bsz, :end_pos // ratio])
+        index_score = (index_score.relu_() * weights.unsqueeze(-1)).sum(dim=2)
+        if world_size > 1:
+            dist.all_reduce(index_score)
+        if start_pos == 0:
+            mask = torch.arange(seqlen // ratio).repeat(seqlen, 1) >= torch.arange(1, seqlen + 1).unsqueeze(1) // ratio
+            index_score += torch.where(mask, float("-inf"), 0)
+        topk_idxs = index_score.topk(min(self.index_topk, end_pos // ratio), dim=-1)[1]
+        if start_pos == 0:
+            mask = topk_idxs >= torch.arange(1, seqlen + 1).unsqueeze(1) // ratio
+            topk_idxs = torch.where(mask, -1, topk_idxs + offset)
+        else:
+            topk_idxs += offset
+        return topk_idxs
+
+
+class Attention(nn.Module):
+    """Multi-head Latent Attention (MLA) with sliding window + optional KV compression.
+    Uses low-rank Q projection (wq_a -> q_norm -> wq_b) and grouped low-rank O projection."""
+    def __init__(self, layer_id: int, args: ModelArgs):
+        super().__init__()
+        self.layer_id = layer_id
+        self.dim = args.dim
+        self.n_heads = args.n_heads
+        self.n_local_heads = args.n_heads // world_size
+        self.q_lora_rank = args.q_lora_rank
+        self.o_lora_rank = args.o_lora_rank
+        self.head_dim = args.head_dim
+        self.rope_head_dim = args.rope_head_dim
+        self.nope_head_dim = args.head_dim - args.rope_head_dim
+        self.n_groups = args.o_groups
+        self.n_local_groups = self.n_groups // world_size
+        self.window_size = args.window_size
+        self.compress_ratio = args.compress_ratios[layer_id]
+        self.eps = args.norm_eps
+
+        self.attn_sink = nn.Parameter(torch.empty(self.n_local_heads, dtype=torch.float32))
+        self.wq_a = Linear(self.dim, self.q_lora_rank, dtype=torch.bfloat16)
+        self.q_norm = RMSNorm(self.q_lora_rank, self.eps)
+        self.wq_b = ColumnParallelLinear(self.q_lora_rank, self.n_heads * self.head_dim, dtype=torch.bfloat16)
+        self.wkv = Linear(self.dim, self.head_dim, dtype=torch.bfloat16)
+        self.kv_norm = RMSNorm(self.head_dim, self.eps)
+
+        self.wo_a = ColumnParallelLinear(self.n_heads * self.head_dim // self.n_groups, self.n_groups * args.o_lora_rank, dtype=torch.bfloat16)
+        self.wo_b = RowParallelLinear(self.n_groups * args.o_lora_rank, self.dim)
+
+        self.softmax_scale = self.head_dim ** -0.5
+
+        if self.compress_ratio:
+            self.compressor = Compressor(args, self.compress_ratio, self.head_dim)
+            if self.compress_ratio == 4:
+                self.indexer = Indexer(args, self.compress_ratio)
+            else:
+                self.indexer = None
+
+        kv_cache_size = args.window_size + (args.max_seq_len // self.compress_ratio if self.compress_ratio else 0)
+        self.register_buffer("kv_cache", torch.zeros(args.max_batch_size, kv_cache_size, self.head_dim), persistent=False)
+        if self.compress_ratio:
+            original_seq_len, rope_theta = args.original_seq_len, args.compress_rope_theta
+        else:
+            # disable YaRN and use base rope_theta in pure sliding-window attention
+            original_seq_len, rope_theta = 0, args.rope_theta
+        freqs_cis = precompute_freqs_cis(self.rope_head_dim, args.max_seq_len, original_seq_len,
+                                         rope_theta, args.rope_factor, args.beta_fast, args.beta_slow)
+        self.register_buffer("freqs_cis", freqs_cis, persistent=False)
+
+    def forward(self, x: torch.Tensor, start_pos: int):
+        bsz, seqlen, _ = x.size()
+        freqs_cis = self.freqs_cis[start_pos:start_pos+seqlen]
+        win = self.window_size
+        ratio = self.compress_ratio
+        rd = self.rope_head_dim
+        if self.compress_ratio and self.compressor.kv_cache is None:
+            self.compressor.kv_cache = self.kv_cache[:, win:]
+            self.compressor.freqs_cis = self.freqs_cis
+            if self.indexer is not None:
+                self.indexer.freqs_cis = self.freqs_cis
+        # q
+        qr = q = self.q_norm(self.wq_a(x))
+        q = self.wq_b(q).unflatten(-1, (self.n_local_heads, self.head_dim))
+        q *= torch.rsqrt(q.square().mean(-1, keepdim=True) + self.eps)
+        apply_rotary_emb(q[..., -rd:], freqs_cis)
+
+        # win kv & topk_idxs
+        kv = self.wkv(x)
+        kv = self.kv_norm(kv)
+        apply_rotary_emb(kv[..., -rd:], freqs_cis)
+        # FP8-simulate non-rope dims to match QAT; rope dims stay bf16 for positional precision
+        # act_quant(kv[..., :-rd], 64, scale_fmt, scale_dtype, True)
+        topk_idxs = get_window_topk_idxs(win, bsz, seqlen, start_pos)
+        if self.compress_ratio:
+            offset = kv.size(1) if start_pos == 0 else win
+            if self.indexer is not None:
+                compress_topk_idxs = self.indexer(x, qr, start_pos, offset)
+            else:
+                compress_topk_idxs = get_compress_topk_idxs(ratio, bsz, seqlen, start_pos, offset)
+            topk_idxs = torch.cat([topk_idxs, compress_topk_idxs], dim=-1)
+        topk_idxs = topk_idxs.int()
+
+        # compress kv & attn
+        if start_pos == 0:
+            if seqlen <= win:
+                self.kv_cache[:bsz, :seqlen] = kv
+            else:
+                cutoff = seqlen % win
+                self.kv_cache[:bsz, cutoff: win], self.kv_cache[:bsz, :cutoff] = kv[:, -win:].split([win - cutoff, cutoff], dim=1)
+            if self.compress_ratio:
+                if (kv_compress := self.compressor(x, start_pos)) is not None:
+                    kv = torch.cat([kv, kv_compress], dim=1)
+            # We performed QAT here, kv could also use fp8 format, though current implementation uses bf16
+            o = sparse_attn(q, kv, self.attn_sink, topk_idxs, self.softmax_scale)
+        else:
+            self.kv_cache[:bsz, start_pos % win] = kv.squeeze(1)
+            if self.compress_ratio:
+                self.compressor(x, start_pos)
+            o = sparse_attn(q, self.kv_cache[:bsz], self.attn_sink, topk_idxs, self.softmax_scale)
+        apply_rotary_emb(o[..., -rd:], freqs_cis, True)
+
+        # o
+        o = o.view(bsz, seqlen, self.n_local_groups, -1)
+        wo_a = self.wo_a.weight.view(self.n_local_groups, self.o_lora_rank, -1)
+        # NOTE: wo_a is FP8 in checkpoint; could do FP8 einsum here for better perf,
+        # but using BF16 for simplicity.
+        o = torch.einsum("bsgd,grd->bsgr", o, wo_a)
+        x = self.wo_b(o.flatten(2))
+        return x
+
+
+class Gate(nn.Module):
+    """MoE gating: computes expert routing scores and selects top-k experts.
+    Supports hash-based routing (first n_hash_layers) where expert indices are
+    predetermined per token ID, and score-based routing (remaining layers)."""
+    def __init__(self, layer_id: int, args: ModelArgs):
+        super().__init__()
+        self.dim = args.dim
+        self.topk = args.n_activated_experts
+        self.score_func = args.score_func
+        self.route_scale = args.route_scale
+        self.hash = layer_id < args.n_hash_layers
+        self.weight = nn.Parameter(torch.empty(args.n_routed_experts, args.dim))
+        if self.hash:
+            self.tid2eid = nn.Parameter(torch.empty(args.vocab_size, args.n_activated_experts, dtype=torch.int32), requires_grad=False)
+            self.bias = None
+        else:
+            self.bias = nn.Parameter(torch.empty(args.n_routed_experts, dtype=torch.float32))
+
+    def forward(self, x: torch.Tensor, input_ids: Optional[torch.Tensor] = None) -> Tuple[torch.Tensor, torch.Tensor]:
+        scores = linear(x.float(), self.weight.float())
+        if self.score_func == "softmax":
+            scores = scores.softmax(dim=-1)
+        elif self.score_func == "sigmoid":
+            scores = scores.sigmoid()
+        else:
+            scores = F.softplus(scores).sqrt()
+        original_scores = scores
+        # Bias shifts scores for expert selection (topk) but does not affect routing weights.
+        if self.bias is not None:
+            scores = scores + self.bias
+        if self.hash:
+            indices = self.tid2eid[input_ids]
+        else:
+            indices = scores.topk(self.topk, dim=-1)[1]
+        weights = original_scores.gather(1, indices)
+        if self.score_func != "softmax":
+            weights /= weights.sum(dim=-1, keepdim=True)
+        weights *= self.route_scale
+        return weights, indices
+
+
+class Expert(nn.Module):
+    """Single MoE expert: SwiGLU FFN (w1, w2, w3). Computation in float32 for stability."""
+    def __init__(self, dim: int, inter_dim: int, dtype=None, swiglu_limit=0):
+        super().__init__()
+        self.w1 = Linear(dim, inter_dim, dtype=dtype)
+        self.w2 = Linear(inter_dim, dim, dtype=dtype)
+        self.w3 = Linear(dim, inter_dim, dtype=dtype)
+        self.swiglu_limit = swiglu_limit
+
+    def forward(self, x: torch.Tensor, weights: Optional[torch.Tensor] = None) -> torch.Tensor:
+        dtype = x.dtype
+        gate = self.w1(x).float()
+        up = self.w3(x).float()
+        if self.swiglu_limit > 0:
+            up = torch.clamp(up, min=-self.swiglu_limit, max=self.swiglu_limit)
+            gate = torch.clamp(gate, max=self.swiglu_limit)
+        x = F.silu(gate) * up
+        if weights is not None:
+            x = weights * x
+        return self.w2(x.to(dtype))
+
+
+class MoE(nn.Module):
+    """Mixture-of-Experts: gate routes each token to top-k routed experts + 1 shared expert.
+    Experts are sharded across TP ranks; each rank handles n_routed_experts // world_size experts."""
+    def __init__(self, layer_id: int, args: ModelArgs):
+        super().__init__()
+        self.layer_id = layer_id
+        self.dim = args.dim
+        assert args.n_routed_experts % world_size == 0, f"Number of experts must be divisible by world size (world_size={world_size})"
+        self.n_routed_experts = args.n_routed_experts
+        self.n_local_experts = args.n_routed_experts // world_size
+        self.n_activated_experts = args.n_activated_experts
+        self.experts_start_idx = rank * self.n_local_experts
+        self.experts_end_idx = self.experts_start_idx + self.n_local_experts
+        self.gate = Gate(layer_id, args)
+        if args.expert_dtype == "fp4":
+            expert_dtype = torch.float4_e2m1fn_x2
+        elif args.expert_dtype == "fp8":
+            expert_dtype = torch.float8_e4m3fn
+        else:
+            None
+        # expert_dtype = torch.float4_e2m1fn_x2 if args.expert_dtype == "fp4" else None
+        self.experts = nn.ModuleList([Expert(args.dim, args.moe_inter_dim, dtype=torch.bfloat16, swiglu_limit=args.swiglu_limit) if self.experts_start_idx <= i < self.experts_end_idx else None
+                                       for i in range(self.n_routed_experts)])
+        assert args.n_shared_experts == 1
+        # no swiglu_limit
+        self.shared_experts = Expert(args.dim, args.moe_inter_dim, dtype=torch.bfloat16)
+
+    def forward(self, x: torch.Tensor, input_ids: torch.Tensor) -> torch.Tensor:
+        shape = x.size()
+        x = x.view(-1, self.dim)
+        weights, indices = self.gate(x, input_ids.flatten())
+        y = torch.zeros_like(x, dtype=torch.float32)
+        ####
+        torch.cuda.synchronize()
+        indices_cpu = indices.flatten().cpu()
+        counts_cpu = torch.bincount(indices_cpu, minlength=self.n_routed_experts)
+        counts = counts_cpu.cuda()
+        #counts = torch.bincount(indices.flatten(), minlength=self.n_routed_experts).tolist()
+        torch.cuda.synchronize()
+
+        for i in range(self.experts_start_idx, self.experts_end_idx):
+            if counts[i] == 0:
+                continue
+            expert = self.experts[i]
+            idx, top = torch.where(indices == i)
+            y[idx] += expert(x[idx], weights[idx, top, None])
+        if world_size > 1:
+            dist.all_reduce(y)
+        y += self.shared_experts(x)
+        return y.type_as(x).view(shape)
+
+
+class Block(nn.Module):
+    """Transformer block with Hyper-Connections (HC) mixing.
+    Instead of a simple residual, HC maintains `hc_mult` copies of the hidden state.
+    hc_pre: reduces hc copies -> 1 via learned weighted sum (pre-weights from Sinkhorn).
+    hc_post: expands 1 -> hc copies via learned post-weights + combination matrix."""
+    def __init__(self, layer_id: int, args: ModelArgs):
+        super().__init__()
+        self.layer_id = layer_id
+        self.norm_eps = args.norm_eps
+        self.attn = Attention(layer_id, args)
+        self.ffn = MoE(layer_id, args)
+        self.attn_norm = RMSNorm(args.dim, self.norm_eps)
+        self.ffn_norm = RMSNorm(args.dim, self.norm_eps)
+        self.hc_mult = hc_mult = args.hc_mult
+        self.hc_sinkhorn_iters = args.hc_sinkhorn_iters
+        self.hc_eps = args.hc_eps
+        mix_hc = (2 + hc_mult) * hc_mult
+        hc_dim = hc_mult * args.dim
+        with set_dtype(torch.float32):
+            self.hc_attn_fn = nn.Parameter(torch.empty(mix_hc, hc_dim))
+            self.hc_ffn_fn = nn.Parameter(torch.empty(mix_hc, hc_dim))
+            self.hc_attn_base = nn.Parameter(torch.empty(mix_hc))
+            self.hc_ffn_base = nn.Parameter(torch.empty(mix_hc))
+            self.hc_attn_scale = nn.Parameter(torch.empty(3))
+            self.hc_ffn_scale = nn.Parameter(torch.empty(3))
+
+    def hc_pre(self, x: torch.Tensor, hc_fn: torch.Tensor, hc_scale: torch.Tensor, hc_base: torch.Tensor):
+        # x: [b,s,hc,d], hc_fn: [mix_hc,hc*d], hc_scale: [3], hc_base: [mix_hc], y: [b,s,hc,d]
+        shape, dtype = x.size(), x.dtype
+        x = x.flatten(2).float()
+        rsqrt = torch.rsqrt(x.square().mean(-1, keepdim=True) + self.norm_eps)
+        mixes = F.linear(x, hc_fn) * rsqrt
+        pre, post, comb = hc_split_sinkhorn(mixes, hc_scale, hc_base, self.hc_mult, self.hc_sinkhorn_iters, self.hc_eps)
+        y = torch.sum(pre.unsqueeze(-1) * x.view(shape), dim=2)
+        return y.to(dtype), post, comb
+
+    def hc_post(self, x: torch.Tensor, residual: torch.Tensor, post: torch.Tensor, comb: torch.Tensor):
+        # x: [b,s,d], residual: [b,s,hc,d], post: [b,s,hc], comb: [b,s,hc,hc], y: [b,s,hc,d]
+        y = post.unsqueeze(-1) * x.unsqueeze(-2) + torch.sum(comb.unsqueeze(-1) * residual.unsqueeze(-2), dim=2)
+        return y.type_as(x)
+
+    def forward(self, x: torch.Tensor, start_pos: int, input_ids: Optional[torch.Tensor]) -> torch.Tensor:
+        residual = x
+        x, post, comb = self.hc_pre(x, self.hc_attn_fn, self.hc_attn_scale, self.hc_attn_base)
+        x = self.attn_norm(x)
+        x = self.attn(x, start_pos)
+        x = self.hc_post(x, residual, post, comb)
+
+        residual = x
+        x, post, comb = self.hc_pre(x, self.hc_ffn_fn, self.hc_ffn_scale, self.hc_ffn_base)
+        x = self.ffn_norm(x)
+        x = self.ffn(x, input_ids)
+        x = self.hc_post(x, residual, post, comb)
+        return x
+
+
+class ParallelHead(nn.Module):
+
+    def __init__(self, vocab_size: int, dim: int, norm_eps: float = 1e-6, hc_eps: float = 1e-6):
+        super().__init__()
+        self.vocab_size = vocab_size
+        self.dim = dim
+        self.norm_eps = norm_eps
+        self.hc_eps = hc_eps
+        self.part_vocab_size = (vocab_size // world_size)
+        # lm_head in the checkpoint is stored in bf16, while the parameter here is stored in fp32 for easier computation of logits later.
+        self.weight = nn.Parameter(torch.empty(self.part_vocab_size, self.dim, dtype=torch.float32))
+
+    def get_logits(self, x):
+        return F.linear(x[:, -1].float(), self.weight)
+
+    def forward(self, x: torch.Tensor, hc_fn: torch.Tensor, hc_scale: torch.Tensor, hc_base: torch.Tensor, norm: RMSNorm):
+        # x: [b,s,hc,d]
+        x = self.hc_head(x, hc_fn, hc_scale, hc_base)
+        logits = self.get_logits(norm(x))
+        if world_size > 1:
+            all_logits = [torch.empty_like(logits) for _ in range(world_size)]
+            dist.all_gather(all_logits, logits)
+            logits = torch.cat(all_logits, dim=-1)
+        return logits
+
+    def hc_head(self, x: torch.Tensor, hc_fn: torch.Tensor, hc_scale: torch.Tensor, hc_base: torch.Tensor):
+        shape, dtype = x.size(), x.dtype
+        x = x.flatten(2).float()
+        rsqrt = torch.rsqrt(x.square().mean(-1, keepdim=True) + self.norm_eps)
+        mixes = F.linear(x, hc_fn) * rsqrt
+        pre = torch.sigmoid(mixes * hc_scale + hc_base) + self.hc_eps
+        y = torch.sum(pre.unsqueeze(-1) * x.view(shape), dim=2)
+        return y.to(dtype)
+
+
+class MTPBlock(Block):
+
+    def __init__(self, layer_id: int, args: ModelArgs):
+        super().__init__(layer_id, args)
+        self.e_proj = Linear(args.dim, args.dim, dtype=torch.bfloat16)
+        self.h_proj = Linear(args.dim, args.dim, dtype=torch.bfloat16)
+        self.enorm = RMSNorm(args.dim, args.norm_eps)
+        self.hnorm = RMSNorm(args.dim, args.norm_eps)
+        self.norm = RMSNorm(args.dim, args.norm_eps)
+        self.hc_mult = hc_mult = args.hc_mult
+        hc_dim = hc_mult * args.dim
+        with set_dtype(torch.float32):
+            self.hc_head_fn = nn.Parameter(torch.empty(hc_mult, hc_dim))
+            self.hc_head_base = nn.Parameter(torch.empty(hc_mult))
+            self.hc_head_scale = nn.Parameter(torch.empty(1))
+        self.embed: ParallelEmbedding = None
+        self.head: ParallelHead = None
+
+    @torch.inference_mode()
+    def forward(self, x: torch.Tensor, start_pos: int, input_ids: torch.Tensor) -> torch.Tensor:
+        # x: [b,s,hc,d]
+        assert self.embed is not None and self.head is not None
+        e = self.embed(input_ids)
+        e = self.enorm(e)
+        x = self.hnorm(x)
+        x = self.e_proj(e).unsqueeze(2) + self.h_proj(x)
+        x = super().forward(x, start_pos, input_ids)
+        logits = self.head(x, self.hc_head_fn, self.hc_head_scale, self.hc_head_base, self.norm)
+        return logits
+
+
+class Transformer(nn.Module):
+    """Full DeepSeek-V4 model: embed -> HC-expand -> N blocks -> HC-head -> logits.
+    Sets global state (world_size, rank, default_dtype, scale_fmt, scale_dtype) in __init__."""
+    def __init__(self, args: ModelArgs):
+        # global world_size, rank, default_dtype, scale_fmt, scale_dtype
+        global world_size, rank
+        world_size = dist.get_world_size() if dist.is_initialized() else 1
+        rank = dist.get_rank() if dist.is_initialized() else 0
+        default_dtype = torch.float8_e4m3fn if args.dtype == "fp8" else torch.bfloat16
+        scale_fmt = "ue8m0" if args.scale_dtype == "fp8" else args.scale_fmt
+        scale_dtype = torch.float8_e8m0fnu if args.scale_dtype == "fp8" else torch.float32
+        super().__init__()
+        self.max_seq_len = args.max_seq_len
+        self.norm_eps = args.norm_eps
+        self.hc_eps = args.hc_eps
+        self.embed = ParallelEmbedding(args.vocab_size, args.dim)
+        self.layers = torch.nn.ModuleList()
+        for layer_id in range(args.n_layers):
+            self.layers.append(Block(layer_id, args))
+        self.norm = RMSNorm(args.dim, self.norm_eps)
+        self.head = ParallelHead(args.vocab_size, args.dim, self.norm_eps, self.hc_eps)
+        self.mtp = torch.nn.ModuleList()
+        for layer_id in range(args.n_mtp_layers):
+            self.mtp.append(MTPBlock(args.n_layers + layer_id, args))
+            self.mtp[-1].embed = self.embed
+            self.mtp[-1].head = self.head
+        self.hc_mult = hc_mult = args.hc_mult
+        hc_dim = hc_mult * args.dim
+        with set_dtype(torch.float32):
+            self.hc_head_fn = nn.Parameter(torch.empty(hc_mult, hc_dim))
+            self.hc_head_base = nn.Parameter(torch.empty(hc_mult))
+            self.hc_head_scale = nn.Parameter(torch.empty(1))
+
+    @torch.inference_mode()
+    def forward(self, input_ids: torch.Tensor, start_pos: int = 0):
+        h = self.embed(input_ids)
+        # Expand to hc_mult copies for Hyper-Connections
+        h = h.unsqueeze(2).repeat(1, 1, self.hc_mult, 1)
+        for layer in self.layers:
+            h = layer(h, start_pos, input_ids)
+        logits = self.head(h, self.hc_head_fn, self.hc_head_scale, self.hc_head_base, self.norm)
+        return logits
+
+
+if __name__ == "__main__":
+    torch.set_default_dtype(torch.bfloat16)
+    torch.set_default_device("cuda")
+    torch.manual_seed(0)
+    args = ModelArgs(n_hash_layers=0)
+    x = torch.randint(0, args.vocab_size, (2, 128))
+    model = Transformer(args)
+
+    print(model(x).size())
+    for i in range(128, 150):
+        print(i, model(x[:, 0:1], i).size())
+
+    h = torch.randn(2, 128, args.hc_mult, args.dim)
+    mtp = model.mtp[0]
+    print(mtp(h, 0, x).size())
+    print(mtp(h[:, 0:1], 1, x[:, 0:1]).size())

inference/start_torch.sh

+#!/usr/bin/env bash
+
+set -euo pipefail
+
+export NCCL_ALGO="Ring"
+export NCCL_PROTO="Simple"
+export MP=8
+export CONFIG="config.json"
+export CKPT_PATH="deepseek-ai/DeepSeek-V4-Flash-bf16-mp8"
+
+torchrun \
+  --nproc-per-node "${MP}" \
+  generate.py \
+  --ckpt-path "${CKPT_PATH}" \
+  --config "${CONFIG}" \
+  --interactive

model.properties

+# 模型唯一标识
+modelCode=2397
+# 模型名称
+modelName=DeepSeek-V4 
+# 模型描述
+modelDescription= DeepSeek-V4：迈向高效百万上下文智能。
+# 运行过程
+processType=推理
+# 算法类别
+appCategory=对话问答
+# 框架类型
+frameType=pytorch
+# 加速卡类型
+accelerateType=BW1100
\ No newline at end of file