Skip to content

GitLab

Commit 049925cc authored by Rayyyyy's avatar Rayyyyy
Browse files

add grop ande change to 24.04

parents
Pipeline #997 failed with stages
in 0 seconds
Hide whitespace changes
Inline Side-by-side
Showing with 3518 additions and 0 deletions
.gitignore 0 → 100644
View file @ 049925cc
checkpoints/*
!checkpoints/README.md
CODE_OF_CONDUCT.md 0 → 100644
View file @ 049925cc
Be excellent to each other.
Contributors.md 0 → 100644
View file @ 049925cc
# Contributors
None
LICENSE.txt 0 → 100644
View file @ 049925cc
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
README.md 0 → 100644
View file @ 049925cc
# Grok1
Grok-1是由xAI从头开始训练的3140亿个参数混合专家模型。
## 论文
暂无
## 模型结构
Grok-1 是一个8个专家（2个活跃）的混合专家（MoE）模型。
<div align=center>
<img src="./doc/MoE.png"/>
</div>
## 算法原理
Grok-1目前的设计规格如下:
- **参数:** 314B
- **体系结构:** Mixture of 8 Experts (MoE)
- **专家利用:** 2 experts used per token
- **层数:** 64
- **多头注意力:** 48 个 head 用于查询，8 个用于键 / 值（KV）
- **嵌入大小:** 6,144
- **Tokenization:** tokenizer 词汇大小为 131,072
- **附加特性:**
- Rotary embeddings (RoPE)
- Supports activation sharding and 8-bit 量化
- **最大序列长度(上下文):** 8,192 tokens
## 环境配置
### Docker（方法一）
-v 路径、docker_name和imageID根据实际情况修改
```bash
docker pull image.sourcefind.cn:5000/dcu/admin/base/jax:0.4.23-ubuntu20.04-dtk24.04-py310
docker run -it -v /path/your_code_data/:/path/your_code_data/ -v /opt/hyhal:/opt/hyhal:ro --shm-size=200G --privileged=true --device=/dev/kfd --device=/dev/dri/ --group-add video --name docker_name imageID bash
cd /your_code_path/grok-1_jax
pip install -r requirements.txt
```
### Dockerfile（方法二）
```bash
cd ./docker
docker build --no-cache -t grok1:latest .
docker run -it -v /path/your_code_data/:/path/your_code_data/ -v /opt/hyhal:/opt/hyhal:ro --shm-size=200G --privileged=true --device=/dev/kfd --device=/dev/dri/ --group-add video --name docker_name imageID bash
cd /your_code_path/grok-1_jax
```
### Anaconda（方法三）
关于本项目DCU显卡所需的特殊深度学习库可从[光合](https://developer.hpccube.com/tool/)开发者社区下载安装。
```
DTK驱动：dtk24.04
python：python3.10
jax: ≥ 0.4.23
gcc: ≥ 9.4.0
```
`Tips：以上dtk驱动、python等DCU相关工具版本需要严格一一对应`
其它非深度学习库参照requirements.txt安装：
```bash
pip install -r requirements.txt
```
## 数据集
暂无
## 训练
官方github未开源微调代码
## 推理
推理所需硬件最低配置参考：
device(s)：K100 64G 8卡
Driver version: 5.16.29.22.20
Vbios version: 5.223.001200k.548101
linux: ubuntu20.04
处理器： Hygon C86 7375 32-core Processor
CPU(s)：128
内存：≥ 600G, 1T以上最优
```bash
python run.py
```
## result
日志信息可以参考**log.txt**文件
<div align=center>
<img src="./doc/end.png"/>
</div>
### 精度
暂无
## 应用场景
### 算法类别
对话问答
### 热点应用行业
制造,广媒,家居,教育
## 预训练权重
下载地址[huggingface_grok-1](https://hf-mirror.com/xai-org/grok-1)
也可以直接使用`./checkpoints/down_models.py`进行下载，模型默认保存地址为：`~/.cache/modelscope/hub/cjc1887415157/grok-1/`。
```bash
python ./checkpoints/down_models.py
mv ~/.cache/modelscope/hub/cjc1887415157/grok-1/ckpt-0 ./
```
## 源码仓库及问题反馈
- https://developer.hpccube.com/codes/modelzoo/grok-1_jax
## 参考资料
- https://github.com/xai-org/grok-1
README_ori.md 0 → 100644
View file @ 049925cc
# Grok-1
image.sourcefind.cn:5000/dcu/admin/base/tensorflow:2.11.0-ubuntu20.04-dtk23.10-py310
This repository contains JAX example code for loading and running the Grok-1 open-weights model.
Make sure to download the checkpoint and place the `ckpt-0` directory in `checkpoints` - see [Downloading the weights](#downloading-the-weights)
Then, run
```shell
pip install -r requirements.txt
python run.py
```
to test the code.
The script loads the checkpoint and samples from the model on a test input.
Due to the large size of the model (314B parameters), a machine with enough GPU memory is required to test the model with the example code.
The implementation of the MoE layer in this repository is not efficient. The implementation was chosen to avoid the need for custom kernels to validate the correctness of the model.
# Model Specifications
Grok-1 is currently designed with the following specifications:
- **Parameters:** 314B
- **Architecture:** Mixture of 8 Experts (MoE)
- **Experts Utilization:** 2 experts used per token
- **Layers:** 64
- **Attention Heads:** 48 for queries, 8 for keys/values
- **Embedding Size:** 6,144
- **Tokenization:** SentencePiece tokenizer with 131,072 tokens
- **Additional Features:**
- Rotary embeddings (RoPE)
- Supports activation sharding and 8-bit quantization
- **Maximum Sequence Length (context):** 8,192 tokens
# Downloading the weights
You can download the weights using a torrent client and this magnet link:
```
magnet:?xt=urn:btih:5f96d43576e3d386c9ba65b883210a393b68210e&tr=https%3A%2F%2Facademictorrents.com%2Fannounce.php&tr=udp%3A%2F%2Ftracker.coppersurfer.tk%3A6969&tr=udp%3A%2F%2Ftracker.opentrackr.org%3A1337%2Fannounce
```
or directly using [HuggingFace 🤗 Hub](https://huggingface.co/xai-org/grok-1):
```
git clone https://github.com/xai-org/grok-1.git && cd grok-1
pip install huggingface_hub[hf_transfer]
huggingface-cli download xai-org/grok-1 --repo-type model --include ckpt-0/* --local-dir checkpoints --local-dir-use-symlinks False
```
# License
The code and associated Grok-1 weights in this release are licensed under the
Apache 2.0 license. The license only applies to the source files in this
repository and the model weights of Grok-1.
checkpoint.py 0 → 100644
View file @ 049925cc
# Copyright 2024 X.AI Corp.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import annotations
import contextlib
import logging
import math
import os
import pickle
import re
import shutil
import sys
import tempfile
from concurrent.futures import ThreadPoolExecutor, wait
from typing import Any, Optional
import jax
import numpy as np
from jax.experimental import multihost_utils
from model import QuantizedWeight8bit
logger = logging.getLogger(__name__)
rank_logger = logging.getLogger("rank")
# Needed for loading the checkpoint with pickle.
sys.modules['__main__'].QuantizedWeight8bit = QuantizedWeight8bit
@contextlib.contextmanager
def copy_to_shm(file: str):
if file.startswith("/dev/shm/"):
# Nothing to do, the file is already in shared memory.
yield file
return
tmp_dir = "/dev/shm/"
fd, tmp_path = tempfile.mkstemp(dir=tmp_dir)
try:
shutil.copyfile(file, tmp_path)
yield tmp_path
finally:
os.remove(tmp_path)
os.close(fd)
@contextlib.contextmanager
def copy_from_shm(file: str):
tmp_dir = "/dev/shm/"
fd, tmp_path = tempfile.mkstemp(dir=tmp_dir)
try:
yield tmp_path
shutil.copyfile(tmp_path, file)
finally:
os.remove(tmp_path)
os.close(fd)
# def fast_unpickle(path: str) -> Any:
# with copy_to_shm(path) as tmp_path:
# with open(tmp_path, "rb") as f:
# return pickle.load(f)
def fast_unpickle(path: str) -> Any:
with open(path, "rb") as f:
return pickle.load(f)
def fast_pickle(obj: Any, path: str) -> None:
with copy_from_shm(path) as tmp_path:
with open(tmp_path, "wb") as f:
pickle.dump(obj, f)
def load_tensors(shaped_arrays, directory, mesh_config, tensor_indices=None):
"""Loads a set of arrays."""
pool = ThreadPoolExecutor(max_workers=32)
fs = list()
num_tensors = 0
num_replicas = 1
data_model_shards = math.prod(mesh_config)
if tensor_indices is None:
iterator = enumerate(shaped_arrays)
else:
iterator = zip(tensor_indices, shaped_arrays)
for i, t in iterator:
if (i % num_replicas) == ((jax.process_index() // data_model_shards) % num_replicas):
idx = (
jax.process_index() // (num_replicas * data_model_shards) * data_model_shards
+ jax.process_index() % data_model_shards
)
fs.append(
pool.submit(fast_unpickle, os.path.join(directory, f"tensor{i:05d}_{idx:03d}"))
)
num_tensors += 1
else:
fs.append(pool.submit(np.zeros, t.shape, dtype=t.dtype))
wait(fs)
return [f.result() for f in fs]
def path_tuple_to_string(path: tuple) -> str:
pieces = []
for elem in path:
if isinstance(elem, jax.tree_util.DictKey):
pieces.append(elem.key)
elif isinstance(elem, jax.tree_util.GetAttrKey):
pieces.append(elem.name)
else:
assert isinstance(elem, (jax.tree_util.FlattenedIndexKey, jax.tree_util.SequenceKey))
return "/".join(pieces)
def get_load_path_str(
init_path_str: str,
load_rename_rules: Optional[list[tuple[str, str]]] = None,
load_exclude_rules: Optional[list[str]] = None,
) -> Optional[str]:
# Exclusion
if load_exclude_rules is not None:
for search_pattern in load_exclude_rules:
if re.search(search_pattern, init_path_str):
return None
# Renaming
load_path_str = init_path_str
if load_rename_rules is not None:
for search_pattern, replacement_pattern in load_rename_rules:
if re.search(search_pattern, load_path_str):
load_path_str = re.sub(search_pattern, replacement_pattern, load_path_str)
break
return load_path_str
def replace_with_load_state(
init_state: Any,
load_state: Any,
load_rename_rules: Optional[list[tuple[str, str]]] = None,
load_exclude_rules: Optional[list[str]] = None,
mesh_config: tuple = (1, 1),
) -> Any:
flatten_load, _ = jax.tree_util.tree_flatten_with_path(load_state)
flatten_init, structure_init = jax.tree_util.tree_flatten_with_path(init_state)
load_map = {path_tuple_to_string(path): tensor for path, tensor in flatten_load}
replaced = []
num_replicas = 1
data_model_shards = math.prod(mesh_config)
for i, (init_path, tensor) in enumerate(flatten_init):
init_path_str = path_tuple_to_string(init_path)
load_path_str = get_load_path_str(init_path_str, load_rename_rules, load_exclude_rules)
if load_path_str is None:
rank_logger.info(f"Excluded from restore: {init_path_str}.")
replaced.append(tensor)
elif load_path_str in load_map:
if load_path_str == init_path_str:
rank_logger.info(f"Restored from ckpt: {init_path_str}.")
else:
rank_logger.info(f"Restored from ckpt: {init_path_str} <-- {load_path_str}.")
replaced.append(load_map[load_path_str])
else:
rank_logger.info(f"Not found in ckpt: {init_path_str}.")
if (i % num_replicas) == ((jax.process_index() // data_model_shards) % num_replicas):
replaced.append(tensor)
else:
replaced.append(np.zeros_like(tensor))
return jax.tree_util.tree_unflatten(structure_init, replaced)
def restore(
checkpoint_path: str,
state_shapes: Any,
mesh,
between_hosts_config,
params_only,
state_sharding,
init_state: Optional[Any] = None,
) -> Any:
ckpt_path = os.path.join(checkpoint_path, "ckpt-0")
rank_logger.info("Loading checkpoint at {}".format(ckpt_path))
ckpt_shapes = state_shapes
ckpt_shapes_with_path, structure = jax.tree_util.tree_flatten_with_path(ckpt_shapes)
ckpt_shapes_flat = [elem[1] for elem in ckpt_shapes_with_path]
loaded_tensors = load_tensors(ckpt_shapes_flat, ckpt_path, between_hosts_config)
state = jax.tree_util.tree_unflatten(structure, loaded_tensors)
# Sanity check to give a better error message.
ckpt_keys = set(state.params.keys())
code_keys = set(state_sharding.params.keys())
if ckpt_keys != code_keys and init_state is None:
missing_in_ckpt = code_keys - ckpt_keys
missing_locally = ckpt_keys - code_keys
raise ValueError(
"Parameters in the code are not matching checkpoint parameters.\n"
"Params missing in checkpoint: {}\nParams missing in code: {}".format(
missing_in_ckpt, missing_locally
)
)
state_sharding = jax.tree_util.tree_map(
lambda x: jax.sharding.PartitionSpec() if x is None else x,
state_sharding,
is_leaf=lambda x: x is None,
)
state = multihost_utils.host_local_array_to_global_array(state, mesh, state_sharding)
if params_only:
state = state.params
return state
checkpoints/README.md 0 → 100644
View file @ 049925cc
# Checkpoint directory
Place Grok-1 checkpoints here so they can be loaded by the example script.
debug.log 0 → 100644
View file @ 049925cc
[0403/192602.428:INFO:skin_prom_main.cpp(112)] skinprom start
[0403/192603.240:INFO:skin_prom_main.cpp(137)] skinprom end
doc/MoE.png 0 → 100644
View file @ 049925cc
doc/MoE.png

69.1 KB

doc/end.png 0 → 100644
View file @ 049925cc
doc/end.png

11.2 KB

docker/Dockerfile 0 → 100644
View file @ 049925cc
FROM docker pull image.sourcefind.cn:5000/dcu/admin/base/jax:0.4.23-ubuntu20.04-dtk24.04-py310
log.txt 0 → 100644
View file @ 049925cc
2024-04-02 20:06:34.521652: E external/xla/xla/stream_executor/plugin_registry.cc:90] Invalid plugin kind specified: DNN
INFO:jax._src.xla_bridge:Unable to initialize backend 'cuda':
INFO:jax._src.xla_bridge:Unable to initialize backend 'tpu': INTERNAL: Failed to open libtpu.so: libtpu.so: cannot open shared object file: No such file or directory
INFO:rank:Initializing mesh for self.local_mesh_config=(1, 8) self.between_hosts_config=(1, 1)...
INFO:rank:Detected 8 devices in mesh
2024-04-02 20:06:38.881608: W external/xla/xla/service/gpu/gpu_compiler.cc:549] GpuCompilationEnvironment of hlo_module jit_convert_element_type:
2024-04-02 20:06:38.881674: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_backend_optimization_level: 3
2024-04-02 20:06:38.881683: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_eliminate_hlo_implicit_broadcast: true
2024-04-02 20:06:38.881689: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_multi_thread_eigen: true
2024-04-02 20:06:38.881695: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cuda_data_dir: "./cuda_sdk_lib"
2024-04-02 20:06:38.881700: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_alias_scope_metadata: true
2024-04-02 20:06:38.881705: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_noalias_metadata: true
2024-04-02 20:06:38.881710: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_invariant_load_metadata: true
2024-04-02 20:06:38.881716: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_force_host_platform_device_count: 1
2024-04-02 20:06:38.881721: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_nans: true
2024-04-02 20:06:38.881725: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_infs: true
2024-04-02 20:06:38.881730: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_allow_excess_precision: true
2024-04-02 20:06:38.881735: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_autotune_level: 4
2024-04-02 20:06:38.881740: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_division: true
2024-04-02 20:06:38.881745: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_functions: true
2024-04-02 20:06:38.882966: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_max_hlo_modules: -1
2024-04-02 20:06:38.882973: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_multiheap_size_constraint_per_heap: -1
2024-04-02 20:06:38.882978: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_async_all_reduce: true
2024-04-02 20:06:38.882983: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_strict_conv_algorithm_picker: true
2024-04-02 20:06:38.882989: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_reduce_combine_threshold_bytes: 31457280
2024-04-02 20:06:38.882994: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_frontend: true
2024-04-02 20:06:38.882999: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_nccl_termination_timeout_seconds: -1
2024-04-02 20:06:38.883004: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_shared_constants: true
2024-04-02 20:06:38.883009: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_scratch_max_megabytes: 4096
2024-04-02 20:06:38.883014: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_simplify_all_fp_conversions: true
2024-04-02 20:06:38.883019: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_xla_runtime_executable: true
2024-04-02 20:06:38.883028: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_shape_checks: RUNTIME
2024-04-02 20:06:38.884206: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_normalize_layouts: true
2024-04-02 20:06:38.884214: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_tiling_and_fusion: true
2024-04-02 20:06:38.884219: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_enable_mlir_pretty_form: true
2024-04-02 20:06:38.884224: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_triton_gemm: true
2024-04-02 20:06:38.884241: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_int8x32_convolution_reordering: true
2024-04-02 20:06:38.884246: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_experimental_deallocation: true
2024-04-02 20:06:38.884251: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_fusion_outlining: true
2024-04-02 20:06:38.884256: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_m_dim: 8
2024-04-02 20:06:38.884273: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_n_dim: 8
2024-04-02 20:06:38.884280: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_k_dim: 8
2024-04-02 20:06:38.885725: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_num_runs_to_instantiate: -1
2024-04-02 20:06:38.885742: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_lhs_enable_gpu_async_tracker: true
2024-04-02 20:06:38.885751: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_inflation_factor: 1
2024-04-02 20:06:38.887291: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_min_graph_size: 5
2024-04-02 20:06:38.887308: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reassociation_for_converted_ar: true
2024-04-02 20:06:38.887318: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_gather_combine_threshold_bytes: 31457280
2024-04-02 20:06:38.887326: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_reduce_scatter_combine_threshold_bytes: 31457280
2024-04-02 20:06:38.887335: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_highest_priority_async_stream: true
2024-04-02 20:06:38.888307: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_auto_spmd_partitioning_memory_budget_ratio: 1.1
2024-04-02 20:06:38.888318: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_padding_bytes: 8388608
2024-04-02 20:06:38.888324: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_triton_fusion_level: 2
2024-04-02 20:06:38.888329: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_eviction_timeout_seconds: 60
2024-04-02 20:06:38.888334: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_gpu2_hal: true
2024-04-02 20:06:38.888339: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_copy_insertion_use_region_analysis: true
2024-04-02 20:06:38.888344: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_permute_decomposer_threshold: 9223372036854775807
2024-04-02 20:06:38.888349: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_split_k_autotuning: true
2024-04-02 20:06:38.888355: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduction_epilogue_fusion: true
2024-04-02 20:06:38.888359: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cublas_fallback: true
2024-04-02 20:06:38.889743: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_filter_kernels_spilling_registers_on_autotuning: true
2024-04-02 20:06:38.889751: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_debug_buffer_assignment_show_max: 15
2024-04-02 20:06:38.889756: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_enable_dumping: true
2024-04-02 20:06:38.889761: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_all_gather_combine_by_dim: true
2024-04-02 20:06:38.889766: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduce_scatter_combine_by_dim: true
2024-04-02 20:06:38.889771: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: FUSION
2024-04-02 20:06:38.889776: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: CUBLAS
2024-04-02 20:06:38.889781: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cub_radix_sort: true
2024-04-02 20:06:38.889786: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_memory_limit_slop_factor: 95
2024-04-02 20:06:38.889797: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_threshold_for_windowed_einsum_mib: 100000
2024-04-02 20:06:43.779722: W external/xla/xla/service/gpu/gpu_compiler.cc:549] GpuCompilationEnvironment of hlo_module jit__threefry_seed:
2024-04-02 20:06:43.779803: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_backend_optimization_level: 3
2024-04-02 20:06:43.779812: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_eliminate_hlo_implicit_broadcast: true
2024-04-02 20:06:43.779819: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_multi_thread_eigen: true
2024-04-02 20:06:43.779825: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cuda_data_dir: "./cuda_sdk_lib"
2024-04-02 20:06:43.779830: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_alias_scope_metadata: true
2024-04-02 20:06:43.779836: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_noalias_metadata: true
2024-04-02 20:06:43.779841: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_invariant_load_metadata: true
2024-04-02 20:06:43.779846: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_force_host_platform_device_count: 1
2024-04-02 20:06:43.779851: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_nans: true
2024-04-02 20:06:43.779856: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_infs: true
2024-04-02 20:06:43.779861: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_allow_excess_precision: true
2024-04-02 20:06:43.779865: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_autotune_level: 4
2024-04-02 20:06:43.779870: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_division: true
2024-04-02 20:06:43.779875: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_functions: true
2024-04-02 20:06:43.779880: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_max_hlo_modules: -1
2024-04-02 20:06:43.779885: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_multiheap_size_constraint_per_heap: -1
2024-04-02 20:06:43.779890: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_async_all_reduce: true
2024-04-02 20:06:43.779894: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_strict_conv_algorithm_picker: true
2024-04-02 20:06:43.779899: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_reduce_combine_threshold_bytes: 31457280
2024-04-02 20:06:43.779904: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_frontend: true
2024-04-02 20:06:43.779909: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_nccl_termination_timeout_seconds: -1
2024-04-02 20:06:43.779914: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_shared_constants: true
2024-04-02 20:06:43.779918: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_scratch_max_megabytes: 4096
2024-04-02 20:06:43.779923: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_simplify_all_fp_conversions: true
2024-04-02 20:06:43.779928: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_xla_runtime_executable: true
2024-04-02 20:06:43.779933: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_shape_checks: RUNTIME
2024-04-02 20:06:43.779938: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_normalize_layouts: true
2024-04-02 20:06:43.779943: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_tiling_and_fusion: true
2024-04-02 20:06:43.779949: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_enable_mlir_pretty_form: true
2024-04-02 20:06:43.782200: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_triton_gemm: true
2024-04-02 20:06:43.782209: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_int8x32_convolution_reordering: true
2024-04-02 20:06:43.782214: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_experimental_deallocation: true
2024-04-02 20:06:43.782225: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_fusion_outlining: true
2024-04-02 20:06:43.782230: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_m_dim: 8
2024-04-02 20:06:43.782235: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_n_dim: 8
2024-04-02 20:06:43.782240: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_k_dim: 8
2024-04-02 20:06:43.782245: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_num_runs_to_instantiate: -1
2024-04-02 20:06:43.782249: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_lhs_enable_gpu_async_tracker: true
2024-04-02 20:06:43.782254: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_inflation_factor: 1
2024-04-02 20:06:43.782266: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_min_graph_size: 5
2024-04-02 20:06:43.782273: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reassociation_for_converted_ar: true
2024-04-02 20:06:43.783512: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_gather_combine_threshold_bytes: 31457280
2024-04-02 20:06:43.783518: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_reduce_scatter_combine_threshold_bytes: 31457280
2024-04-02 20:06:43.783523: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_highest_priority_async_stream: true
2024-04-02 20:06:43.783528: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_auto_spmd_partitioning_memory_budget_ratio: 1.1
2024-04-02 20:06:43.783532: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_padding_bytes: 8388608
2024-04-02 20:06:43.783537: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_triton_fusion_level: 2
2024-04-02 20:06:43.783542: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_eviction_timeout_seconds: 60
2024-04-02 20:06:43.783547: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_gpu2_hal: true
2024-04-02 20:06:43.783552: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_copy_insertion_use_region_analysis: true
2024-04-02 20:06:43.783557: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_permute_decomposer_threshold: 9223372036854775807
2024-04-02 20:06:43.783562: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_split_k_autotuning: true
2024-04-02 20:06:43.783568: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduction_epilogue_fusion: true
2024-04-02 20:06:43.784956: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cublas_fallback: true
2024-04-02 20:06:43.784963: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_filter_kernels_spilling_registers_on_autotuning: true
2024-04-02 20:06:43.784968: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_debug_buffer_assignment_show_max: 15
2024-04-02 20:06:43.784973: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_detailed_logging: true
2024-04-02 20:06:43.784978: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_enable_dumping: true
2024-04-02 20:06:43.784983: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_all_gather_combine_by_dim: true
2024-04-02 20:06:43.784988: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduce_scatter_combine_by_dim: true
2024-04-02 20:06:43.784993: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: FUSION
2024-04-02 20:06:43.784998: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: CUBLAS
2024-04-02 20:06:43.785003: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cub_radix_sort: true
2024-04-02 20:06:43.785008: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_memory_limit_slop_factor: 95
2024-04-02 20:06:43.785013: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_threshold_for_windowed_einsum_mib: 100000
INFO:rank:partition rules: <bound method LanguageModelConfig.partition_rules of LanguageModelConfig(model=TransformerConfig(emb_size=6144, key_size=128, num_q_heads=48, num_kv_heads=8, num_layers=64, vocab_size=131072, widening_factor=8, attn_output_multiplier=0.08838834764831845, name=None, num_experts=8, capacity_factor=1.0, num_selected_experts=2, init_scale=1.0, shard_activations=True, data_axis='data', model_axis='model'), vocab_size=131072, pad_token=0, eos_token=2, sequence_len=8192, model_size=6144, embedding_init_scale=1.0, embedding_multiplier_scale=78.38367176906169, output_multiplier_scale=0.5773502691896257, name=None, fprop_dtype=<class 'jax.numpy.bfloat16'>, model_type=None, init_scale_override=None, shard_embeddings=True)>
INFO:rank:(1, 256, 6144)
INFO:rank:(1, 256, 131072)
INFO:rank:State sharding type: <class 'model.TrainingState'>
INFO:rank:(1, 256, 6144)
INFO:rank:(1, 256, 131072)
INFO:rank:Loading checkpoint at ./checkpoints/ckpt-0
2024-04-02 20:16:02.345285: W external/xla/xla/service/gpu/gpu_compiler.cc:549] GpuCompilationEnvironment of hlo_module jit_broadcast_in_dim:
2024-04-02 20:16:02.345369: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_backend_optimization_level: 3
2024-04-02 20:16:02.345378: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_eliminate_hlo_implicit_broadcast: true
2024-04-02 20:16:02.345386: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_multi_thread_eigen: true
2024-04-02 20:16:02.345393: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cuda_data_dir: "./cuda_sdk_lib"
2024-04-02 20:16:02.345399: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_alias_scope_metadata: true
2024-04-02 20:16:02.345404: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_noalias_metadata: true
2024-04-02 20:16:02.345410: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_invariant_load_metadata: true
2024-04-02 20:16:02.345416: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_force_host_platform_device_count: 1
2024-04-02 20:16:02.345422: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_nans: true
2024-04-02 20:16:02.345428: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_infs: true
2024-04-02 20:16:02.345434: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_allow_excess_precision: true
2024-04-02 20:16:02.345439: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_autotune_level: 4
2024-04-02 20:16:02.345445: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_division: true
2024-04-02 20:16:02.345451: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_functions: true
2024-04-02 20:16:02.345457: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_max_hlo_modules: -1
2024-04-02 20:16:02.345462: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_multiheap_size_constraint_per_heap: -1
2024-04-02 20:16:02.345467: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_async_all_reduce: true
2024-04-02 20:16:02.345473: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_strict_conv_algorithm_picker: true
2024-04-02 20:16:02.345478: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_reduce_combine_threshold_bytes: 31457280
2024-04-02 20:16:02.345483: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_frontend: true
2024-04-02 20:16:02.345488: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_nccl_termination_timeout_seconds: -1
2024-04-02 20:16:02.345494: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_shared_constants: true
2024-04-02 20:16:02.345499: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_scratch_max_megabytes: 4096
2024-04-02 20:16:02.345504: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_simplify_all_fp_conversions: true
2024-04-02 20:16:02.345508: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_xla_runtime_executable: true
2024-04-02 20:16:02.345513: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_shape_checks: RUNTIME
2024-04-02 20:16:02.345525: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_normalize_layouts: true
2024-04-02 20:16:02.345530: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_tiling_and_fusion: true
2024-04-02 20:16:02.345536: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_enable_mlir_pretty_form: true
2024-04-02 20:16:02.345542: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_triton_gemm: true
2024-04-02 20:16:02.345547: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_int8x32_convolution_reordering: true
2024-04-02 20:16:02.345551: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_experimental_deallocation: true
2024-04-02 20:16:02.345556: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_fusion_outlining: true
2024-04-02 20:16:02.345561: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_m_dim: 8
2024-04-02 20:16:02.345567: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_n_dim: 8
2024-04-02 20:16:02.345572: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_k_dim: 8
2024-04-02 20:16:02.345577: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_num_runs_to_instantiate: -1
2024-04-02 20:16:02.345582: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_lhs_enable_gpu_async_tracker: true
2024-04-02 20:16:02.345587: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_inflation_factor: 1
2024-04-02 20:16:02.345592: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_min_graph_size: 5
2024-04-02 20:16:02.345597: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reassociation_for_converted_ar: true
2024-04-02 20:16:02.345603: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_gather_combine_threshold_bytes: 31457280
2024-04-02 20:16:02.345607: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_reduce_scatter_combine_threshold_bytes: 31457280
2024-04-02 20:16:02.345612: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_highest_priority_async_stream: true
2024-04-02 20:16:02.345617: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_auto_spmd_partitioning_memory_budget_ratio: 1.1
2024-04-02 20:16:02.345622: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_padding_bytes: 8388608
2024-04-02 20:16:02.345630: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_triton_fusion_level: 2
2024-04-02 20:16:02.345638: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_eviction_timeout_seconds: 60
2024-04-02 20:16:02.345646: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_gpu2_hal: true
2024-04-02 20:16:02.345654: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_copy_insertion_use_region_analysis: true
2024-04-02 20:16:02.345660: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_permute_decomposer_threshold: 9223372036854775807
2024-04-02 20:16:02.345665: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_split_k_autotuning: true
2024-04-02 20:16:02.345677: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduction_epilogue_fusion: true
2024-04-02 20:16:02.345683: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cublas_fallback: true
2024-04-02 20:16:02.345688: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_filter_kernels_spilling_registers_on_autotuning: true
2024-04-02 20:16:02.345692: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_debug_buffer_assignment_show_max: 15
2024-04-02 20:16:02.345698: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_enable_dumping: true
2024-04-02 20:16:02.345702: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_all_gather_combine_by_dim: true
2024-04-02 20:16:02.345707: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduce_scatter_combine_by_dim: true
2024-04-02 20:16:02.345717: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: FUSION
2024-04-02 20:16:02.345723: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: CUBLAS
2024-04-02 20:16:02.345729: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cub_radix_sort: true
2024-04-02 20:16:02.345735: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_memory_limit_slop_factor: 95
2024-04-02 20:16:02.345740: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_threshold_for_windowed_einsum_mib: 100000
INFO:rank:(1, 8192, 6144)
INFO:rank:(1, 8192, 131072)
2024-04-02 20:16:11.323926: W external/xla/xla/service/gpu/gpu_compiler.cc:549] GpuCompilationEnvironment of hlo_module jit__threefry_split:
2024-04-02 20:16:11.323992: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_backend_optimization_level: 3
2024-04-02 20:16:11.324001: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_eliminate_hlo_implicit_broadcast: true
2024-04-02 20:16:11.324007: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_multi_thread_eigen: true
2024-04-02 20:16:11.324013: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cuda_data_dir: "./cuda_sdk_lib"
2024-04-02 20:16:11.324018: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_alias_scope_metadata: true
2024-04-02 20:16:11.324023: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_noalias_metadata: true
2024-04-02 20:16:11.324028: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_invariant_load_metadata: true
2024-04-02 20:16:11.324033: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_force_host_platform_device_count: 1
2024-04-02 20:16:11.324038: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_nans: true
2024-04-02 20:16:11.324044: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_infs: true
2024-04-02 20:16:11.324049: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_allow_excess_precision: true
2024-04-02 20:16:11.324054: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_autotune_level: 4
2024-04-02 20:16:11.324059: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_division: true
2024-04-02 20:16:11.324064: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_functions: true
2024-04-02 20:16:11.324069: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_max_hlo_modules: -1
2024-04-02 20:16:11.324074: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_multiheap_size_constraint_per_heap: -1
2024-04-02 20:16:11.324079: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_async_all_reduce: true
2024-04-02 20:16:11.324084: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_strict_conv_algorithm_picker: true
2024-04-02 20:16:11.324089: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_reduce_combine_threshold_bytes: 31457280
2024-04-02 20:16:11.324094: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_frontend: true
2024-04-02 20:16:11.324099: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_nccl_termination_timeout_seconds: -1
2024-04-02 20:16:11.324104: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_shared_constants: true
2024-04-02 20:16:11.324109: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_scratch_max_megabytes: 4096
2024-04-02 20:16:11.324114: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_simplify_all_fp_conversions: true
2024-04-02 20:16:11.324119: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_xla_runtime_executable: true
2024-04-02 20:16:11.324124: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_shape_checks: RUNTIME
2024-04-02 20:16:11.326912: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_normalize_layouts: true
2024-04-02 20:16:11.326919: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_tiling_and_fusion: true
2024-04-02 20:16:11.326932: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_enable_mlir_pretty_form: true
2024-04-02 20:16:11.326938: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_triton_gemm: true
2024-04-02 20:16:11.326943: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_int8x32_convolution_reordering: true
2024-04-02 20:16:11.326948: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_experimental_deallocation: true
2024-04-02 20:16:11.326953: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_fusion_outlining: true
2024-04-02 20:16:11.326957: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_m_dim: 8
2024-04-02 20:16:11.326962: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_n_dim: 8
2024-04-02 20:16:11.326967: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_k_dim: 8
2024-04-02 20:16:11.326972: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_num_runs_to_instantiate: -1
2024-04-02 20:16:11.326977: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_lhs_enable_gpu_async_tracker: true
2024-04-02 20:16:11.327978: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_inflation_factor: 1
2024-04-02 20:16:11.327984: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_min_graph_size: 5
2024-04-02 20:16:11.327989: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reassociation_for_converted_ar: true
2024-04-02 20:16:11.327994: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_gather_combine_threshold_bytes: 31457280
2024-04-02 20:16:11.327999: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_reduce_scatter_combine_threshold_bytes: 31457280
2024-04-02 20:16:11.328005: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_highest_priority_async_stream: true
2024-04-02 20:16:11.328010: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_auto_spmd_partitioning_memory_budget_ratio: 1.1
2024-04-02 20:16:11.328015: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_padding_bytes: 8388608
2024-04-02 20:16:11.328020: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_triton_fusion_level: 2
2024-04-02 20:16:11.328025: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_eviction_timeout_seconds: 60
2024-04-02 20:16:11.328030: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_gpu2_hal: true
2024-04-02 20:16:11.329152: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_copy_insertion_use_region_analysis: true
2024-04-02 20:16:11.329158: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_permute_decomposer_threshold: 9223372036854775807
2024-04-02 20:16:11.329163: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_split_k_autotuning: true
2024-04-02 20:16:11.329168: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduction_epilogue_fusion: true
2024-04-02 20:16:11.329173: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cublas_fallback: true
2024-04-02 20:16:11.329179: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_filter_kernels_spilling_registers_on_autotuning: true
2024-04-02 20:16:11.329184: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_debug_buffer_assignment_show_max: 15
2024-04-02 20:16:11.329189: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_detailed_logging: true
2024-04-02 20:16:11.329194: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_enable_dumping: true
2024-04-02 20:16:11.329199: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_all_gather_combine_by_dim: true
2024-04-02 20:16:11.329204: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduce_scatter_combine_by_dim: true
2024-04-02 20:16:11.329210: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: FUSION
2024-04-02 20:16:11.330869: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: CUBLAS
2024-04-02 20:16:11.330876: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cub_radix_sort: true
2024-04-02 20:16:11.330881: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_memory_limit_slop_factor: 95
2024-04-02 20:16:11.330886: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_threshold_for_windowed_einsum_mib: 100000
2024-04-02 20:16:11.776457: W external/xla/xla/service/gpu/gpu_compiler.cc:549] GpuCompilationEnvironment of hlo_module pjit_apply_fn:
2024-04-02 20:16:11.776524: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_backend_optimization_level: 3
2024-04-02 20:16:11.776533: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_eliminate_hlo_implicit_broadcast: true
2024-04-02 20:16:11.776539: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_multi_thread_eigen: true
2024-04-02 20:16:11.776545: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cuda_data_dir: "./cuda_sdk_lib"
2024-04-02 20:16:11.776551: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_alias_scope_metadata: true
2024-04-02 20:16:11.776556: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_noalias_metadata: true
2024-04-02 20:16:11.776561: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_invariant_load_metadata: true
2024-04-02 20:16:11.776566: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_force_host_platform_device_count: 1
2024-04-02 20:16:11.776571: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_nans: true
2024-04-02 20:16:11.776576: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_infs: true
2024-04-02 20:16:11.776581: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_allow_excess_precision: true
2024-04-02 20:16:11.776586: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_autotune_level: 4
2024-04-02 20:16:11.776591: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_division: true
2024-04-02 20:16:11.776595: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_functions: true
2024-04-02 20:16:11.776600: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_max_hlo_modules: -1
2024-04-02 20:16:11.776605: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_multiheap_size_constraint_per_heap: -1
2024-04-02 20:16:11.776610: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_async_all_reduce: true
2024-04-02 20:16:11.776615: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_strict_conv_algorithm_picker: true
2024-04-02 20:16:11.776620: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_reduce_combine_threshold_bytes: 31457280
2024-04-02 20:16:11.776625: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_frontend: true
2024-04-02 20:16:11.776629: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_nccl_termination_timeout_seconds: -1
2024-04-02 20:16:11.776634: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_shared_constants: true
2024-04-02 20:16:11.776639: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_scratch_max_megabytes: 4096
2024-04-02 20:16:11.776644: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_simplify_all_fp_conversions: true
2024-04-02 20:16:11.780464: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_xla_runtime_executable: true
2024-04-02 20:16:11.780471: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_shape_checks: RUNTIME
2024-04-02 20:16:11.780477: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_normalize_layouts: true
2024-04-02 20:16:11.780482: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_tiling_and_fusion: true
2024-04-02 20:16:11.780487: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_enable_mlir_pretty_form: true
2024-04-02 20:16:11.780492: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_triton_gemm: true
2024-04-02 20:16:11.780503: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_int8x32_convolution_reordering: true
2024-04-02 20:16:11.780508: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_experimental_deallocation: true
2024-04-02 20:16:11.780513: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_fusion_outlining: true
2024-04-02 20:16:11.780518: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_m_dim: 8
2024-04-02 20:16:11.780522: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_n_dim: 8
2024-04-02 20:16:11.780527: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_k_dim: 8
2024-04-02 20:16:11.780532: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_num_runs_to_instantiate: -1
2024-04-02 20:16:11.782305: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_lhs_enable_gpu_async_tracker: true
2024-04-02 20:16:11.782311: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_inflation_factor: 1
2024-04-02 20:16:11.782316: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_min_graph_size: 5
2024-04-02 20:16:11.782321: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reassociation_for_converted_ar: true
2024-04-02 20:16:11.782326: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_gather_combine_threshold_bytes: 31457280
2024-04-02 20:16:11.782331: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_reduce_scatter_combine_threshold_bytes: 31457280
2024-04-02 20:16:11.782336: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_highest_priority_async_stream: true
2024-04-02 20:16:11.782341: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_auto_spmd_partitioning_memory_budget_ratio: 1.1
2024-04-02 20:16:11.782346: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_padding_bytes: 8388608
2024-04-02 20:16:11.782351: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_triton_fusion_level: 2
2024-04-02 20:16:11.782356: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_eviction_timeout_seconds: 60
2024-04-02 20:16:11.782361: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_gpu2_hal: true
2024-04-02 20:16:11.788516: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_copy_insertion_use_region_analysis: true
2024-04-02 20:16:11.788522: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_permute_decomposer_threshold: 9223372036854775807
2024-04-02 20:16:11.788528: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_split_k_autotuning: true
2024-04-02 20:16:11.788532: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduction_epilogue_fusion: true
2024-04-02 20:16:11.788537: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cublas_fallback: true
2024-04-02 20:16:11.788542: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_filter_kernels_spilling_registers_on_autotuning: true
2024-04-02 20:16:11.788547: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_debug_buffer_assignment_show_max: 15
2024-04-02 20:16:11.788553: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_detailed_logging: true
2024-04-02 20:16:11.788558: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_enable_dumping: true
2024-04-02 20:16:11.788563: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_all_gather_combine_by_dim: true
2024-04-02 20:16:11.788567: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduce_scatter_combine_by_dim: true
2024-04-02 20:16:11.788572: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: FUSION
2024-04-02 20:16:11.788579: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: CUBLAS
2024-04-02 20:16:11.789651: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cub_radix_sort: true
2024-04-02 20:16:11.789663: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_memory_limit_slop_factor: 95
2024-04-02 20:16:11.789668: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_threshold_for_windowed_einsum_mib: 100000
INFO:runners:Precompile 1024
INFO:rank:(1, 1, 6144)
INFO:rank:(1, 1, 131072)
2024-04-02 20:16:37.688908: W external/xla/xla/service/gpu/gpu_compiler.cc:549] GpuCompilationEnvironment of hlo_module pjit__unnamed_wrapped_function_:
2024-04-02 20:16:37.688982: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_backend_optimization_level: 3
2024-04-02 20:16:37.688992: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_eliminate_hlo_implicit_broadcast: true
2024-04-02 20:16:37.688998: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_multi_thread_eigen: true
2024-04-02 20:16:37.689004: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cuda_data_dir: "./cuda_sdk_lib"
2024-04-02 20:16:37.689010: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_alias_scope_metadata: true
2024-04-02 20:16:37.689015: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_noalias_metadata: true
2024-04-02 20:16:37.689020: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_invariant_load_metadata: true
2024-04-02 20:16:37.689025: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_force_host_platform_device_count: 1
2024-04-02 20:16:37.689030: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_nans: true
2024-04-02 20:16:37.689034: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_infs: true
2024-04-02 20:16:37.689039: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_allow_excess_precision: true
2024-04-02 20:16:37.689044: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_autotune_level: 4
2024-04-02 20:16:37.689049: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_division: true
2024-04-02 20:16:37.689054: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_functions: true
2024-04-02 20:16:37.689059: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_max_hlo_modules: -1
2024-04-02 20:16:37.689064: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_multiheap_size_constraint_per_heap: -1
2024-04-02 20:16:37.689069: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_async_all_reduce: true
2024-04-02 20:16:37.689074: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_strict_conv_algorithm_picker: true
2024-04-02 20:16:37.690290: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_reduce_combine_threshold_bytes: 31457280
2024-04-02 20:16:37.690297: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_frontend: true
2024-04-02 20:16:37.690302: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_nccl_termination_timeout_seconds: -1
2024-04-02 20:16:37.690307: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_shared_constants: true
2024-04-02 20:16:37.690312: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_scratch_max_megabytes: 4096
2024-04-02 20:16:37.690317: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_simplify_all_fp_conversions: true
2024-04-02 20:16:37.690322: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_xla_runtime_executable: true
2024-04-02 20:16:37.690327: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_shape_checks: RUNTIME
2024-04-02 20:16:37.690332: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_normalize_layouts: true
2024-04-02 20:16:37.690337: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_tiling_and_fusion: true
2024-04-02 20:16:37.690342: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_enable_mlir_pretty_form: true
2024-04-02 20:16:37.690347: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_triton_gemm: true
2024-04-02 20:16:37.690364: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_int8x32_convolution_reordering: true
2024-04-02 20:16:37.690371: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_experimental_deallocation: true
2024-04-02 20:16:37.690376: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_fusion_outlining: true
2024-04-02 20:16:37.690381: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_m_dim: 8
2024-04-02 20:16:37.690386: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_n_dim: 8
2024-04-02 20:16:37.690390: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_k_dim: 8
2024-04-02 20:16:37.690395: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_num_runs_to_instantiate: -1
2024-04-02 20:16:37.690400: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_lhs_enable_gpu_async_tracker: true
2024-04-02 20:16:37.690405: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_inflation_factor: 1
2024-04-02 20:16:37.690409: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_min_graph_size: 5
2024-04-02 20:16:37.690414: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reassociation_for_converted_ar: true
2024-04-02 20:16:37.690419: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_gather_combine_threshold_bytes: 31457280
2024-04-02 20:16:37.690424: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_reduce_scatter_combine_threshold_bytes: 31457280
2024-04-02 20:16:37.690429: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_highest_priority_async_stream: true
2024-04-02 20:16:37.690434: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_auto_spmd_partitioning_memory_budget_ratio: 1.1
2024-04-02 20:16:37.690438: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_padding_bytes: 8388608
2024-04-02 20:16:37.690443: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_triton_fusion_level: 2
2024-04-02 20:16:37.690448: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_eviction_timeout_seconds: 60
2024-04-02 20:16:37.690453: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_gpu2_hal: true
2024-04-02 20:16:37.690457: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_copy_insertion_use_region_analysis: true
2024-04-02 20:16:37.690462: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_permute_decomposer_threshold: 9223372036854775807
2024-04-02 20:16:37.690467: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_split_k_autotuning: true
2024-04-02 20:16:37.690472: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduction_epilogue_fusion: true
2024-04-02 20:16:37.690477: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cublas_fallback: true
2024-04-02 20:16:37.690482: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_filter_kernels_spilling_registers_on_autotuning: true
2024-04-02 20:16:37.690486: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_debug_buffer_assignment_show_max: 15
2024-04-02 20:16:37.690491: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_detailed_logging: true
2024-04-02 20:16:37.690496: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_enable_dumping: true
2024-04-02 20:16:37.690501: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_all_gather_combine_by_dim: true
2024-04-02 20:16:37.690505: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduce_scatter_combine_by_dim: true
2024-04-02 20:16:37.690510: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: FUSION
2024-04-02 20:16:37.690515: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: CUBLAS
2024-04-02 20:16:37.690519: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cub_radix_sort: true
2024-04-02 20:16:37.690524: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_memory_limit_slop_factor: 95
2024-04-02 20:16:37.690539: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_threshold_for_windowed_einsum_mib: 100000
2024-04-02 20:18:16.404824: E external/xla/xla/service/rendezvous.cc:31] This thread has been waiting for 10 seconds and may be stuck:
2024-04-02 20:18:18.721123: E external/xla/xla/service/rendezvous.cc:31] This thread has been waiting for 10 seconds and may be stuck:
2024-04-02 20:18:20.185634: E external/xla/xla/service/rendezvous.cc:36] Thread is unstuck! Warning above was a false-positive. Perhaps the timeout is too short.
2024-04-02 20:18:20.185783: E external/xla/xla/service/rendezvous.cc:36] Thread is unstuck! Warning above was a false-positive. Perhaps the timeout is too short.
INFO:runners:Compiling...
INFO:rank:(1, 1, 6144)
INFO:rank:(1, 1, 131072)
2024-04-02 20:19:38.707084: W external/xla/xla/service/gpu/gpu_compiler.cc:549] GpuCompilationEnvironment of hlo_module pjit_apply_fn:
2024-04-02 20:19:38.707204: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_backend_optimization_level: 3
2024-04-02 20:19:38.707213: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_eliminate_hlo_implicit_broadcast: true
2024-04-02 20:19:38.707219: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_multi_thread_eigen: true
2024-04-02 20:19:38.707225: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cuda_data_dir: "./cuda_sdk_lib"
2024-04-02 20:19:38.707230: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_alias_scope_metadata: true
2024-04-02 20:19:38.707235: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_noalias_metadata: true
2024-04-02 20:19:38.707240: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_invariant_load_metadata: true
2024-04-02 20:19:38.707245: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_force_host_platform_device_count: 1
2024-04-02 20:19:38.707250: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_nans: true
2024-04-02 20:19:38.707255: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_infs: true
2024-04-02 20:19:38.707266: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_allow_excess_precision: true
2024-04-02 20:19:38.707272: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_autotune_level: 4
2024-04-02 20:19:38.707277: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_division: true
2024-04-02 20:19:38.707281: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_functions: true
2024-04-02 20:19:38.707286: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_max_hlo_modules: -1
2024-04-02 20:19:38.707291: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_multiheap_size_constraint_per_heap: -1
2024-04-02 20:19:38.707296: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_async_all_reduce: true
2024-04-02 20:19:38.707301: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_strict_conv_algorithm_picker: true
2024-04-02 20:19:38.707311: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_reduce_combine_threshold_bytes: 31457280
2024-04-02 20:19:38.707316: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_frontend: true
2024-04-02 20:19:38.707321: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_nccl_termination_timeout_seconds: -1
2024-04-02 20:19:38.707326: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_shared_constants: true
2024-04-02 20:19:38.708908: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_scratch_max_megabytes: 4096
2024-04-02 20:19:38.708917: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_simplify_all_fp_conversions: true
2024-04-02 20:19:38.708922: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_xla_runtime_executable: true
2024-04-02 20:19:38.708927: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_shape_checks: RUNTIME
2024-04-02 20:19:38.708942: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_normalize_layouts: true
2024-04-02 20:19:38.708948: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_tiling_and_fusion: true
2024-04-02 20:19:38.708952: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_enable_mlir_pretty_form: true
2024-04-02 20:19:38.708957: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_triton_gemm: true
2024-04-02 20:19:38.708963: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_int8x32_convolution_reordering: true
2024-04-02 20:19:38.708968: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_experimental_deallocation: true
2024-04-02 20:19:38.708974: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_fusion_outlining: true
2024-04-02 20:19:38.710184: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_m_dim: 8
2024-04-02 20:19:38.710191: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_n_dim: 8
2024-04-02 20:19:38.710197: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_k_dim: 8
2024-04-02 20:19:38.710202: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_num_runs_to_instantiate: -1
2024-04-02 20:19:38.710207: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_lhs_enable_gpu_async_tracker: true
2024-04-02 20:19:38.710212: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_inflation_factor: 1
2024-04-02 20:19:38.710217: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_min_graph_size: 5
2024-04-02 20:19:38.710222: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reassociation_for_converted_ar: true
2024-04-02 20:19:38.710227: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_gather_combine_threshold_bytes: 31457280
2024-04-02 20:19:38.710232: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_reduce_scatter_combine_threshold_bytes: 31457280
2024-04-02 20:19:38.710239: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_highest_priority_async_stream: true
2024-04-02 20:19:38.711294: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_auto_spmd_partitioning_memory_budget_ratio: 1.1
2024-04-02 20:19:38.711301: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_padding_bytes: 8388608
2024-04-02 20:19:38.711306: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_triton_fusion_level: 2
2024-04-02 20:19:38.711311: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_eviction_timeout_seconds: 60
2024-04-02 20:19:38.711315: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_gpu2_hal: true
2024-04-02 20:19:38.711320: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_copy_insertion_use_region_analysis: true
2024-04-02 20:19:38.711325: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_permute_decomposer_threshold: 9223372036854775807
2024-04-02 20:19:38.711330: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_split_k_autotuning: true
2024-04-02 20:19:38.711335: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduction_epilogue_fusion: true
2024-04-02 20:19:38.711339: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cublas_fallback: true
2024-04-02 20:19:38.711345: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_filter_kernels_spilling_registers_on_autotuning: true
2024-04-02 20:19:38.711350: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_debug_buffer_assignment_show_max: 15
2024-04-02 20:19:38.711356: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_detailed_logging: true
2024-04-02 20:19:38.712443: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_enable_dumping: true
2024-04-02 20:19:38.712450: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_all_gather_combine_by_dim: true
2024-04-02 20:19:38.712455: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduce_scatter_combine_by_dim: true
2024-04-02 20:19:38.712465: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: FUSION
2024-04-02 20:19:38.712469: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: CUBLAS
2024-04-02 20:19:38.712474: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cub_radix_sort: true
2024-04-02 20:19:38.712479: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_memory_limit_slop_factor: 95
2024-04-02 20:19:38.712484: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_threshold_for_windowed_einsum_mib: 100000
INFO:runners:Done compiling.
2024-04-02 20:24:05.210361: E external/xla/xla/service/rendezvous.cc:31] This thread has been waiting for 10 seconds and may be stuck:
2024-04-02 20:24:05.606087: E external/xla/xla/service/rendezvous.cc:31] This thread has been waiting for 10 seconds and may be stuck:
2024-04-02 20:24:05.609344: E external/xla/xla/service/rendezvous.cc:31] This thread has been waiting for 10 seconds and may be stuck:
2024-04-02 20:24:05.634597: E external/xla/xla/service/rendezvous.cc:31] This thread has been waiting for 10 seconds and may be stuck:
2024-04-02 20:24:05.868559: E external/xla/xla/service/rendezvous.cc:31] This thread has been waiting for 10 seconds and may be stuck:
2024-04-02 20:24:05.871576: E external/xla/xla/service/rendezvous.cc:31] This thread has been waiting for 10 seconds and may be stuck:
2024-04-02 20:24:15.149612: E external/xla/xla/service/rendezvous.cc:36] Thread is unstuck! Warning above was a false-positive. Perhaps the timeout is too short.
2024-04-02 20:24:15.149871: E external/xla/xla/service/rendezvous.cc:36] Thread is unstuck! Warning above was a false-positive. Perhaps the timeout is too short.
2024-04-02 20:24:15.151427: E external/xla/xla/service/rendezvous.cc:36] Thread is unstuck! Warning above was a false-positive. Perhaps the timeout is too short.
2024-04-02 20:24:15.153198: E external/xla/xla/service/rendezvous.cc:36] Thread is unstuck! Warning above was a false-positive. Perhaps the timeout is too short.
2024-04-02 20:24:15.154609: E external/xla/xla/service/rendezvous.cc:36] Thread is unstuck! Warning above was a false-positive. Perhaps the timeout is too short.
2024-04-02 20:24:15.155907: E external/xla/xla/service/rendezvous.cc:36] Thread is unstuck! Warning above was a false-positive. Perhaps the timeout is too short.
2024-04-02 20:25:18.110561: E external/xla/xla/service/rendezvous.cc:31] This thread has been waiting for 10 seconds and may be stuck:
2024-04-02 20:25:18.124629: E external/xla/xla/service/rendezvous.cc:31] This thread has been waiting for 10 seconds and may be stuck:
2024-04-02 20:25:18.131321: E external/xla/xla/service/rendezvous.cc:31] This thread has been waiting for 10 seconds and may be stuck:
2024-04-02 20:25:18.256010: E external/xla/xla/service/rendezvous.cc:31] This thread has been waiting for 10 seconds and may be stuck:
2024-04-02 20:25:18.275243: E external/xla/xla/service/rendezvous.cc:31] This thread has been waiting for 10 seconds and may be stuck:
2024-04-02 20:25:18.437802: E external/xla/xla/service/rendezvous.cc:31] This thread has been waiting for 10 seconds and may be stuck:
2024-04-02 20:25:18.466272: E external/xla/xla/service/rendezvous.cc:31] This thread has been waiting for 10 seconds and may be stuck:
2024-04-02 20:25:23.385495: E external/xla/xla/service/rendezvous.cc:36] Thread is unstuck! Warning above was a false-positive. Perhaps the timeout is too short.
2024-04-02 20:25:23.385636: E external/xla/xla/service/rendezvous.cc:36] Thread is unstuck! Warning above was a false-positive. Perhaps the timeout is too short.
2024-04-02 20:25:23.385793: E external/xla/xla/service/rendezvous.cc:36] Thread is unstuck! Warning above was a false-positive. Perhaps the timeout is too short.
2024-04-02 20:25:23.387155: E external/xla/xla/service/rendezvous.cc:36] Thread is unstuck! Warning above was a false-positive. Perhaps the timeout is too short.
2024-04-02 20:25:23.388295: E external/xla/xla/service/rendezvous.cc:36] Thread is unstuck! Warning above was a false-positive. Perhaps the timeout is too short.
2024-04-02 20:25:23.388400: E external/xla/xla/service/rendezvous.cc:36] Thread is unstuck! Warning above was a false-positive. Perhaps the timeout is too short.
2024-04-02 20:25:23.389957: E external/xla/xla/service/rendezvous.cc:36] Thread is unstuck! Warning above was a false-positive. Perhaps the timeout is too short.
2024-04-02 20:30:06.824387: W external/xla/xla/service/gpu/gpu_compiler.cc:549] GpuCompilationEnvironment of hlo_module jit_convert_element_type:
2024-04-02 20:30:06.824465: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_backend_optimization_level: 3
2024-04-02 20:30:06.824477: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_eliminate_hlo_implicit_broadcast: true
2024-04-02 20:30:06.824486: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_multi_thread_eigen: true
2024-04-02 20:30:06.824495: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cuda_data_dir: "./cuda_sdk_lib"
2024-04-02 20:30:06.824501: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_alias_scope_metadata: true
2024-04-02 20:30:06.824508: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_noalias_metadata: true
2024-04-02 20:30:06.824513: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_invariant_load_metadata: true
2024-04-02 20:30:06.824521: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_force_host_platform_device_count: 1
2024-04-02 20:30:06.824530: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_nans: true
2024-04-02 20:30:06.824549: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_infs: true
2024-04-02 20:30:06.824557: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_allow_excess_precision: true
2024-04-02 20:30:06.824565: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_autotune_level: 4
2024-04-02 20:30:06.824573: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_division: true
2024-04-02 20:30:06.824581: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_functions: true
2024-04-02 20:30:06.824590: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_max_hlo_modules: -1
2024-04-02 20:30:06.824596: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_multiheap_size_constraint_per_heap: -1
2024-04-02 20:30:06.824602: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_async_all_reduce: true
2024-04-02 20:30:06.824608: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_strict_conv_algorithm_picker: true
2024-04-02 20:30:06.824613: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_reduce_combine_threshold_bytes: 31457280
2024-04-02 20:30:06.824620: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_frontend: true
2024-04-02 20:30:06.824626: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_nccl_termination_timeout_seconds: -1
2024-04-02 20:30:06.824633: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_shared_constants: true
2024-04-02 20:30:06.824638: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_scratch_max_megabytes: 4096
2024-04-02 20:30:06.824643: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_simplify_all_fp_conversions: true
2024-04-02 20:30:06.824647: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_xla_runtime_executable: true
2024-04-02 20:30:06.824655: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_shape_checks: RUNTIME
2024-04-02 20:30:06.824664: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_normalize_layouts: true
2024-04-02 20:30:06.824670: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_tiling_and_fusion: true
2024-04-02 20:30:06.824677: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_enable_mlir_pretty_form: true
2024-04-02 20:30:06.824685: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_triton_gemm: true
2024-04-02 20:30:06.824700: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_int8x32_convolution_reordering: true
2024-04-02 20:30:06.824706: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_experimental_deallocation: true
2024-04-02 20:30:06.824711: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_fusion_outlining: true
2024-04-02 20:30:06.824718: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_m_dim: 8
2024-04-02 20:30:06.824725: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_n_dim: 8
2024-04-02 20:30:06.824732: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_k_dim: 8
2024-04-02 20:30:06.824738: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_num_runs_to_instantiate: -1
2024-04-02 20:30:06.824742: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_lhs_enable_gpu_async_tracker: true
2024-04-02 20:30:06.824748: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_inflation_factor: 1
2024-04-02 20:30:06.824756: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_min_graph_size: 5
2024-04-02 20:30:06.824762: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reassociation_for_converted_ar: true
2024-04-02 20:30:06.824767: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_gather_combine_threshold_bytes: 31457280
2024-04-02 20:30:06.824771: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_reduce_scatter_combine_threshold_bytes: 31457280
2024-04-02 20:30:06.824776: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_highest_priority_async_stream: true
2024-04-02 20:30:06.824781: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_auto_spmd_partitioning_memory_budget_ratio: 1.1
2024-04-02 20:30:06.824786: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_padding_bytes: 8388608
2024-04-02 20:30:06.824795: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_triton_fusion_level: 2
2024-04-02 20:30:06.824801: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_eviction_timeout_seconds: 60
2024-04-02 20:30:06.824808: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_gpu2_hal: true
2024-04-02 20:30:06.824814: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_copy_insertion_use_region_analysis: true
2024-04-02 20:30:06.824819: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_permute_decomposer_threshold: 9223372036854775807
2024-04-02 20:30:06.824826: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_split_k_autotuning: true
2024-04-02 20:30:06.824831: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduction_epilogue_fusion: true
2024-04-02 20:30:06.824838: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cublas_fallback: true
2024-04-02 20:30:06.824844: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_filter_kernels_spilling_registers_on_autotuning: true
2024-04-02 20:30:06.824849: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_debug_buffer_assignment_show_max: 15
2024-04-02 20:30:06.824856: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_enable_dumping: true
2024-04-02 20:30:06.824861: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_all_gather_combine_by_dim: true
2024-04-02 20:30:06.824866: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduce_scatter_combine_by_dim: true
2024-04-02 20:30:06.824872: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: FUSION
2024-04-02 20:30:06.824880: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: CUBLAS
2024-04-02 20:30:06.824887: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cub_radix_sort: true
2024-04-02 20:30:06.824894: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_memory_limit_slop_factor: 95
2024-04-02 20:30:06.824904: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_threshold_for_windowed_einsum_mib: 100000
2024-04-02 20:30:06.870851: W external/xla/xla/service/gpu/gpu_compiler.cc:549] GpuCompilationEnvironment of hlo_module jit_broadcast_in_dim:
2024-04-02 20:30:06.870932: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_backend_optimization_level: 3
2024-04-02 20:30:06.870941: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_eliminate_hlo_implicit_broadcast: true
2024-04-02 20:30:06.870950: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_multi_thread_eigen: true
2024-04-02 20:30:06.870955: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cuda_data_dir: "./cuda_sdk_lib"
2024-04-02 20:30:06.870960: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_alias_scope_metadata: true
2024-04-02 20:30:06.870966: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_noalias_metadata: true
2024-04-02 20:30:06.870970: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_invariant_load_metadata: true
2024-04-02 20:30:06.870976: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_force_host_platform_device_count: 1
2024-04-02 20:30:06.870981: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_nans: true
2024-04-02 20:30:06.870986: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_infs: true
2024-04-02 20:30:06.870991: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_allow_excess_precision: true
2024-04-02 20:30:06.870996: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_autotune_level: 4
2024-04-02 20:30:06.871001: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_division: true
2024-04-02 20:30:06.871006: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_functions: true
2024-04-02 20:30:06.873138: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_max_hlo_modules: -1
2024-04-02 20:30:06.873146: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_multiheap_size_constraint_per_heap: -1
2024-04-02 20:30:06.873153: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_async_all_reduce: true
2024-04-02 20:30:06.873158: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_strict_conv_algorithm_picker: true
2024-04-02 20:30:06.873164: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_reduce_combine_threshold_bytes: 31457280
2024-04-02 20:30:06.873169: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_frontend: true
2024-04-02 20:30:06.873174: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_nccl_termination_timeout_seconds: -1
2024-04-02 20:30:06.873179: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_shared_constants: true
2024-04-02 20:30:06.873184: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_scratch_max_megabytes: 4096
2024-04-02 20:30:06.873190: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_simplify_all_fp_conversions: true
2024-04-02 20:30:06.873196: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_xla_runtime_executable: true
2024-04-02 20:30:06.875107: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_shape_checks: RUNTIME
2024-04-02 20:30:06.875114: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_normalize_layouts: true
2024-04-02 20:30:06.875119: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_tiling_and_fusion: true
2024-04-02 20:30:06.875124: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_enable_mlir_pretty_form: true
2024-04-02 20:30:06.875130: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_triton_gemm: true
2024-04-02 20:30:06.875135: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_int8x32_convolution_reordering: true
2024-04-02 20:30:06.875140: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_experimental_deallocation: true
2024-04-02 20:30:06.875157: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_fusion_outlining: true
2024-04-02 20:30:06.875162: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_m_dim: 8
2024-04-02 20:30:06.875170: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_n_dim: 8
2024-04-02 20:30:06.878198: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_k_dim: 8
2024-04-02 20:30:06.878207: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_num_runs_to_instantiate: -1
2024-04-02 20:30:06.878213: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_lhs_enable_gpu_async_tracker: true
2024-04-02 20:30:06.878219: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_inflation_factor: 1
2024-04-02 20:30:06.878224: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_min_graph_size: 5
2024-04-02 20:30:06.878229: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reassociation_for_converted_ar: true
2024-04-02 20:30:06.878234: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_gather_combine_threshold_bytes: 31457280
2024-04-02 20:30:06.878240: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_reduce_scatter_combine_threshold_bytes: 31457280
2024-04-02 20:30:06.878244: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_highest_priority_async_stream: true
2024-04-02 20:30:06.878249: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_auto_spmd_partitioning_memory_budget_ratio: 1.1
2024-04-02 20:30:06.878255: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_padding_bytes: 8388608
2024-04-02 20:30:06.878266: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_triton_fusion_level: 2
2024-04-02 20:30:06.878271: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_eviction_timeout_seconds: 60
2024-04-02 20:30:06.878276: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_gpu2_hal: true
2024-04-02 20:30:06.878281: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_copy_insertion_use_region_analysis: true
2024-04-02 20:30:06.880486: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_permute_decomposer_threshold: 9223372036854775807
2024-04-02 20:30:06.880492: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_split_k_autotuning: true
2024-04-02 20:30:06.880497: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduction_epilogue_fusion: true
2024-04-02 20:30:06.880503: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cublas_fallback: true
2024-04-02 20:30:06.880509: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_filter_kernels_spilling_registers_on_autotuning: true
2024-04-02 20:30:06.880513: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_debug_buffer_assignment_show_max: 15
2024-04-02 20:30:06.880519: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_enable_dumping: true
2024-04-02 20:30:06.880523: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_all_gather_combine_by_dim: true
2024-04-02 20:30:06.880529: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduce_scatter_combine_by_dim: true
2024-04-02 20:30:06.880535: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: FUSION
2024-04-02 20:30:06.880540: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: CUBLAS
2024-04-02 20:30:06.880545: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cub_radix_sort: true
2024-04-02 20:30:06.880551: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_memory_limit_slop_factor: 95
2024-04-02 20:30:06.881900: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_threshold_for_windowed_einsum_mib: 100000
2024-04-02 20:30:07.020453: W external/xla/xla/service/gpu/gpu_compiler.cc:549] GpuCompilationEnvironment of hlo_module jit__squeeze:
2024-04-02 20:30:07.020526: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_backend_optimization_level: 3
2024-04-02 20:30:07.020542: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_eliminate_hlo_implicit_broadcast: true
2024-04-02 20:30:07.020549: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_multi_thread_eigen: true
2024-04-02 20:30:07.020554: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cuda_data_dir: "./cuda_sdk_lib"
2024-04-02 20:30:07.020559: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_alias_scope_metadata: true
2024-04-02 20:30:07.020564: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_noalias_metadata: true
2024-04-02 20:30:07.020569: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_invariant_load_metadata: true
2024-04-02 20:30:07.020574: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_force_host_platform_device_count: 1
2024-04-02 20:30:07.020579: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_nans: true
2024-04-02 20:30:07.020584: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_infs: true
2024-04-02 20:30:07.020589: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_allow_excess_precision: true
2024-04-02 20:30:07.020594: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_autotune_level: 4
2024-04-02 20:30:07.020599: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_division: true
2024-04-02 20:30:07.020603: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_functions: true
2024-04-02 20:30:07.020608: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_max_hlo_modules: -1
2024-04-02 20:30:07.020613: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_multiheap_size_constraint_per_heap: -1
2024-04-02 20:30:07.020621: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_async_all_reduce: true
2024-04-02 20:30:07.023268: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_strict_conv_algorithm_picker: true
2024-04-02 20:30:07.023276: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_reduce_combine_threshold_bytes: 31457280
2024-04-02 20:30:07.023281: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_frontend: true
2024-04-02 20:30:07.023286: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_nccl_termination_timeout_seconds: -1
2024-04-02 20:30:07.023291: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_shared_constants: true
2024-04-02 20:30:07.023296: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_scratch_max_megabytes: 4096
2024-04-02 20:30:07.023300: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_simplify_all_fp_conversions: true
2024-04-02 20:30:07.023305: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_xla_runtime_executable: true
2024-04-02 20:30:07.023310: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_shape_checks: RUNTIME
2024-04-02 20:30:07.023316: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_normalize_layouts: true
2024-04-02 20:30:07.023321: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_tiling_and_fusion: true
2024-04-02 20:30:07.023325: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_enable_mlir_pretty_form: true
2024-04-02 20:30:07.023330: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_triton_gemm: true
2024-04-02 20:30:07.023334: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_int8x32_convolution_reordering: true
2024-04-02 20:30:07.023339: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_experimental_deallocation: true
2024-04-02 20:30:07.023344: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_fusion_outlining: true
2024-04-02 20:30:07.023351: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_m_dim: 8
2024-04-02 20:30:07.024915: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_n_dim: 8
2024-04-02 20:30:07.024927: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_k_dim: 8
2024-04-02 20:30:07.024932: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_num_runs_to_instantiate: -1
2024-04-02 20:30:07.024937: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_lhs_enable_gpu_async_tracker: true
2024-04-02 20:30:07.024942: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_inflation_factor: 1
2024-04-02 20:30:07.024947: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_min_graph_size: 5
2024-04-02 20:30:07.024951: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reassociation_for_converted_ar: true
2024-04-02 20:30:07.024956: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_gather_combine_threshold_bytes: 31457280
2024-04-02 20:30:07.024961: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_reduce_scatter_combine_threshold_bytes: 31457280
2024-04-02 20:30:07.024965: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_highest_priority_async_stream: true
2024-04-02 20:30:07.024970: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_auto_spmd_partitioning_memory_budget_ratio: 1.1
2024-04-02 20:30:07.024975: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_padding_bytes: 8388608
2024-04-02 20:30:07.024980: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_triton_fusion_level: 2
2024-04-02 20:30:07.024984: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_eviction_timeout_seconds: 60
2024-04-02 20:30:07.024990: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_gpu2_hal: true
2024-04-02 20:30:07.024994: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_copy_insertion_use_region_analysis: true
2024-04-02 20:30:07.024999: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_permute_decomposer_threshold: 9223372036854775807
2024-04-02 20:30:07.025004: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_split_k_autotuning: true
2024-04-02 20:30:07.025009: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduction_epilogue_fusion: true
2024-04-02 20:30:07.025013: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cublas_fallback: true
2024-04-02 20:30:07.025020: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_filter_kernels_spilling_registers_on_autotuning: true
2024-04-02 20:30:07.026867: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_debug_buffer_assignment_show_max: 15
2024-04-02 20:30:07.026875: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_enable_dumping: true
2024-04-02 20:30:07.026880: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_all_gather_combine_by_dim: true
2024-04-02 20:30:07.026885: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduce_scatter_combine_by_dim: true
2024-04-02 20:30:07.026890: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: FUSION
2024-04-02 20:30:07.026895: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: CUBLAS
2024-04-02 20:30:07.026900: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cub_radix_sort: true
2024-04-02 20:30:07.026905: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_memory_limit_slop_factor: 95
2024-04-02 20:30:07.026910: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_threshold_for_windowed_einsum_mib: 100000
2024-04-02 20:30:07.065030: W external/xla/xla/service/gpu/gpu_compiler.cc:549] GpuCompilationEnvironment of hlo_module jit_scatter:
2024-04-02 20:30:07.065103: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_backend_optimization_level: 3
2024-04-02 20:30:07.065111: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_eliminate_hlo_implicit_broadcast: true
2024-04-02 20:30:07.065118: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_multi_thread_eigen: true
2024-04-02 20:30:07.065133: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cuda_data_dir: "./cuda_sdk_lib"
2024-04-02 20:30:07.065138: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_alias_scope_metadata: true
2024-04-02 20:30:07.065143: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_noalias_metadata: true
2024-04-02 20:30:07.065148: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_llvm_enable_invariant_load_metadata: true
2024-04-02 20:30:07.065153: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_force_host_platform_device_count: 1
2024-04-02 20:30:07.065158: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_nans: true
2024-04-02 20:30:07.065162: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_infs: true
2024-04-02 20:30:07.065167: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_allow_excess_precision: true
2024-04-02 20:30:07.065172: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_autotune_level: 4
2024-04-02 20:30:07.065177: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_division: true
2024-04-02 20:30:07.065187: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_fast_math_honor_functions: true
2024-04-02 20:30:07.067717: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_max_hlo_modules: -1
2024-04-02 20:30:07.067726: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_multiheap_size_constraint_per_heap: -1
2024-04-02 20:30:07.067731: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_async_all_reduce: true
2024-04-02 20:30:07.067736: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_strict_conv_algorithm_picker: true
2024-04-02 20:30:07.067741: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_reduce_combine_threshold_bytes: 31457280
2024-04-02 20:30:07.067746: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_frontend: true
2024-04-02 20:30:07.067750: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_nccl_termination_timeout_seconds: -1
2024-04-02 20:30:07.067755: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_shared_constants: true
2024-04-02 20:30:07.067760: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_scratch_max_megabytes: 4096
2024-04-02 20:30:07.067764: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_simplify_all_fp_conversions: true
2024-04-02 20:30:07.067772: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_xla_runtime_executable: true
2024-04-02 20:30:07.069614: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_shape_checks: RUNTIME
2024-04-02 20:30:07.069621: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_normalize_layouts: true
2024-04-02 20:30:07.069627: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_tiling_and_fusion: true
2024-04-02 20:30:07.069632: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_dump_enable_mlir_pretty_form: true
2024-04-02 20:30:07.069637: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_triton_gemm: true
2024-04-02 20:30:07.069642: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cudnn_int8x32_convolution_reordering: true
2024-04-02 20:30:07.069648: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_experimental_deallocation: true
2024-04-02 20:30:07.069653: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_enable_mlir_fusion_outlining: true
2024-04-02 20:30:07.069658: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_m_dim: 8
2024-04-02 20:30:07.069663: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_n_dim: 8
2024-04-02 20:30:07.069668: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_cpu_matmul_tiling_k_dim: 8
2024-04-02 20:30:07.069673: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_num_runs_to_instantiate: -1
2024-04-02 20:30:07.069678: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_lhs_enable_gpu_async_tracker: true
2024-04-02 20:30:07.072042: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_inflation_factor: 1
2024-04-02 20:30:07.072048: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_min_graph_size: 5
2024-04-02 20:30:07.072053: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reassociation_for_converted_ar: true
2024-04-02 20:30:07.072058: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_all_gather_combine_threshold_bytes: 31457280
2024-04-02 20:30:07.072063: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_reduce_scatter_combine_threshold_bytes: 31457280
2024-04-02 20:30:07.072068: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_highest_priority_async_stream: true
2024-04-02 20:30:07.072072: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_auto_spmd_partitioning_memory_budget_ratio: 1.1
2024-04-02 20:30:07.072077: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_redzone_padding_bytes: 8388608
2024-04-02 20:30:07.072081: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_triton_fusion_level: 2
2024-04-02 20:30:07.072086: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_graph_eviction_timeout_seconds: 60
2024-04-02 20:30:07.072091: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_gpu2_hal: true
2024-04-02 20:30:07.072095: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_copy_insertion_use_region_analysis: true
2024-04-02 20:30:07.072104: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_collective_permute_decomposer_threshold: 9223372036854775807
2024-04-02 20:30:07.073627: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_split_k_autotuning: true
2024-04-02 20:30:07.073634: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduction_epilogue_fusion: true
2024-04-02 20:30:07.073639: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_cublas_fallback: true
2024-04-02 20:30:07.073644: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_filter_kernels_spilling_registers_on_autotuning: true
2024-04-02 20:30:07.073649: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_debug_buffer_assignment_show_max: 15
2024-04-02 20:30:07.073654: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_enable_dumping: true
2024-04-02 20:30:07.073659: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_all_gather_combine_by_dim: true
2024-04-02 20:30:07.073664: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_reduce_scatter_combine_by_dim: true
2024-04-02 20:30:07.073669: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: FUSION
2024-04-02 20:30:07.073674: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_command_buffer: CUBLAS
2024-04-02 20:30:07.073679: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_enable_cub_radix_sort: true
2024-04-02 20:30:07.073684: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_memory_limit_slop_factor: 95
2024-04-02 20:30:07.073690: W external/xla/xla/service/gpu/gpu_compiler.cc:549] xla_gpu_threshold_for_windowed_einsum_mib: 100000
infer time： 4.5299530029296875e-06 秒
Output for prompt: The answer to life the universe and everything is of course 42.
But what is the answer to the question of how to get a job in the games industry?
Well, it’s not 42.
It’s not even 42000.
It’s actually 420000.
That’s the number of people who applied for jobs at EA last year.
And that’s just EA.
So how do you get a job in the games industry?
model.properties 0 → 100644
View file @ 049925cc
# 模型唯一标识
modelCode=571
# 模型名称
modelName=grok-1_jax
# 模型描述
modelDescription=Grok-1是由xAI从头开始训练的3140亿个参数混合专家模型。
# 应用场景
appScenario=推理,对话问答,制造,广媒,家居,教育
# 框架类型
frameType=jax
model.py 0 → 100644
View file @ 049925cc
# Copyright 2024 X.AI Corp.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import functools
import logging
import re
from dataclasses import dataclass
from typing import Any, Callable, Dict, List, NamedTuple, Optional, Sequence, Tuple, Union
import haiku as hk
import jax
import jax.experimental.maps
import jax.numpy as jnp
from jax import config, tree_util
from jax.experimental.shard_map import shard_map
from jax.lax import with_sharding_constraint as pjit_sharding_constraint
from jax.sharding import PartitionSpec
from jax.sharding import PartitionSpec as P
config.update("jax_spmd_mode", "allow_all")
logger = logging.getLogger(__name__)
rank_logger = logging.getLogger("rank")
@dataclass
class QuantizedWeight8bit:
weight: jnp.array
scales: jnp.array
@property
def shape(self):
return self.weight.shape
tree_util.register_pytree_node(
QuantizedWeight8bit,
lambda qw: ([qw.weight, qw.scales], ()),
lambda _, children: QuantizedWeight8bit(children[0], children[1]),
)
class TrainingState(NamedTuple):
"""Container for the training state."""
params: hk.Params
def _match(qs, ks):
"""Return True if regexes in qs match any window of strings in tuple ks."""
# compile regexes and force complete match
qts = tuple(map(lambda x: re.compile(x + "$"), qs))
for i in range(len(ks) - len(qs) + 1):
matches = [x.match(y) for x, y in zip(qts, ks[i:])]
if matches and all(matches):
return True
return False
def with_sharding_constraint(x, constraint):
if jax.experimental.maps.thread_resources.env.physical_mesh.empty:
return x
else:
return pjit_sharding_constraint(x, constraint)
def cast_bfloat16(x):
if x.dtype.kind == "f":
return x.astype(jnp.bfloat16)
else:
return x
def ffn_size(emb_size, widening_factor):
_ffn_size = int(widening_factor * emb_size) * 2 // 3
_ffn_size = _ffn_size + (8 - _ffn_size) % 8 # ensure it's a multiple of 8
logger.debug(f"emd_size: {emb_size} adjusted ffn_size: {_ffn_size}")
return _ffn_size
def apply_rules(rules):
def _apply_rules(path, value):
del value # Unused.
path_list = [str(i.key).split("/") for i in path if isinstance(i, jax.tree_util.DictKey)]
flattened_path = jax.tree_util.tree_flatten(path_list)[0]
for rule, replacement in rules:
if _match(rule, flattened_path):
if isinstance(replacement, PartitionSpec):
if "layer_stack" in flattened_path:
replacement = PartitionSpec(None, *replacement)
rank_logger.debug(f"Apply {replacement} to {flattened_path} with rule {rule}")
return replacement
rank_logger.info(f"{flattened_path} no matching found!")
return None
return _apply_rules
TRANSFORMER_PARTITION_RULES = [
# attention
(("multi_head_attention", "(query|key|value)", "w"), P("data", "model")),
(("multi_head_attention", "(query|key|value)", "b"), P(None)),
(("multi_head_attention", "linear", "w"), P("model", "data")),
(("multi_head_attention", "linear", "b"), P(None)),
# mlp
((r"decoder_layer_[0-9]+", "linear", "w"), P("data", "model")),
((r"decoder_layer_[0-9]+", "linear", "b"), P(None)),
((r"decoder_layer_[0-9]+", "linear_v", "w"), P("data", "model")),
((r"decoder_layer_[0-9]+", "linear_v", "b"), P(None)),
(
(r"decoder_layer_[0-9]+", "linear_1", "w"),
P(
"model",
"data",
),
),
((r"decoder_layer_[0-9]+", "linear_1", "b"), P(None)),
# layer norms
((r"decoder_layer_[0-9]+", "layer_norm", "offset"), P(None)),
((r"decoder_layer_[0-9]+", "layer_norm", "scale"), P(None)),
((r"decoder_layer_[0-9]+", "layer_norm_1", "offset"), P(None)),
((r"decoder_layer_[0-9]+", "layer_norm_1", "scale"), P(None)),
# rms norms
((r"decoder_layer_[0-9]+", "rms_norm", "scale"), P(None)),
((r"decoder_layer_[0-9]+", "rms_norm_1", "scale"), P(None)),
((r"decoder_layer_[0-9]+", "rms_norm_2", "scale"), P(None)),
((r"decoder_layer_[0-9]+", "rms_norm_3", "scale"), P(None)),
# router
(("router", "w"), P("data")),
# moe mlp
(("moe", "linear", "w"), P(None, "data", "model")),
(("moe", "linear", "b"), P(None)),
(("moe", "linear_v", "w"), P(None, "data", "model")),
(("moe", "linear_v", "b"), P(None)),
(("moe", "linear_1", "w"), P(None, "model", "data")),
(("moe", "linear_1", "b"), P(None)),
# layer norms
(("moe", "layer_norm", "offset"), P(None)),
(("moe", "layer_norm", "scale"), P(None)),
(("moe", "layer_norm_1", "offset"), P(None)),
(("moe", "layer_norm_1", "scale"), P(None)),
# rms norms
(("moe", "rms_norm", "scale"), P(None)),
(("moe", "rms_norm_1", "scale"), P(None)),
(("moe", "rms_norm_2", "scale"), P(None)),
(("moe", "rms_norm_3", "scale"), P(None)),
]
LM_PARTITION_RULES = [
# Embedding layer.
(
("language_model", "positional_embeddings"),
P(None, ("data", "model")),
),
(
("language_model", "in_out_embed", "embeddings"),
P(None, ("data", "model")),
),
# Final RMSNorm.
(("language_model", "rms_norm"), P(None)),
]
TOP_K = 8
class KVMemory(NamedTuple):
k: Optional[jax.Array]
v: Optional[jax.Array]
step: Optional[jax.Array]
def init_layer_memories(
batch_size: int,
sequence_len: int,
num_kv_heads: int,
key_size: int,
num_layers: int,
step: Optional[jax.Array] = None,
dtype=jnp.bfloat16,
):
return [
KVMemory(
k=jnp.zeros((batch_size, sequence_len, num_kv_heads, key_size), dtype=dtype),
v=jnp.zeros((batch_size, sequence_len, num_kv_heads, key_size), dtype=dtype),
step=step,
)
for _ in range(num_layers)
]
class Memory(NamedTuple):
# Self-attention key/value cache.
layers: List[KVMemory]
class Router(hk.Module):
def __init__(
self,
num_selected_experts: int,
data_axis: Union[str, Tuple[str, ...]] = "data",
model_axis: Union[str, Tuple[str, ...]] = "model",
shard_activations: bool = False,
mesh: Any = None,
name: str = "router",
):
super().__init__(name)
self.shard_activations = shard_activations
self.data_axis = data_axis
self.model_axis = model_axis
self.mesh = mesh
self.num_selected_experts = num_selected_experts
def compute_routing_prob(
self, inputs: jax.Array, padding_mask: Optional[jax.Array], num_experts: int
):
return self._compute_routing_prob(inputs, padding_mask, num_experts)
@hk.transparent
def _compute_routing_prob(
self,
inputs: jax.Array,
padding_mask: Optional[jax.Array],
num_experts: int,
):
# Using fp32 for the routing prob computation.
inputs = jax.lax.convert_element_type(inputs, jnp.float32)
# [batch_size, seq_len, num_experts]
routing_logits = self._router_weights(inputs, num_experts, sharding=P("data"))
assert routing_logits.dtype == jnp.float32
routing_probs = jax.nn.softmax(routing_logits)
if padding_mask is not None:
routing_probs *= padding_mask
return routing_probs, routing_logits, 0
@hk.transparent
def _router_weights(
self,
x: jax.Array,
num_experts: int,
sharding: Optional[P] = None,
):
fprop_dtype = x.dtype
if not x.shape:
raise ValueError("Input must not be scalar.")
input_size = self.input_size = x.shape[-1]
w = hk.get_parameter(
"w", [input_size, num_experts], jnp.float32, init=hk.initializers.Constant(0)
)
if sharding:
w = with_sharding_constraint(w, sharding)
out = jnp.dot(x, w.astype(fprop_dtype))
return out
class MoELayer(hk.Module):
def __init__(
self,
num_experts: int,
layer_fn: Callable,
router: Router,
mesh: Any = None,
shard_activations: bool = False,
data_axis: Union[str, Tuple[str, ...]] = "data",
model_axis: Union[str, Tuple[str, ...]] = "model",
name: Optional[str] = "moe",
):
super().__init__(name)
self.num_experts = num_experts
self.layer_fn = layer_fn
self.router = router
self.mesh = mesh
self.shard_activations = shard_activations
self.data_axis = data_axis
self.model_axis = model_axis
@hk.transparent
def _inference_call(self, inputs: jax.Array, padding_mask: Optional[jax.Array] = None):
routing_probs, _, _ = self.router.compute_routing_prob(
inputs, padding_mask, self.num_experts
)
expert_gate, expert_index = jax.lax.top_k(routing_probs, k=self.router.num_selected_experts)
tmp = jnp.reshape(inputs, (inputs.shape[0] * inputs.shape[1], inputs.shape[2]))
broad_inputs = jnp.tile(tmp[:, jnp.newaxis, :], (1, self.router.num_selected_experts, 1))
broad_inputs = jnp.reshape(
broad_inputs, (broad_inputs.shape[0] * broad_inputs.shape[1], broad_inputs.shape[2])
)
init_fn, _ = hk.transform(self.layer_fn)
vmapped_init_fn = jax.vmap(init_fn, in_axes=0, out_axes=0)
lifted_init_fn = hk.experimental.transparent_lift(vmapped_init_fn)
# Fetch the vmapped params of the DenseBlock.
params = lifted_init_fn(
jax.random.split(jax.random.PRNGKey(1), self.num_experts),
jnp.zeros((self.num_experts, 1, 1, inputs.shape[-1])),
)
# Index and prob are in the shape [m, 2] indicating which token assigned to which experts.
# b: num_expert
# m: token or sequence dim
# k: input embed dim
# n: output embed dim
# e: the number of experts chosen for each token
@functools.partial(
shard_map,
mesh=self.mesh,
in_specs=(
P(self.data_axis, None),
P(None, None, self.model_axis),
P(None, None, self.model_axis),
P(None),
P(None),
),
out_specs=P(self.data_axis, self.model_axis),
check_rep=False,
)
def moe_slow_matmul1(input, weight, scales, index, prob):
weight = weight * scales
one_hot_indices = jax.nn.one_hot(index.reshape(-1), 8, axis=0)
all_expert_output = jnp.einsum("mk,bkn->bmn", input, weight)
output = jnp.einsum("bm,bmn->mn", one_hot_indices, all_expert_output)
return output
@functools.partial(
shard_map,
mesh=self.mesh,
in_specs=(
P(self.data_axis, self.model_axis),
P(None, self.model_axis, None),
P(None, self.model_axis, None),
P(None),
P(None),
),
out_specs=P(self.data_axis, None),
check_rep=False,
)
def moe_slow_matmul2(input, weight, scales, index, prob):
weight = weight * scales
one_hot_indices = jax.nn.one_hot(index.reshape(-1), 8, axis=0)
all_expert_output = jnp.einsum("mk,bkn->bmn", input, weight)
output = jnp.einsum("bm,bmn->mn", one_hot_indices, all_expert_output)
return jax.lax.psum(output, axis_name="model")
if hasattr(params["linear"]["w"], "scales"):
x = moe_slow_matmul1(
broad_inputs,
params["linear_v"]["w"].weight,
params["linear_v"]["w"].scales,
expert_index,
expert_gate,
)
y = moe_slow_matmul1(
broad_inputs,
params["linear"]["w"].weight,
params["linear"]["w"].scales,
expert_index,
expert_gate,
)
y = jax.nn.gelu(y)
out = moe_slow_matmul2(
x * y,
params["linear_1"]["w"].weight,
params["linear_1"]["w"].scales,
expert_index,
expert_gate,
)
out = jnp.reshape(
out,
[
inputs.shape[0],
inputs.shape[1],
self.router.num_selected_experts,
out.shape[-1],
],
)
out = expert_gate[:, :, :, None].astype(jnp.bfloat16) * out
out = jnp.sum(out, axis=2)
out = out.astype(jnp.bfloat16)
else:
# This is only here so that we can construct a valid init_fn with this code.
return inputs
return out
def __call__(self, inputs: jax.Array, padding_mask: jax.Array):
return self._inference_call(inputs)
class MHAOutput(NamedTuple):
"""Outputs of the multi-head attention operation."""
embeddings: jax.Array
memory: Any
class DecoderOutput(NamedTuple):
embeddings: jax.Array
memory: Any
class TransformerOutput(NamedTuple):
embeddings: jax.Array
memory: Any
@dataclass
class TransformerConfig:
emb_size: int
key_size: int
num_q_heads: int
num_kv_heads: int
num_layers: int
vocab_size: int = 128 * 1024
widening_factor: float = 4.0
attn_output_multiplier: float = 1.0
name: Optional[str] = None
num_experts: int = -1
capacity_factor: float = 1.0
num_selected_experts: int = 1
init_scale: float = 1.0
shard_activations: bool = False
# Used for activation sharding.
data_axis: Union[str, Tuple[str, ...]] = "data"
model_axis: Union[str, Tuple[str, ...]] = "model"
def __post_init__(self):
if isinstance(self.data_axis, list):
self.data_axis = tuple(self.data_axis)
if isinstance(self.model_axis, list):
self.model_axis = tuple(self.model_axis)
def partition_rules(self):
return TRANSFORMER_PARTITION_RULES
def make(self, mesh=None) -> "Transformer":
data_axis = tuple(self.data_axis) if isinstance(self.data_axis, list) else self.data_axis
model_axis = (
tuple(self.model_axis) if isinstance(self.model_axis, list) else self.model_axis
)
return Transformer(
num_q_heads=self.num_q_heads,
num_kv_heads=self.num_kv_heads,
widening_factor=self.widening_factor,
key_size=self.key_size,
init_scale=self.init_scale,
mesh=mesh,
attn_output_multiplier=self.attn_output_multiplier,
shard_activations=self.shard_activations,
num_layers=self.num_layers,
num_experts=self.num_experts,
num_selected_experts=self.num_selected_experts,
data_axis=data_axis,
model_axis=model_axis,
)
def get_memory_sharding(self):
return Memory(
layers=[
KVMemory(
k=P(self.data_axis, self.model_axis),
v=P(self.data_axis, self.model_axis),
step=P(self.data_axis),
)
for _ in range(self.num_layers)
],
)
def hk_rms_norm(
x: jax.Array,
fixed_scale=False,
sharding=P(None),
) -> jax.Array:
"""Applies a unique LayerNorm to x with default settings."""
ln = RMSNorm(axis=-1, create_scale=not fixed_scale, sharding=sharding)
return ln(x)
def make_attention_mask(
query_input: jax.Array,
key_input: jax.Array,
pairwise_fn: Callable[..., Any] = jnp.multiply,
dtype: Any = jnp.bfloat16,
):
"""Mask-making helper for attention weights.
In case of 1d inputs (i.e., `[batch..., len_q]`, `[batch..., len_kv]`, the
attention weights will be `[batch..., heads, len_q, len_kv]` and this
function will produce `[batch..., 1, len_q, len_kv]`.
Args:
query_input: a batched, flat input of query_length size
key_input: a batched, flat input of key_length size
pairwise_fn: broadcasting elementwise comparison function
dtype: mask return dtype
Returns:
A `[batch..., 1, len_q, len_kv]` shaped mask for 1d attention.
"""
mask = pairwise_fn(jnp.expand_dims(query_input, axis=-1), jnp.expand_dims(key_input, axis=-2))
mask = jnp.expand_dims(mask, axis=-3)
return mask.astype(dtype)
class Linear(hk.Linear):
def __init__(
self,
output_size: int,
with_bias: bool = True,
sharding: Optional[P] = None,
mesh: Any = None,
name: Optional[str] = None,
shard_axis: int = 0,
):
super().__init__(
output_size=output_size,
with_bias=with_bias,
name=name,
)
self.sharding = sharding
self.mesh = mesh
self.shard_axis = shard_axis
def __call__(
self,
inputs: jax.Array,
) -> jax.Array:
"""Computes a linear transform of the input."""
fprop_dtype = inputs.dtype
if not inputs.shape:
raise ValueError("Input must not be scalar.")
input_size = self.input_size = inputs.shape[-1]
output_size = self.output_size
w = hk.get_parameter(
"w", [input_size, output_size], jnp.float32, init=hk.initializers.Constant(0)
)
if hasattr(w, "scales"):
shape = inputs.shape
inputs = jnp.reshape(inputs, (-1, shape[-1]))
@functools.partial(
shard_map,
mesh=self.mesh,
in_specs=(self.sharding, self.sharding),
out_specs=self.sharding,
check_rep=False,
)
def mul(w, s):
return w.astype(s.dtype) * s
w = mul(w.weight, w.scales)
out = jnp.dot(inputs, w.astype(fprop_dtype))
if self.with_bias:
b = hk.get_parameter(
"b", [self.output_size], jnp.float32, init=hk.initializers.Constant(0)
)
b = jnp.broadcast_to(b, out.shape)
out = out + b.astype(fprop_dtype)
return out
class RMSNorm(hk.RMSNorm):
def __init__(
self,
axis: Union[int, Sequence[int], slice],
eps: float = 1e-5,
name: Optional[str] = None,
create_scale: bool = True,
sharding: Optional[P] = None,
):
super().__init__(axis, eps, create_scale=create_scale, name=name)
self.sharding = sharding
def __call__(self, inputs: jax.Array):
fprop_dtype = inputs.dtype
param_shape = (inputs.shape[-1],)
if self.create_scale:
scale = hk.get_parameter(
"scale",
param_shape,
dtype=jnp.float32,
init=hk.initializers.Constant(0),
)
if self.sharding:
scale = with_sharding_constraint(scale, self.sharding)
scale = jnp.broadcast_to(scale.astype(jnp.float32), inputs.shape)
else:
scale = 1.0
inputs = inputs.astype(jnp.float32)
scale = scale.astype(jnp.float32)
mean_squared = jnp.mean(jnp.square(inputs), axis=[-1], keepdims=True)
mean_squared = jnp.broadcast_to(mean_squared, inputs.shape)
normed_inputs = inputs * jax.lax.rsqrt(mean_squared + self.eps)
outputs = scale * normed_inputs
return outputs.astype(fprop_dtype)
def rotate_half(
x: jax.Array,
) -> jax.Array:
"""Obtain the rotated counterpart of each feature"""
x1, x2 = jnp.split(x, 2, axis=-1)
return jnp.concatenate((-x2, x1), axis=-1)
class RotaryEmbedding(hk.Module):
"""Applies rotary embeddings (RoPE) to the input sequence tensor,
as described in https://arxiv.org/abs/2104.09864.
Attributes:
dim (int): Dimensionality of the feature vectors
base_exponent (int): Base exponent to compute embeddings from
"""
def __init__(
self,
dim: int,
name: Optional[str] = None,
base_exponent: int = 10000,
):
super().__init__(name)
self.dim = dim
self.base_exponent = base_exponent
assert self.dim % 2 == 0
def __call__(
self,
x: jax.Array,
seq_dim: int,
offset: jax.Array,
const_position: Optional[int] = None,
t: Optional[jax.Array] = None,
) -> jax.Array:
fprop_dtype = x.dtype
# Compute the per-dimension frequencies
exponents = jnp.arange(0, self.dim, 2, dtype=jnp.float32)
inv_freq = jnp.asarray(
1.0 / (self.base_exponent ** (exponents / self.dim)), dtype=jnp.float32
)
if jnp.shape(offset) == ():
# Offset can be a scalar or one offset per batch element.
offset = jnp.expand_dims(offset, 0)
# Compute the per element phase (to pass into sin and cos)
if const_position:
t = const_position * jnp.ones(
(
1,
x.shape[seq_dim],
),
dtype=jnp.float32,
)
elif t is None:
t = jnp.arange(x.shape[seq_dim], dtype=jnp.float32) + jnp.expand_dims(offset, -1)
phase = jnp.einsum("bi,j->bij", t, inv_freq)
phase = jnp.tile(phase, reps=(1, 2))[:, :, None, :]
x = x * jnp.cos(phase) + rotate_half(x) * jnp.sin(phase)
x = x.astype(fprop_dtype)
return x
class MultiHeadAttention(hk.Module):
def __init__(
self,
num_q_heads: int,
num_kv_heads: int,
key_size: int,
*,
with_bias: bool = True,
value_size: Optional[int] = None,
model_size: Optional[int] = None,
attn_output_multiplier: 1.0,
data_axis: Union[str, Tuple[str, ...]] = "data",
model_axis: Union[str, Tuple[str, ...]] = "model",
name: Optional[str] = None,
):
super().__init__(name=name)
self.num_q_heads = num_q_heads
self.num_kv_heads = num_kv_heads
self.key_size = key_size
self.value_size = value_size or key_size
self.model_size = model_size or key_size * num_q_heads
self.data_axis = data_axis
self.model_axis = model_axis
self.attn_output_multiplier = attn_output_multiplier
self.with_bias = with_bias
def __call__(
self,
query: jax.Array,
key: Optional[jax.Array],
value: Optional[jax.Array],
mask: Optional[jax.Array] = None,
kv_memory: Optional[KVMemory] = None,
mesh: Any = None,
) -> MHAOutput:
# In shape hints below, we suppress the leading dims [...] for brevity.
# Hence e.g. [A, B] should be read in every case as [..., A, B].
sequence_length = query.shape[1]
projection = self._linear_projection
use_memory = False
if kv_memory is not None:
if kv_memory.k is None:
assert kv_memory.v is None
assert key is not None
assert value is not None
else:
assert kv_memory.v is not None
use_memory = True
else:
assert key is not None
assert value is not None
# Check that the keys and values have consistent batch size and sequence length.
if not use_memory:
assert key.shape[:2] == value.shape[:2], f"key/value shape: {key.shape}/{value.shape}"
if mask is not None:
assert mask.ndim == 4
assert mask.shape[0] in {
1,
query.shape[0],
}, f"mask/query shape: {mask.shape}/{query.shape}"
if not use_memory:
assert key.shape[0] in {
1,
query.shape[0],
}, f"key/query shape: {key.shape}/{query.shape}"
assert mask.shape[1] == 1
assert mask.shape[2] in {
1,
query.shape[1],
}, f"mask/query shape: {mask.shape}/{query.shape}"
if not use_memory:
assert mask.shape[3] in {
1,
key.shape[1],
}, f"mask/query shape: {mask.shape}/{key.shape}"
# Compute key/query/values (overload K/Q/V to denote the respective sizes).
assert self.num_q_heads % self.num_kv_heads == 0
query_heads = projection(
query,
self.key_size,
self.num_q_heads,
name="query",
sharding=P("data", "model"),
mesh=mesh,
) # [B, T', H, Q=K]
new_memory = None
key_heads = projection(
key,
self.key_size,
self.num_kv_heads,
name="key",
sharding=P("data", "model"),
mesh=mesh,
) # [B, T, H, K]
value_heads = projection(
value,
self.value_size,
self.num_kv_heads,
name="value",
sharding=P("data", "model"),
mesh=mesh,
) # [B, T, H, V]
rotate = RotaryEmbedding(dim=self.key_size, base_exponent=int(1e4))
key_heads = rotate(key_heads, seq_dim=1, offset=(kv_memory.step if kv_memory else 0))
query_heads = rotate(query_heads, seq_dim=1, offset=(kv_memory.step if kv_memory else 0))
@functools.partial(jax.vmap)
def update_into(mem, start, update):
return jax.lax.dynamic_update_slice_in_dim(mem, update, start, axis=0)
if kv_memory:
if mesh is not None:
@functools.partial(
shard_map,
mesh=mesh,
in_specs=(
P("data", None, "model"),
P("data"),
P("data", None, "model"),
),
out_specs=P("data", None, "model"),
check_rep=False,
)
def update_into_shmap(mems, starts, updates):
return update_into(mems, starts, updates)
key_heads = update_into_shmap(kv_memory.k, kv_memory.step, key_heads)
value_heads = update_into_shmap(kv_memory.v, kv_memory.step, value_heads)
else:
key_heads = update_into(kv_memory.k, kv_memory.step, key_heads)
value_heads = update_into(kv_memory.v, kv_memory.step, value_heads)
new_step = kv_memory.step + sequence_length
memory_mask = jnp.arange(kv_memory.k.shape[1]) < new_step[:, None]
memory_mask = memory_mask[:, None, None, :] # [B, H, T, T]
if mask is not None:
mask = memory_mask * mask
else:
mask = memory_mask
new_memory = KVMemory(
k=key_heads,
v=value_heads,
step=new_step,
)
# Add separate dimension for grouped query heads.
query_heads = with_sharding_constraint(query_heads, P(self.data_axis, None, "model", None))
key_heads = with_sharding_constraint(key_heads, P(self.data_axis, None, "model", None))
value_heads = with_sharding_constraint(value_heads, P(self.data_axis, None, "model", None))
b, t, h, d = query_heads.shape
_, _, kv_h, _ = key_heads.shape
assert h % kv_h == 0, f"query_heads {h} must be a multiple of kv_heads {kv_h}"
query_heads = jnp.reshape(query_heads, (b, t, kv_h, h // kv_h, d))
query_heads = with_sharding_constraint(
query_heads, P(self.data_axis, None, "model", None, None)
)
# Compute attention weights.
# Attention softmax is always carried out in fp32.
attn_logits = jnp.einsum("...thHd,...Thd->...hHtT", query_heads, key_heads).astype(
jnp.float32
)
attn_logits *= self.attn_output_multiplier
max_attn_val = jnp.array(30.0, dtype=attn_logits.dtype)
attn_logits = max_attn_val * jnp.tanh(attn_logits / max_attn_val)
mask = mask[:, :, None, :, :]
if mask is not None:
if mask.ndim != attn_logits.ndim:
raise ValueError(
f"Mask dimensionality {mask.ndim} must match logits dimensionality "
f"{attn_logits.ndim} for {mask.shape}/{attn_logits.shape}."
)
attn_logits = jnp.where(mask, attn_logits, -1e30)
attn_weights = jax.nn.softmax(attn_logits).astype(query.dtype) # [H, T', T]
# Weight the values by the attention and flatten the head vectors.
attn = jnp.einsum("...hHtT,...Thd->...thHd", attn_weights, value_heads)
attn = with_sharding_constraint(attn, P(self.data_axis, None, "model", None, None))
leading_dims = attn.shape[:2]
attn = jnp.reshape(attn, (*leading_dims, -1)) # [T', H*V]
attn = with_sharding_constraint(attn, P(self.data_axis, None, "model"))
# Apply another projection to get the final embeddings.
final_projection = Linear(
self.model_size,
with_bias=False,
sharding=P("model", "data"),
mesh=mesh,
)
return MHAOutput(final_projection(attn), new_memory)
@hk.transparent
def _linear_projection(
self,
x: jax.Array,
head_size: int,
num_heads: int,
sharding: Optional[P] = None,
name: Optional[str] = None,
mesh: Any = None,
) -> jax.Array:
y = Linear(
num_heads * head_size,
with_bias=False,
name=name,
sharding=sharding,
mesh=mesh,
)(x)
*leading_dims, _ = x.shape
return y.reshape((*leading_dims, num_heads, head_size))
@dataclass
class MHABlock(hk.Module):
"""A MHA Block"""
num_q_heads: int
num_kv_heads: int
key_size: int
attn_output_multiplier: float = 1.0
mesh: Any = None
data_axis: Union[str, Tuple[str, ...]] = "data"
model_axis: Union[str, Tuple[str, ...]] = "model"
@hk.transparent
def __call__(
self,
inputs: jax.Array, # [B, T, D]
mask: jax.Array, # [B, 1, T, T] or [B, 1, 1, T] or B[1, 1, 1, 1]
layer_memory: Optional[KVMemory],
) -> MHAOutput:
_, _, model_size = inputs.shape
assert mask.ndim == 4, f"shape: {mask.shape}"
assert mask.shape[2] in {1, inputs.shape[1]}, str(mask.shape)
assert mask.shape[3] in {1, inputs.shape[1]}, str(mask.shape)
side_input = inputs
def attn_block(query, key, value, mask, memory) -> MHAOutput:
return MultiHeadAttention(
num_q_heads=self.num_q_heads,
num_kv_heads=self.num_kv_heads,
key_size=self.key_size,
model_size=model_size,
data_axis=self.data_axis,
model_axis=self.model_axis,
attn_output_multiplier=self.attn_output_multiplier,
)(
query,
key,
value,
mask,
memory,
mesh=self.mesh,
)
attn_output = attn_block(inputs, side_input, side_input, mask, layer_memory)
h_attn = attn_output.embeddings
return attn_output._replace(embeddings=h_attn)
@dataclass
class DenseBlock(hk.Module):
num_q_heads: int
num_kv_heads: int
key_size: int
widening_factor: float = 4.0
sharding_constraint: bool = False
mesh: Any = None
@hk.transparent
def __call__(
self,
inputs: jax.Array, # [B, T, D]
) -> jax.Array: # [B, T, D]
_, _, model_size = inputs.shape
h_v = Linear(
ffn_size(
model_size,
self.widening_factor,
),
with_bias=False,
mesh=self.mesh,
sharding=P("data", "model"),
name="linear_v",
)(inputs)
h_w1 = jax.nn.gelu(
Linear(
ffn_size(
model_size,
self.widening_factor,
),
with_bias=False,
mesh=self.mesh,
sharding=P("data", "model"),
)(inputs)
)
h_dense = Linear(
model_size,
with_bias=False,
sharding=P("model", "data"),
mesh=self.mesh,
shard_axis=1,
)(h_w1 * h_v)
return h_dense
@dataclass
class DecoderLayer(hk.Module):
"""A transformer stack."""
num_q_heads: int
num_kv_heads: int
key_size: int
num_layers: int
# MoE.
num_experts: int
layer_index: Optional[int] = None
num_selected_experts: int = 1
widening_factor: float = 4.0
name: Optional[str] = None
data_axis: Union[str, Tuple[str, ...]] = "data"
model_axis: Union[str, Tuple[str, ...]] = "model"
shard_activations: bool = False
attn_output_multiplier: float = 1.0
mesh: Any = None
def __call__(
self,
inputs: jax.Array, # [B, T, D]
mask: jax.Array, # [B, 1, T, T] or [B, 1, 1, T]
padding_mask: Optional[jax.Array],
layer_memory: Optional[KVMemory],
) -> DecoderOutput:
"""Transforms input embedding sequences to output embedding sequences."""
def layer_norm(x):
return hk_rms_norm(x)
if self.shard_activations:
sharding = P(self.data_axis, None, self.model_axis)
else:
sharding = P(self.data_axis, None)
h = with_sharding_constraint(inputs, sharding)
attn_output = MHABlock(
num_q_heads=self.num_q_heads,
num_kv_heads=self.num_kv_heads,
key_size=self.key_size,
attn_output_multiplier=self.attn_output_multiplier,
mesh=self.mesh,
data_axis=self.data_axis,
model_axis=self.model_axis,
)(layer_norm(h), mask, layer_memory)
h_attn = attn_output.embeddings
h_attn = layer_norm(h_attn)
h += h_attn
h = with_sharding_constraint(h, sharding)
def base_dense_block(h):
h = DenseBlock(
num_q_heads=self.num_q_heads,
num_kv_heads=self.num_kv_heads,
key_size=self.key_size,
widening_factor=self.widening_factor,
sharding_constraint=False,
mesh=self.mesh,
)(h)
return h
if self.num_experts > 1:
rank_logger.debug("Using MoE!")
router = Router(
num_selected_experts=self.num_selected_experts,
shard_activations=self.shard_activations,
data_axis=self.data_axis,
model_axis=self.model_axis,
mesh=self.mesh,
)
h_dense = MoELayer(
num_experts=self.num_experts,
mesh=self.mesh,
layer_fn=base_dense_block,
router=router,
shard_activations=self.shard_activations,
data_axis=self.data_axis,
model_axis=self.model_axis,
)(layer_norm(h), padding_mask)
else:
h_dense = base_dense_block(layer_norm(h))
h_dense = layer_norm(h_dense)
h += h_dense
h = with_sharding_constraint(h, sharding)
return DecoderOutput(
embeddings=h,
memory=attn_output.memory,
)
class LanguageModelOutput(NamedTuple):
logits: jax.Array
model_state: Any
class InOutEmbed(hk.Embed):
"""Module for embedding tokens in a low-dimensional space."""
def __init__(
self,
vocab_size: Optional[int] = None,
embed_dim: Optional[int] = None,
sharding: Optional[P] = None,
name: Optional[str] = None,
):
super().__init__(
vocab_size=vocab_size,
embed_dim=embed_dim,
name=name,
)
self.sharding = sharding
@property
def embeddings(self):
embed_mat = hk.get_parameter(
"embeddings",
[self.vocab_size, self.embed_dim],
dtype=jnp.float32,
init=hk.initializers.Constant(0),
)
if self.sharding:
embed_mat = with_sharding_constraint(embed_mat, self.sharding)
return embed_mat
def decode(
self,
inputs: jax.Array,
) -> jax.Array:
return jnp.dot(inputs, self.embeddings.T.astype(inputs.dtype))
@dataclass
class LanguageModelConfig:
"""An autoregressive transformer-based language model."""
model: Optional[TransformerConfig]
vocab_size: int
pad_token: int
eos_token: int
sequence_len: int
model_size: int = 0
embedding_init_scale: float = 1.0
embedding_multiplier_scale: float = 1.0
output_multiplier_scale: float = 1.0
name: Optional[str] = None
fprop_dtype: Any = jnp.bfloat16
model_type: Optional[str] = None
init_scale_override: Optional[float] = None
shard_embeddings: bool = True
_initialized = False
def initialize(self):
# We cannot specify [] as a default value (it is mutable), hence None.
model_config = self.model
assert self.init_scale_override is None, (
"Overriding model initialize scale is supported only for predefined models."
)
if self.model_size == 0:
self.model_size = model_config.emb_size
assert self.model is not None, "Model could not be initialized."
self._initialized = True
return self
def make(self, *args, **kwargs):
if not self._initialized:
logger.warning(
f"LanguageModel {self.name} is not initialized. Initializing for one replica."
)
self.initialize()
return LanguageModel(
model=self.model.make(*args, **kwargs),
config=self,
fprop_dtype=self.fprop_dtype,
mesh=kwargs.get("mesh", None),
)
def partition_rules(self):
return LM_PARTITION_RULES + self.model.partition_rules()
def layer_norm(x, model):
return hk_rms_norm(x)
@dataclass
class LanguageModel(hk.Module):
"""An autoregressive transformer-based language model."""
model: "Transformer"
config: LanguageModelConfig
fprop_dtype: Any = jnp.bfloat16
name: Optional[str] = None
mesh: Any = None
def __call__(
self,
tokens: jax.Array,
memory: Optional[Memory] = None,
*,
batch: Dict[str, jax.Array] = {},
last_hid_only: bool = False,
length: Optional[jax.Array] = None,
) -> LanguageModelOutput:
"""Forward pass, producing a sequence of logits."""
del batch # Unused.
config = self.config
input_mask = jnp.greater(tokens, config.pad_token)
# Embed the input tokens and positions.
in_out_embed = InOutEmbed(
self.config.vocab_size,
embed_dim=self.config.model_size,
sharding=P(None, ("data", "model")),
)
input_embeddings = in_out_embed(tokens).astype(config.fprop_dtype)
input_embeddings = with_sharding_constraint(
input_embeddings, P("data", None, self.model.model_axis)
)
input_embeddings *= config.embedding_multiplier_scale
model_output = self.model(
input_embeddings,
input_mask,
memory=memory,
) # [B, T, D]
embeddings, model_state = model_output.embeddings, model_output.memory
if self.model.shard_activations:
embeddings = with_sharding_constraint(
embeddings, P("data", None, self.model.model_axis)
)
else:
embeddings = with_sharding_constraint(embeddings, P("data", None))
rank_logger.debug(f"Final embedding shape: {embeddings.shape}")
embeddings = layer_norm(embeddings, self.model)
assert embeddings.dtype == self.fprop_dtype
if last_hid_only:
last_step = jnp.maximum(jnp.sum(input_mask.astype(jnp.int32), axis=1) - 1, 0)
last_hid = jax.vmap(lambda x, i: x[i], in_axes=0, out_axes=0)(embeddings, last_step)
return last_hid
if length is not None:
last_step = jnp.maximum(length.astype(jnp.int32) - 1, 0)
embeddings = jax.vmap(lambda x, i: x[i], in_axes=0, out_axes=0)(embeddings, last_step)
embeddings = jnp.expand_dims(embeddings, axis=1)
# Decode the embeddings (here, we use tied weights).
rank_logger.info(embeddings.shape)
out = in_out_embed.decode(embeddings)
rank_logger.info(out.shape)
out *= config.output_multiplier_scale
if self.model.shard_activations:
out = with_sharding_constraint(out, P("data", None, self.model.model_axis))
else:
out = with_sharding_constraint(out, P("data", None))
return LanguageModelOutput(
logits=out,
model_state=model_state,
)
def init_memory(self, batch_size: int, seq_len: int, dtype=jnp.bfloat16):
return self.model.init_memory(batch_size=batch_size, sequence_len=seq_len, dtype=dtype)
def prefill_memory(self, prompts, memory):
# Pad to the left and right align?
# Basically assume prompt is already padded
model_output = self(prompts, memory=memory)
return model_output.logits, model_output.model_state
@dataclass
class Transformer(hk.Module):
"""A transformer stack."""
num_q_heads: int
num_kv_heads: int
key_size: int
widening_factor: float
init_scale: float
mesh: Any
attn_output_multiplier: float
shard_activations: bool
num_layers: int
# MoE
num_experts: int
num_selected_experts: int
name: Optional[str] = None
# Used for activation sharding
data_axis: Union[str, Tuple[str, ...]] = "data"
model_axis: Union[str, Tuple[str, ...]] = "model"
def init_memory(self, batch_size: int, sequence_len: int, dtype=jnp.bfloat16):
return Memory(
layers=init_layer_memories(
batch_size,
sequence_len,
self.num_kv_heads,
self.key_size,
self.num_layers,
step=jnp.zeros(batch_size, dtype=jnp.int32),
dtype=dtype,
),
)
def __call__(
self,
embeddings: jax.Array, # [B, T, D]
mask: jax.Array, # [B, T]
memory: Optional[Memory],
) -> TransformerOutput:
"""Transforms input embedding sequences to output embedding sequences."""
fprop_dtype = embeddings.dtype
_, seq_len, model_size = embeddings.shape
padding_mask = mask.copy()
mask = mask[:, None, None, :] # [B, H=1, T'=1, T]
# Compute causal mask for autoregressive sequence modelling.
causal_mask = jnp.tril(jnp.ones((1, 1, seq_len, seq_len))).astype(
fprop_dtype
) # [B=1, H=1, T, T]
mask = mask * causal_mask # [B, H=1, T, T]
h = embeddings
kv_memories = []
def block(
h,
mask,
padding_mask,
memory,
layer_index: Optional[int] = None,
widening_factor: Optional[int] = None,
name: Optional[str] = None,
) -> DecoderOutput:
return DecoderLayer(
num_q_heads=self.num_q_heads,
num_kv_heads=self.num_kv_heads,
key_size=self.key_size,
widening_factor=widening_factor or self.widening_factor,
num_layers=self.num_layers,
mesh=self.mesh,
data_axis=self.data_axis,
model_axis=self.model_axis,
attn_output_multiplier=self.attn_output_multiplier,
shard_activations=self.shard_activations,
# MoE.
num_experts=self.num_experts,
num_selected_experts=self.num_selected_experts,
name=name,
layer_index=layer_index,
)(
h,
mask,
padding_mask,
memory,
)
for i in range(self.num_layers):
decoder_output = block(
h,
mask,
padding_mask,
memory.layers[i] if memory else None,
layer_index=i,
name=f"decoder_layer_{i}",
)
h, new_kv_memory = (
decoder_output.embeddings,
decoder_output.memory,
)
kv_memories.append(new_kv_memory)
return TransformerOutput(
embeddings=h,
memory=Memory(layers=kv_memories),
)
pyproject.toml 0 → 100644
View file @ 049925cc
[tool.ruff]
indent-width = 4
line-length = 100
[tool.ruff.lint]
ignore = [
"E722",
"E731",
"E741",
"F405",
"E402",
"F403",
]
select = ["ISC001"]
run.py 0 → 100644
View file @ 049925cc
# Copyright 2024 X.AI Corp.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import time
import logging
from model import LanguageModelConfig, TransformerConfig, QuantizedWeight8bit as QW8Bit
from runners import InferenceRunner, ModelRunner, sample_from_model
CKPT_PATH = "./checkpoints/"
def main():
grok_1_model = LanguageModelConfig(
vocab_size=128 * 1024,
pad_token=0,
eos_token=2,
sequence_len=8192,
embedding_init_scale=1.0,
output_multiplier_scale=0.5773502691896257,
embedding_multiplier_scale=78.38367176906169,
model=TransformerConfig(
emb_size=48 * 128,
widening_factor=8,
key_size=128,
num_q_heads=48,
num_kv_heads=8,
num_layers=64,
attn_output_multiplier=0.08838834764831845,
shard_activations=True,
# MoE.
num_experts=8,
num_selected_experts=2,
# Activation sharding.
data_axis="data",
model_axis="model",
),
)
inference_runner = InferenceRunner(
pad_sizes=(1024,),
runner=ModelRunner(
model=grok_1_model,
bs_per_device=0.125,
checkpoint_path=CKPT_PATH,
),
name="local",
load=CKPT_PATH,
tokenizer_path="./tokenizer.model",
local_mesh_config=(1, 8),
between_hosts_config=(1, 1),
)
inference_runner.initialize()
start_time = time.time()
gen = inference_runner.run()
print("infer time：", time.time() - start_time, "秒")
inp = "The answer to life the universe and everything is of course"
print(f"Output for prompt: {inp}", sample_from_model(gen, inp, max_len=100, temperature=0.01))
if __name__ == "__main__":
logging.basicConfig(level=logging.INFO)
main()
runners.py 0 → 100644
View file @ 049925cc
# Copyright 2024 X.AI Corp.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import bisect
import functools
import logging
import math
import re
from dataclasses import dataclass
from typing import Any, Callable, NamedTuple, Optional, Tuple
import haiku as hk
import jax
import jax.experimental.pjit as pjit
import jax.numpy as jnp
import numpy as np
import sentencepiece
from jax.experimental import mesh_utils
from jax.sharding import PartitionSpec as P
from jax.typing import ArrayLike
import checkpoint as xai_checkpoint
from model import (
LanguageModelConfig,
LanguageModelOutput,
TrainingState,
apply_rules,
Memory,
KVMemory,
)
logger = logging.getLogger(__name__)
rank_logger = logging.getLogger("rank")
TOP_K = 8
class SampleSettings(NamedTuple):
temperature: ArrayLike
nucleus_p: ArrayLike
mask: ArrayLike
# Whether a given batch element is actively used. [B]
active: ArrayLike
class SampleOutput(NamedTuple):
token_id: ArrayLike
prob: ArrayLike
top_k_token_ids: ArrayLike
top_k_probs: ArrayLike
def insert_slice(memory: Memory, slice, length, i):
slice = Memory(
layers=[
KVMemory(layer.k, layer.v, step=jnp.array([length]))
for layer in slice.layers
],
)
return jax.tree_map(lambda m, u: jax.lax.dynamic_update_index_in_dim(m, u[0], i, axis=0),
memory, slice)
def pad_to_size(x, size):
if x.shape[0] > size:
# Left truncate if the context is too long.
x = x[-size:]
return np.pad(x, [0, size - x.shape[0]], mode="constant", constant_values=0)
def top_p_filter(logits: jax.Array, top_p: jax.Array) -> jax.Array:
"""Performs nucleus filtering on logits."""
assert logits.ndim == top_p.ndim, f"Expected {logits.ndim} equal {top_p.ndim}"
sorted_logits = jax.lax.sort(logits, is_stable=False)
sorted_probs = jax.nn.softmax(sorted_logits)
threshold_idx = jnp.argmax(jnp.cumsum(sorted_probs, -1) >= 1 - top_p, axis=-1)
threshold_largest_logits = jnp.take_along_axis(
sorted_logits, threshold_idx[..., jnp.newaxis], axis=-1
)
assert threshold_largest_logits.shape == logits.shape[:-1] + (1,)
mask = logits >= threshold_largest_logits
# Set unused logits to -inf.
logits = jnp.where(mask, logits, -1e10)
return logits
def sample_token(
rngs: jax.random.PRNGKey,
lm_outputs: LanguageModelOutput,
settings: SampleSettings,
) -> SampleOutput:
# Expand the settings shape to match the logit shape.
settings = SampleSettings(
temperature=jnp.expand_dims(settings.temperature, (1, 2)), # Input [B], output [B, 1, 1].
nucleus_p=jnp.expand_dims(settings.nucleus_p, (1, 2)), # Input [B], output [B, 1, 1].
mask=jnp.expand_dims(settings.mask, 1), # Input [B, V], output [B, 1, V].
active=settings.active, # [B].
)
logits = lm_outputs.logits / settings.temperature.astype(lm_outputs.logits.dtype)
# Mask out all disallowed tokens by assigning them a near-zero probability.
logits = jnp.where(settings.mask, logits, -1e10)
# Mask out all tokens that don't fall into the p-th percentile.
logits = top_p_filter(logits, settings.nucleus_p.astype(logits.dtype))
new_token = jax.vmap(jax.random.categorical)(rngs, logits)
probabilities = jax.nn.softmax(logits)
token_prob = jnp.take_along_axis(probabilities, jnp.expand_dims(new_token, 1), axis=2)
token_prob = jnp.squeeze(token_prob, 1)
# Gather the top-k tokens and probabilities.
top_k_probs, top_k_token_ids = jax.lax.top_k(probabilities, TOP_K)
top_k_probs = jnp.squeeze(top_k_probs, 1)
top_k_token_ids = jnp.squeeze(top_k_token_ids, 1)
return SampleOutput(
new_token,
token_prob,
top_k_token_ids,
top_k_probs,
)
@dataclass
class ModelRunner:
model: LanguageModelConfig
bs_per_device: float = 2.0
load_rename_rules: Optional[list[tuple[str, str]]] = None
load_exclude_rules: Optional[list[str]] = None
rng_seed: int = 42 # Initial rng seed.
transform_forward: bool = False
checkpoint_path: str = ""
def make_forward_fn(self, mesh: Any):
def forward(tokens):
print("forward ...")
out = self.model.make(mesh=mesh)(tokens)
return out, None
if self.transform_forward:
forward = hk.transform(forward)
return forward
def initialize(
self,
init_data,
local_mesh_config: tuple[int, int],
between_hosts_config: tuple[int, int],
):
num_replicas = math.prod(between_hosts_config)
self.model.initialize()
self.model.fprop_dtype = jnp.bfloat16
num_local_gpus = len(jax.local_devices())
# Calculate the global batch size from the local batch size.
self.batch_size = int(self.bs_per_device * num_local_gpus * num_replicas)
# Calculate the batch size per host from the global batch size.
self.local_batch_size = self.batch_size // jax.process_count()
self.local_mesh_config = local_mesh_config
self.between_hosts_config = between_hosts_config
rank_logger.info(
f"Initializing mesh for {self.local_mesh_config=} {self.between_hosts_config=}..."
)
self.mesh = make_mesh(self.local_mesh_config, self.between_hosts_config)
self.forward = self.make_forward_fn(mesh=self.mesh)
self.logits_fn = hk.transform(lambda tokens: self.forward(tokens)[0])
self.eval_forward = self.make_forward_fn(mesh=self.mesh)
self.logits_eval_fn = hk.transform(lambda tokens: self.eval_forward(tokens)[0])
if self.transform_forward:
self.state_sharding = self.get_state_sharding(init_data)
rank_logger.info(f"State sharding type: {type(self.state_sharding)}")
self.init_fn = pjit.pjit(self.init, out_shardings=self.state_sharding)
def init(self, rng: jax.Array, data) -> TrainingState:
assert self.transform_forward
rng, init_rng = jax.random.split(rng)
params = self.forward.init(init_rng, data["inputs"])
return TrainingState(params=params)
def get_state_sharding(self, init_data):
assert self.transform_forward
rng = jax.random.PRNGKey(self.rng_seed)
rank_logger.info(f"partition rules: {self.model.partition_rules}")
with self.mesh:
shapes = jax.eval_shape(self.init, rng, init_data)
sharding = jax.tree_util.tree_map_with_path(
apply_rules(self.model.partition_rules()),
shapes,
)
return sharding
def load_or_init(
self,
init_data: Any,
from_checkpoint: bool = True,
init_fn: Optional[Callable] = None,
):
rng = jax.random.PRNGKey(self.rng_seed)
if not self.checkpoint_path or not from_checkpoint:
rank_logger.info("Initializing model...")
with self.mesh:
if init_fn is not None:
state = init_fn(rng, init_data)
else:
assert self.transform_forward
state = self.init_fn(rng, init_data)
rank_logger.info("Model state is newly initialized.")
else:
with self.mesh:
if init_fn:
state_shapes = jax.eval_shape(init_fn, rng, init_data)
else:
assert self.transform_forward
state_shapes = jax.eval_shape(self.init_fn, rng, init_data)
init_state = None
state = xai_checkpoint.restore(
checkpoint_path=self.checkpoint_path,
state_shapes=state_shapes,
mesh=self.mesh,
between_hosts_config=self.between_hosts_config,
state_sharding=self.state_sharding,
init_state=init_state,
params_only=True,
)
del init_state
return state
@dataclass
class Request:
prompt: str
temperature: float
nucleus_p: float
rng_seed: int
max_len: int
@dataclass
class InferenceRunner:
name: str
runner: Any
load: str
tokenizer_path: str = "/tmp/xai_data/tokenizer.model"
local_mesh_config: Tuple[int, int] = (1, 1)
between_hosts_config: Tuple[int, int] = (1, 1)
pad_sizes: tuple[int] = (1024,)
def get_pad_bucket(self, size):
i = bisect.bisect_left(self.pad_sizes, size)
return self.pad_sizes[min(i, len(self.pad_sizes) - 1)]
def initialize(self):
runner = self.runner
self.runner.transform_forward = True
dummy_data = dict(
inputs=np.zeros((1, 256), dtype=np.int32),
targets=np.zeros((1, 256), dtype=np.int32),
)
runner.initialize(
dummy_data,
local_mesh_config=self.local_mesh_config,
between_hosts_config=self.between_hosts_config,
)
self.tokenizer = sentencepiece.SentencePieceProcessor(model_file=self.tokenizer_path)
max_len = runner.model.sequence_len
self.vocab_size = self.runner.model.vocab_size
params = runner.load_or_init(dummy_data)
self.params = params
def pad_to_max_len(x):
if len(x.shape) > 1:
pad_width = max_len - x.shape[1]
return jnp.pad(x, [(0, 0), (0, pad_width), (0, 0), (0, 0)])
else:
return x
@functools.lru_cache
def lm():
return runner.model.make(mesh=runner.mesh)
def hk_forward(
tokens,
memory=None,
length=None,
active=None,
) -> LanguageModelOutput:
if memory is not None:
assert active is not None
layers = []
for l in memory.layers:
# Reset steps to 0 for inactive requests to avoid unnecessary computations.
step = jnp.where(active, l.step, jnp.zeros_like(l.step))
layers.append(l._replace(step=step))
memory = memory._replace(layers=layers)
return lm()(tokens, memory, length=length)
def hk_sample_step(rngs, last_output: SampleOutput, memory, settings):
rngs, rngs_ = jax.vmap(jax.random.split, out_axes=1)(rngs)
lm_outputs = hk_forward(last_output.token_id, memory=memory, active=settings.active)
sample_result = sample_token(rngs_, lm_outputs, settings)
return rngs, sample_result, lm_outputs.model_state
def hk_new_memory(batch_size, sequence_len):
return lm().init_memory(batch_size, sequence_len)
def hk_prefill_memory(
rngs,
memory,
settings,
last_output,
prompt,
length,
rng_seed,
new_settings,
i,
):
rng = jax.random.PRNGKey(seed=rng_seed)
rng, rng_ = jax.random.split(rng)
# Allocate new memory for this sample. The memory length is equal to the length of the
# prompt.
slice = hk_new_memory(1, prompt.shape[0])
# Move the settings for this individual batch entry into the joint settings tensor.
settings = jax.tree_map(
lambda o, v: jax.lax.dynamic_update_index_in_dim(o, v, i, axis=0),
settings,
new_settings,
)
# Get the settings for the batch entry from the joint settings tensor.
settings_slice = jax.tree_map(lambda t: jnp.expand_dims(t[i], axis=0), settings)
# Process the first n-1 tokens of the prompt.
lm_outputs = hk_forward(
jnp.expand_dims(prompt, 0),
memory=slice,
length=jnp.expand_dims(length, 0),
active=settings_slice.active,
)
# The forward pass doesn't correctly set the `step` counter inside the memory. Manually
# override it so `hk_forward` uses the correct context length in the next call.
slice = lm_outputs.model_state
slice = slice._replace(
layers=[l._replace(step=jnp.array([length])) for l in slice.layers]
)
# Sample the actual output token.
rng_ = jnp.expand_dims(rng_, 0)
new_output = sample_token(rng_, lm_outputs, settings_slice)
# Update the KV cache/memory.
slice = jax.tree_map(pad_to_max_len, slice)
memory = insert_slice(memory, slice, length, i)
rng = jnp.expand_dims(rng, 0)
rngs = jax.lax.dynamic_update_index_in_dim(rngs, rng, i, axis=0)
# Move the network outputs for this batch entry into the joint output tensor.
last_output = jax.tree_util.tree_map(
lambda last, new: jax.lax.dynamic_update_index_in_dim(last, new, i, axis=0),
last_output,
new_output,
)
return rngs, last_output, memory, settings
sample_step_ = hk.without_apply_rng(hk.transform(hk_sample_step))
prefill_memory_ = hk.without_apply_rng(hk.transform(hk_prefill_memory))
new_memory_ = hk.without_apply_rng(hk.transform(hk_new_memory))
forward_ = hk.without_apply_rng(hk.transform(hk_forward))
rng = jax.random.PRNGKey(42)
dummy_tokens = jnp.zeros((1, max_len), jnp.int32)
with runner.mesh:
shapes = jax.eval_shape(forward_.init, rng, dummy_tokens)
self.params_sharding = jax.tree_util.tree_map_with_path(
apply_rules(runner.model.partition_rules()),
shapes,
)
ds = P("data")
ms = runner.model.model.get_memory_sharding()
self.sample_step = pjit.pjit(
sample_step_.apply,
in_shardings=(self.params_sharding, None, ds, ms, None),
out_shardings=(None, ds, ms),
donate_argnums=3,
)
self.prefill_memory = pjit.pjit(
functools.partial(prefill_memory_.apply),
in_shardings=(
self.params_sharding,
None,
ms,
None,
ds,
None,
None,
None,
None,
None,
),
out_shardings=(None, ds, ms, None),
donate_argnums=(2,),
)
self.new_memory = pjit.pjit(
new_memory_.apply,
static_argnums=(1, 2),
out_shardings=ms,
)
def run(self):
"""Generator that accepts prompts."""
runner = self.runner
mesh = runner.mesh
max_len = runner.model.sequence_len
batch_size = runner.batch_size
params = self.params
rngs = jax.random.split(jax.random.PRNGKey(1), batch_size)
with mesh:
memory = self.new_memory(params, batch_size, max_len)
settings = SampleSettings(
temperature=np.zeros((batch_size,), dtype=np.float32),
nucleus_p=np.zeros((batch_size,), dtype=np.float32),
mask=np.ones((batch_size, self.vocab_size), dtype=np.int32),
active=np.zeros((batch_size), dtype=np.int32),
)
last_output = SampleOutput(
token_id=np.zeros((batch_size, 1), dtype=np.int32),
prob=np.zeros((batch_size, 1), dtype=jnp.bfloat16),
top_k_token_ids=np.zeros((batch_size, TOP_K), dtype=np.int32),
top_k_probs=np.zeros((batch_size, TOP_K), dtype=jnp.bfloat16),
)
prompt = np.array([300, 400, 500, 600, 600, 700, 800])
new_settings = SampleSettings(
temperature=np.float32(1),
nucleus_p=np.float32(1),
mask=np.ones((self.vocab_size,), dtype=np.int32),
active=np.zeros((), dtype=np.int32),
)
rng_seed = np.uint64(1)
for size in self.pad_sizes:
if size > runner.model.sequence_len:
break
logger.info("Precompile {}".format(size))
prompt_len = len(prompt)
prompt = pad_to_size(prompt, size)
rngs, last_output, memory, settings = self.prefill_memory(
params,
rngs,
memory,
settings,
last_output,
prompt,
prompt_len,
rng_seed,
new_settings,
0,
)
with runner.mesh:
logger.info("Compiling...")
rngs, last_output, memory = self.sample_step(
params, rngs, last_output, memory, settings
)
logger.info("Done compiling.")
all_tokens = []
free_slots = list(range(batch_size))
requests = [None] * batch_size
first_output = [None] * batch_size
jax.tree_map(lambda x: x.copy_to_host_async(), last_output)
prev_token = last_output
step = 0
total_num_tokens = 0
total_num_sequences = 0
with mesh:
while True:
while free_slots:
request: Optional[Request] = yield
tokens = self.tokenizer.encode(request.prompt)
temperature = request.temperature
nucleus_p = request.nucleus_p
rng_seed = request.rng_seed
i = free_slots.pop()
prompt = np.array(tokens, dtype=np.int32)
prompt_len = len(prompt)
prompt = pad_to_size(prompt, self.get_pad_bucket(prompt.shape[0]))
# All tokens are allowed.
mask = np.ones((self.vocab_size,), dtype=np.int32)
new_settings = SampleSettings(
temperature=np.float32(temperature),
nucleus_p=np.float32(nucleus_p),
mask=mask,
active=np.ones((), dtype=np.int32),
)
rng_seed = np.uint64(rng_seed)
rngs, last_output, memory, settings = self.prefill_memory(
params,
rngs,
memory,
settings,
last_output,
prompt,
prompt_len,
rng_seed,
new_settings,
i,
)
jax.tree_map(lambda x: x.copy_to_host_async(), last_output)
first_output[i] = last_output
requests[i] = request
total_num_sequences += 1
rngs, last_output, memory = self.sample_step(
params, rngs, last_output, memory, settings
)
total_num_tokens += batch_size - len(free_slots)
# prev_token should already be on the host.
prev_token = jax.tree_map(np.array, prev_token)
for i in range(batch_size):
if requests[i] is not None:
if first_output[i] is not None:
first_output_i = jax.tree_map(np.array, first_output[i])
all_tokens.append(int(first_output_i.token_id[i][0]))
first_output[i] = None
continue
all_tokens.append(int(prev_token.token_id[i][0]))
cont = len(all_tokens) < requests[i].max_len
if not cont:
output_str = self.tokenizer.decode(all_tokens)
requests[i] = None
free_slots.append(i)
all_tokens = []
settings = settings._replace(active=settings.active.at[i].set(0))
yield output_str
jax.tree_map(lambda x: x.copy_to_host_async(), last_output)
prev_token = last_output
step += 1
def make_mesh(
local_mesh_config: tuple[int, ...], between_hosts_config: tuple[int, ...]
) -> jax.sharding.Mesh:
assert len(local_mesh_config) == 2
assert len(between_hosts_config) == 2
rank_logger.info("Detected %s devices in mesh", jax.device_count())
device_mesh = mesh_utils.create_hybrid_device_mesh(
local_mesh_config,
between_hosts_config,
devices=jax.devices(),
process_is_granule=True,
)
rank_logger.debug(re.sub("\n+", "\n", f"Job device mesh is:\n{device_mesh}"))
return jax.sharding.Mesh(device_mesh, ("data", "model"))
def sample_from_model(server, prompt, max_len, temperature):
next(server)
inp = Request(
prompt=prompt,
temperature=temperature,
nucleus_p=1.0,
rng_seed=42,
max_len=max_len,
)
return server.send(inp)
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment