Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Feature] XTuner Lite #974

Merged
merged 104 commits into from
Dec 27, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
104 commits
Select commit Hold shift + click to select a range
f6e6a43
minimum dependency sft
pppppM Jun 12, 2024
3b6237e
fix dispatch
pppppM Jun 13, 2024
3404262
add timer
pppppM Jun 13, 2024
fb96828
add tgs
pppppM Jun 13, 2024
cf4176a
internlm2 tp
pppppM Jun 13, 2024
ce8a5c1
rms support tp
pppppM Jun 14, 2024
ebaa2e1
gradient checkpointing
pppppM Jun 18, 2024
69b7f29
lazy load pretrain
pppppM Jun 19, 2024
6f87c28
temp
pppppM Jul 11, 2024
37afd2b
fix bugs
pppppM Jul 15, 2024
3789fb8
add data pipeline example
pppppM Jul 15, 2024
bbcfc03
fix lints
pppppM Jul 15, 2024
bdbd7b1
remove useless code
pppppM Jul 15, 2024
e31822f
fix hard pack bug
pppppM Jul 16, 2024
9f40981
add comments
pppppM Jul 16, 2024
90eeaab
clean code
pppppM Jul 16, 2024
2b90046
add shard strategy
pppppM Jul 16, 2024
2620215
support cpu offload
pppppM Jul 16, 2024
1893d9c
support cpu offload
pppppM Jul 16, 2024
e4f295b
trust remote code
pppppM Jul 16, 2024
1026668
fix soft packer bug
pppppM Jul 16, 2024
cc25ee9
fix soft packer bug
pppppM Jul 17, 2024
4f38fd3
fix soft packer bug
pppppM Jul 17, 2024
399cd87
refactor data pipeline
pppppM Jul 17, 2024
6cb8e5a
fixup
pppppM Jul 17, 2024
123633a
fix pad tokens bug
pppppM Jul 17, 2024
4980af5
check input_ids and labels
pppppM Jul 18, 2024
1d02ff9
check input_ids and labels in collator
pppppM Jul 18, 2024
871f210
fix load local datasets bug
pppppM Jul 18, 2024
c8a2225
fix load cache datasts
pppppM Jul 18, 2024
dbe954b
restore dset order
pppppM Jul 18, 2024
cae140a
save cached infos
pppppM Jul 19, 2024
c6c9991
accelerate start up
pppppM Jul 22, 2024
2020150
avoid all gather cached datasets
pppppM Jul 22, 2024
5566469
fixup
pppppM Jul 22, 2024
c0a4316
fix cache bug
pppppM Jul 22, 2024
629fe63
Support group length (#4)
HIT-cwh Jul 22, 2024
98c193e
support llava
pppppM Jul 22, 2024
482518e
support packer cache
pppppM Jul 23, 2024
03431b9
refactor dist load
pppppM Jul 24, 2024
12df56c
Add sp tp (#5)
HIT-cwh Jul 26, 2024
2bbe3c2
adapt new parallel load
pppppM Jul 26, 2024
03353ba
Merge branch 'lite' of github.com:pppppM/xtuner into lite
pppppM Jul 26, 2024
ab4ed14
fix setup_parallel (#7)
HIT-cwh Jul 26, 2024
2c587ad
Merge branch 'lite' of github.com:pppppM/xtuner into lite
pppppM Jul 29, 2024
45b6da9
fix some bugs
pppppM Jul 29, 2024
502061b
add remote codes
pppppM Jul 29, 2024
d1845e1
add convert script
pppppM Jul 29, 2024
ca8b8b0
support load image from ceph
pppppM Jul 29, 2024
380054b
support load image from ceph
pppppM Jul 29, 2024
1ab2a9a
fix cache dataset bugs
pppppM Jul 29, 2024
fbc344b
support mulit images
pppppM Jul 30, 2024
17a98cf
support llava interleave
pppppM Jul 31, 2024
be383c9
fix load timeout
pppppM Aug 5, 2024
42a7bf6
refactor datasets: optimize the cache mechanism and clean up code
pppppM Oct 9, 2024
4d14462
distinguish dataset components based on algorithms
pppppM Oct 9, 2024
7574b72
support fsdp2+3d parallel
pppppM Oct 9, 2024
98bbc17
fix lints
pppppM Oct 9, 2024
5f41ac4
support contiguous batching
pppppM Oct 9, 2024
6a8e296
refactor parallel
pppppM Oct 9, 2024
7272bed
zero wasting ppo
pppppM Oct 9, 2024
c4a6121
support asend npu
pppppM Oct 10, 2024
99477de
fix openai convert
pppppM Oct 10, 2024
1ee2e82
fix npu bugs
pppppM Oct 10, 2024
5507b7c
fix npu bug
pppppM Oct 10, 2024
058d3bc
dispatch npu flash attn
pppppM Oct 10, 2024
aef09c7
adapt asend npu
pppppM Oct 10, 2024
13c1151
fix ppo losses
pppppM Oct 12, 2024
5da870e
steady increase in reward
pppppM Oct 13, 2024
f69e80a
faster ppo
pppppM Oct 15, 2024
ab88224
fix top-p generate
pppppM Oct 23, 2024
5523b3e
support internlm3
pppppM Oct 23, 2024
4c8605c
baseline 2.5
pppppM Oct 23, 2024
1b0340c
Merge branch 'main' into ppo+cache
pppppM Oct 23, 2024
f7e9cbe
fix internlm3
pppppM Oct 24, 2024
8dfd02f
(ing)support hard pack
pppppM Nov 4, 2024
0624229
support qwen2
pppppM Nov 4, 2024
6d82eb8
fix dataset bugs
pppppM Nov 4, 2024
918b9b9
baseline
pppppM Nov 4, 2024
a905683
del ppo.py
pppppM Nov 4, 2024
877fd09
fixup
pppppM Nov 4, 2024
75ab733
support hybrid sp
pppppM Nov 8, 2024
5b4126c
fix hybrid sp
pppppM Nov 8, 2024
f6bf99d
qwen2 + hybird sp
pppppM Nov 12, 2024
569c725
fix requirements
pppppM Nov 12, 2024
32c7a63
avoid re-initialize dist
pppppM Nov 13, 2024
fd8dab5
support group pack
pppppM Nov 13, 2024
dada500
pretrain (#13)
HIT-cwh Nov 14, 2024
c6a9ad1
Moe pretrain (#14)
HIT-cwh Nov 14, 2024
32421cf
support prompt length control (#15)
ShuaibinLi Nov 19, 2024
da2c8a8
support VLM Base (#16)
hhaAndroid Nov 22, 2024
cd0c3f1
accelerate tokenize
pppppM Nov 26, 2024
728ff55
Updata InternVL (#17)
hhaAndroid Nov 27, 2024
fcd0836
fix rand sampler (#18)
hhaAndroid Nov 27, 2024
a12f0ed
llama support transformers >= 4.45 (#19)
HIT-cwh Nov 29, 2024
99344c1
convert fsdp1 to fsdp2 in sft.py
pppppM Dec 6, 2024
6caad42
[Feature] Support Liger Kernel (#20)
pppppM Dec 9, 2024
6c6dfad
filter data by max length (#21)
pppppM Dec 9, 2024
07c175f
fix causal forward, prefetch, and remote code (#22)
pppppM Dec 10, 2024
5e85e19
[Enhancement] Accelerating Data Pipeline (#23)
pppppM Dec 11, 2024
657d74f
[Enhancement] Ensure data integrity when the sampling ratio is more t…
pppppM Dec 11, 2024
fad0a4d
[Fix] Pass in temperature during generation (#25)
pppppM Dec 12, 2024
9192ec8
Support Janus and fix some error (#27)
hhaAndroid Dec 16, 2024
a02bbb2
remove useless code
pppppM Dec 26, 2024
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion .pre-commit-config-zh-cn.yaml
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
exclude: ^tests/data/|^xtuner/tools/model_converters/modeling_internlm2_reward/
exclude: ^tests/data/|^xtuner/tools/model_converters/modeling_internlm2_reward/|^xtuner/_lite/modelings/|^xtuner/_lite/accelerate/dispatches/huggingface/
repos:
- repo: https://gitee.com/openmmlab/mirrors-flake8
rev: 5.0.4
Expand Down
2 changes: 1 addition & 1 deletion .pre-commit-config.yaml
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
exclude: ^tests/data/|^xtuner/tools/model_converters/modeling_internlm2_reward/
exclude: ^tests/data/|^xtuner/tools/model_converters/modeling_internlm2_reward/|^xtuner/_lite/modelings/|^xtuner/_lite/accelerate/dispatches/huggingface/
repos:
- repo: https://github.com/PyCQA/flake8
rev: 5.0.4
Expand Down
1 change: 1 addition & 0 deletions requirements/lmdeploy.txt
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
lmdeploy>=0.6.2 --no-deps
15 changes: 6 additions & 9 deletions requirements/runtime.txt
Original file line number Diff line number Diff line change
@@ -1,16 +1,11 @@
# Minimum 0.40.0.post4 to fix some 4-bit precision bugs
bitsandbytes>=0.40.0.post4
# Minimum 2.16.0 to fix some bugs, see https://github.com/huggingface/datasets/pull/6444
datasets>=2.16.0
einops
# Minimum 0.1.2 to fix some bugs, see https://github.com/InternLM/lagent/pull/44
lagent>=0.1.2
einop
# Avoid `import cv2` failed
opencv-python==4.7.0.72
# Minimum 0.10.3 to support distributed evaluation for MMBench
# see https://github.com/open-mmlab/mmengine/pull/1469
mmengine>=0.10.3
openpyxl
# Minimum 0.4.0 to support QLoRA, see https://github.com/huggingface/peft/pull/476
peft>=0.4.0
scikit-image
scipy
SentencePiece
Expand All @@ -23,5 +18,7 @@ torchvision
# https://github.com/huggingface/transformers/blob/v4.38.0/src/transformers/models/llama/modeling_llama.py#L921-L923
# transformers >= 4.43.0 use _flash_attention_forward but not self._flash_attention_forward
# to calculate attn output which lead to bc braeking
transformers>=4.36.0,!=4.38.0,!=4.38.1,!=4.38.2,<=4.42.4
transformers>=4.45
transformers_stream_generator
loguru
pydantic
4 changes: 3 additions & 1 deletion setup.py
Original file line number Diff line number Diff line change
Expand Up @@ -117,10 +117,12 @@ def gen_packages_items():
'Programming Language :: Python :: 3.8',
'Programming Language :: Python :: 3.9',
'Programming Language :: Python :: 3.10',
'Programming Language :: Python :: 3.11',
'Programming Language :: Python :: 3.12',
'Topic :: Utilities',
],
# Python maximum version <3.11, to support mpi4py-mpich
python_requires='>=3.8, <3.11',
python_requires='>=3.8, <3.13',
license='Apache License 2.0',
install_requires=parse_requirements('requirements/runtime.txt'),
extras_require={
Expand Down
Loading
Loading