[BUG]: ModuleNotFoundError: No module named 'colossalai._analyzer'

[a1xiaoyaozi]

🐛 Describe the bug

command

git clone https://github.com/hpcaitech/ColossalAI.git
cd ColossalAI
# install colossalai
pip install .
# install chat
cd applications/Chat
pip install .
git clone https://github.com/hpcaitech/transformers
cd transformers
pip install .
cd ../examples
# train
torchrun --standalone --nproc_per_node=4 train_sft.py \
    --pretrain "./LLaMa-7B" \
    --model 'llama' \
    --strategy colossalai_zero2 \
    --log_interval 10 \
    --save_path  ./Coati-7B \
    --dataset ./dataset/instinwild_en.json \
    --batch_size 4 \
    --accimulation_steps 8 \
    --lr 2e-5 \
    --max_datasets_size 512 \
    --max_epochs 1

result

Traceback (most recent call last): File "/root/ColossalAI/applications/Chat/examples/train_sft.py", line 7, in from coati.dataset import DataCollatorForSupervisedDataset, SFTDataset, SupervisedDataset File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/coati/dataset/init.py", line 1, in from .prompt_dataset import PromptDataset File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/coati/dataset/prompt_dataset.py", line 12, in from colossalai.logging import get_dist_logger File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/colossalai/init.py", line 1, in from .initialize import ( File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/colossalai/initialize.py", line 18, in from colossalai.amp import AMP_TYPE, convert_to_amp File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/colossalai/amp/init.py", line 11, in from .apex_amp import convert_to_apex_amp File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/colossalai/amp/apex_amp/init.py", line 4, in from .apex_amp import ApexAMPOptimizer File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/colossalai/amp/apex_amp/apex_amp.py", line 13, in from colossalai.nn.optimizer import ColossalaiOptimizer File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/colossalai/nn/init.py", line 1, in from ._ops import * File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/colossalai/nn/_ops/init.py", line 1, in from .addmm import colo_addmm File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/colossalai/nn/_ops/addmm.py", line 6, in from ._utils import GeneralTensor, Number, convert_to_colo_tensor, reduce_grad, reduce_input File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/colossalai/nn/_ops/_utils.py", line 7, in from colossalai.nn.layer.utils import divide File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/colossalai/nn/layer/init.py", line 7, in from .moe import * File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/colossalai/nn/layer/moe/init.py", line 1, in from .checkpoint import load_moe_model, save_moe_model File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/colossalai/nn/layer/moe/checkpoint.py", line 5, in from .experts import MoeExperts File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/colossalai/nn/layer/moe/experts.py", line 12, in from colossalai.zero.legacy.init_ctx import no_shard_zero_decrator File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/colossalai/zero/init.py", line 1, in from .gemini import ( File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/colossalai/zero/gemini/init.py", line 3, in from .gemini_ddp import GeminiDDP, ZeroDDP File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/colossalai/zero/gemini/gemini_ddp.py", line 17, in from colossalai.utils.model.experimental import LazyTensor File "/root/miniconda3/envs/coati/lib/python3.9/site-packages/colossalai/utils/model/experimental.py", line 10, in from colossalai._analyzer._subclasses import MetaTensor ModuleNotFoundError: No module named 'colossalai._analyzer'

Environment

CUDA 11.0 cuDNN 8 Python 3.9 Pytorch 1.13.1+cu117

[allaccs]

Worked yesterday for me, stopped working today.

Workaround : git checkout e6a132a449d10489e65544e77f49190e2817b587 Use the commit from yesterday and it works, I guess an update broke the code.

[tuzeao]

Meeted the same problem too. You need to checkout the old version which is around yesterday morning or before.

[FrankLeeeee]

Yes, the source code is not stable until an official release.

[RestartNeverLate]

我通过复制colossalai/_analyzer 到site-packages/colossalai/ 解决了这个问题 cp -r _analyzer/ /path/to/site-packages/colossalai/

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

I solved this problem by copying colossalai/_analyzer to site-packages/colossalai/ cp -r _analyzer/ /path/to/site-packages/colossalai/

[FrankLeeeee]

I will try to fix this bug today.

[tianbuwei]

same question

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

same question

[jamesunnc]

Please does anyone get any specific advice on how to solve this import error?

[wangxiaobo007]

Please does anyone get any specific advice on how to solve this import error?

我通过复制colossalai/_analyzer 到site-packages/colossalai/ 解决了这个问题 cp -r _analyzer/ /path/to/site-packages/colossalai/

I refer to the above solution and it can now run normally, but I always feel that this method is not very friendly

[jamesunnc]

Many thanks, it works now.

[LDH007]

我通过复制colossalai/_analyzer 到site-packages/colossalai/ 解决了这个问题 cp -r _analyzer/ /path/to/site-packages/colossalai/

awesome!

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

I solved this problem by copying colossalai/_analyzer to site-packages/colossalai/ cp -r _analyzer/ /path/to/site-packages/colossalai/

awesome!

[catqaq]

any progress?