ByteDance-Seed / ByteDance-Seed/Bagel

ValueError: At least one of the model submodule will be offloaded to disk, please pass along an `offload_folder`.

Open
#42 3 comments 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
6.2k
Forks
545
PR merge metrics
No merged PRs in 30d

Description

# 环境
```bash
root@0b008caae888:/app/BAGEL# python -V
Python 3.11.12
root@0b008caae888:/app/BAGEL# pip -V
pip 25.1.1 from /opt/conda/lib/python3.11/site-packages/pip (python 3.11)
root@0b008caae888:/app/BAGEL#
root@0b008caae888:/app/BAGEL# nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2025 NVIDIA Corporation
Built on Wed_Jan_15_19:20:09_PST_2025
Cuda compilation tools, release 12.8, V12.8.61
Build cuda_12.8.r12.8/compiler.35404655_0
root@0b008caae888:/app/BAGEL#
root@0b008caae888:/app/BAGEL# nvidia-smi
Sun May 25 01:09:21 2025
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.216.01 Driver Version: 535.216.01 CUDA Version: 12.2 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA H20 On | 00000000:A4:00.0 Off | 0 |
| N/A 34C P0 119W / 500W | 17540MiB / 97871MiB | 0% Default |
| | | Disabled |
+-----------------------------------------+----------------------+----------------------+

+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
+---------------------------------------------------------------------------------------+
root@0b008caae888:/app/BAGEL#

root@0b008caae888:/app/BAGEL#pip list | grep -E "decord|einops|huggingface_hub|matplotlib|numpy|opencv_python|pyarrow|PyYAML|Requests|safetensors|scipy|sentencepiece|torch|torchvision|transformers|flash_attn|accelerate|wandb"
accelerate 1.7.0
decord 0.6.0
einops 0.8.1
flash_attn 2.7.4.post1
matplotlib 3.10.3
matplotlib-inline 0.1.7
numpy 2.2.6
pyarrow 20.0.0
PyYAML 6.0.2
safetensors 0.5.3
scipy 1.15.3
sentencepiece 0.2.0
torch 2.7.0+cu128
torchaudio 2.7.0+cu128
torchelastic 0.2.2
torchvision 0.22.0+cu128
transformers 4.53.0.dev0
root@0b008caae888:/app/BAGEL#

root@0b008caae888:/app/BAGEL# neofetch
.-/+oossssoo+/-. root@0b008caae888
`:+ssssssssssssssssss+:` -----------------
-+ssssssssssssssssssyyssss+- OS: Ubuntu 22.04.5 LTS x86_64
.ossssssssssssssssssdMMMNysssso. Host: CVM 3.0
/ssssssssssshdmmNNmmyNMMMMhssssss/ Kernel: 5.4.241-1-tlinux4-0017.16
+ssssssssshmydMMMMMMMNddddyssssssss+ Uptime: 3 hours, 45 mins
/sssssssshNMMMyhhyyyyhmNMMMNhssssssss/ Packages: 735 (dpkg)
.ssssssssdMMMNhsssssssssshNMMMdssssssss. Shell: bash 5.1.16
+sssshhhyNMMNyssssssssssssyNMMMysssssss+ Resolution: 1024x768
ossyNMMMNyMMhsssssssssssssshmmmhssssssso CPU: AMD EPYC 9K84 (16) @ 2.600GHz
ossyNMMMNyMMhsssssssssssssshmmmhssssssso GPU: NVIDIA 83:00.0 NVIDIA Corporation Device 2329
+sssshhhyNMMNyssssssssssssyNMMMysssssss+ GPU: NVIDIA 23:00.0 NVIDIA Corporation Device 2329
.ssssssssdMMMNhsssssssssshNMMMdssssssss. GPU: NVIDIA a4:00.0 NVIDIA Corporation Device 2329
/sssssssshNMMMyhhyyyyhdNMMMNhssssssss/ GPU: NVIDIA a3:00.0 NVIDIA Corporation Device 2329
+sssssssssdmydMMMMMMMMddddyssssssss+ GPU: NVIDIA 34:00.0 NVIDIA Corporation Device 2329
/ssssssssssshdmNNNNmyNMMMMhssssss/ GPU: NVIDIA 43:00.0 NVIDIA Corporation Device 2329
.ossssssssssssssssssdMMMNysssso. GPU: NVIDIA 63:00.0 NVIDIA Corporation Device 2329
-+sssssssssssssssssyyyssss+- GPU: NVIDIA 33:00.0 NVIDIA Corporation Device 2329
`:+ssssssssssssssssss+:` Memory: 48MiB / 65536MiB
.-/+oossssoo+/-.


```

# 错误信息如下
```python
root@0b008caae888:/workspace# ./start.sh
Directory already exists.
Already up to date.
Directory already exists.
Traceback (most recent call last):
File "/app/BAGEL/app.py", line 89, in
model = load_checkpoint_and_dispatch(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/conda/lib/python3.11/site-packages/accelerate/big_modeling.py", line 628, in load_checkpoint_and_dispatch
load_checkpoint_in_model(
File "/opt/conda/lib/python3.11/site-packages/accelerate/utils/modeling.py", line 1857, in load_checkpoint_in_model
raise ValueError(
ValueError: At least one of the model submodule will be offloaded to disk, please pass along an `offload_folder`.
root@0b008caae888:/workspace#
```

Contributor guide

No contributing guide indexed for this repository

Research direction

Start with app.py at line 89 and reproduce the failure by running ./start.sh in the reported environment. Read the checkpoint-loading call and its Accelerate traceback to determine the expected disk-offload configuration. Done means the application starts without the reported ValueError.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
machine-learning
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.