#6386·ColossalAI

[BUG]: `AssertionError: div_scale should remain default` when using Gemini plugin with CPUAdam

Author: sdjasjCreated Nov 1, 2025Updated Apr 1, 2026
Labelsbug

Is there an existing issue for this bug?

  • I have searched the existing issues

The bug has not been fixed in the latest main branch

  • I have checked the latest main branch

Do you feel comfortable sharing a concise (minimal) script that reproduces the error? :)

Yes, I will share a minimal reproducible script.

Describe the bug

When using the Gemini plugin together with the CPUAdam optimizer, I encountered the following error: assert div_scale == -1, "div_scale should remain default" However, after checking the Gemini plugin source code, I found that these two should be compatible. The relevant assertion in the code is:

python
assert type(optim) in _AVAIL_OPTIM_LIST, (
    "You should use an optimizer in the available list:\n" f"{_AVAIL_OPTIM_LIST}"
)

and _AVAIL_OPTIM_LIST is defined as:

python
_AVAIL_OPTIM_LIST = {FusedAdam, CPUAdam, HybridAdam}

which clearly includes CPUAdam in the supported list.

Below is the script main.py that reproduces the issue:

python
import json
import os
import random
import sys
import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import Dataset, DataLoader
import colossalai
from colossalai.booster import Booster
from colossalai.nn.optimizer import CPUAdam, Lamb
from colossalai.booster.plugin import GeminiPlugin

class RandomDataset(Dataset):
    def __init__(self, num_samples=32 * 100, input_dim=1024, num_classes=10):
        self.x = torch.randn(num_samples, input_dim)
        self.y = torch.randint(0, num_classes, (num_samples,))

    def __len__(self):
        return len(self.x)

    def __getitem__(self, idx):
        return self.x[idx], self.y[idx]


class MLP(nn.Module):
    def __init__(self, input_dim=1024, hidden_dim=512, num_layers=10, num_classes=10):
        super().__init__()
        layers = []
        for i in range(num_layers):
            in_dim = input_dim if i == 0 else hidden_dim
            layers.append(nn.Linear(in_dim, hidden_dim))
            layers.append(nn.ReLU())
        layers.append(nn.Linear(hidden_dim, num_classes))
        self.net = nn.Sequential(*layers)

    def forward(self, x):
        return self.net(x)



def main():
    seed = 1024
    colossalai.launch_from_torch(seed=seed)
    plugin = GeminiPlugin()

    booster = Booster(plugin=plugin)

    model = MLP()
    optimizer = CPUAdam(model.parameters(), lr=1e-3)
    criterion = nn.CrossEntropyLoss()

    dataset = RandomDataset()
    train_dataloader = DataLoader(dataset, batch_size=32, shuffle=False)

    model, optimizer, criterion, train_dataloader, _ = booster.boost(model, optimizer, criterion, train_dataloader)

    precision = getattr(plugin, "precision", "fp16")
    dtype_map = {"fp16": torch.float16, "bf16": torch.bfloat16, "fp32": torch.float32}
    dtype = dtype_map.get(precision, torch.float16)

    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    model.train()
    for epoch in range(1):
        total_loss = 0
        for step, (x, y) in enumerate(train_dataloader):

            x = x.to(device=device, dtype=dtype)
            y = y.to(device=device)

            optimizer.zero_grad()
            output = model(x)
            loss = criterion(output, y)

            booster.backward(loss, optimizer)
            optimizer.step()

            total_loss += loss.item()
            print(f"[Epoch {epoch}] step {step}, loss = {loss.item():.4f}")

        avg_loss = total_loss / len(train_dataloader)
        print(f"Epoch {epoch} finished, average loss = {avg_loss:.4f}")


if __name__ == "__main__":
    main()

Running the following command:

bash
 colossalai run --nproc_per_node 4 main.py

produces the following log output:

W1112 09:36:44.760368 340847 site-packages/torch/distributed/run.py:793] 
W1112 09:36:44.760368 340847 site-packages/torch/distributed/run.py:793] *****************************************
W1112 09:36:44.760368 340847 site-packages/torch/distributed/run.py:793] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
W1112 09:36:44.760368 340847 site-packages/torch/distributed/run.py:793] *****************************************
[11/12/25 09:36:48] INFO     colossalai - colossalai - INFO:                    
                             /home/yanzhen/miniconda3/envs/colossal/lib/python3.
                             9/site-packages/colossalai/initialize.py:75 launch 
                    INFO     colossalai - colossalai - INFO: Distributed        
                             environment is initialized, world size: 4          
                    WARNING  colossalai - colossalai - WARNING:                 
                             /home/yanzhen/miniconda3/envs/colossal/lib/python3.
                             9/site-packages/colossalai/booster/plugin/gemini_pl
                             ugin.py:492 __init__                               
                    WARNING  colossalai - colossalai - WARNING:                 
                             enable_async_reduce sets pin_memory=True to achieve
                             best performance, which is not implicitly set.     
[rank0]:[W1112 09:36:49.581515762 ProcessGroupNCCL.cpp:4115] [PG ID 0 PG GUID 0 Rank 0]  using GPU 0 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect.Specify device_ids in barrier() to force use of a particular device,or call init_process_group() with a device_id.
[rank1]:[W1112 09:36:49.666183352 ProcessGroupNCCL.cpp:4115] [PG ID 0 PG GUID 0 Rank 1]  using GPU 1 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect.Specify device_ids in barrier() to force use of a particular device,or call init_process_group() with a device_id.
[rank3]:[W1112 09:36:49.739092293 ProcessGroupNCCL.cpp:4115] [PG ID 0 PG GUID 0 Rank 3]  using GPU 3 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect.Specify device_ids in barrier() to force use of a particular device,or call init_process_group() with a device_id.
[rank2]:[W1112 09:36:49.019903841 ProcessGroupNCCL.cpp:4115] [PG ID 0 PG GUID 0 Rank 2]  using GPU 2 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect.Specify device_ids in barrier() to force use of a particular device,or call init_process_group() with a device_id.
/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/utils/common.py:59: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly.  To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage()
  if data.storage().size() > 0:
/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/utils/common.py:59: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly.  To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage()
  if data.storage().size() > 0:
/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/utils/common.py:59: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly.  To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage()
  if data.storage().size() > 0:
/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/utils/common.py:59: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly.  To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage()
  if data.storage().size() > 0:
[Epoch 0] step 0, loss = 2.2969[Epoch 0] step 0, loss = 2.2969

[Epoch 0] step 0, loss = 2.2969
[Epoch 0] step 0, loss = 2.2969
[Epoch 0] step 1, loss = 2.2988
[Epoch 0] step 1, loss = 2.2988
[Epoch 0] step 1, loss = 2.2988
[Epoch 0] step 1, loss = 2.2988
[rank3]: Traceback (most recent call last):
[rank3]:   File "/home/yanzhen/distributed_test/colossalAI/test/bug1.py", line 86, in <module>
[rank3]:     main()
[rank3]:   File "/home/yanzhen/distributed_test/colossalAI/test/bug1.py", line 76, in main
[rank3]:     optimizer.step()
[rank3]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/zero/gemini/gemini_optimizer.py", line 288, in step
[rank3]:     ret = self.optim.step(div_scale=combined_scale, *args, **kwargs)
[rank3]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/optim/optimizer.py", line 487, in wrapper
[rank3]:     out = func(*args, **kwargs)
[rank3]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
[rank3]:     return func(*args, **kwargs)
[rank3]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/nn/optimizer/cpu_adam.py", line 193, in step
[rank3]:     assert div_scale == -1, "div_scale should remain default"
[rank3]: AssertionError: div_scale should remain default
[rank2]: Traceback (most recent call last):
[rank2]:   File "/home/yanzhen/distributed_test/colossalAI/test/bug1.py", line 86, in <module>
[rank2]:     main()
[rank2]:   File "/home/yanzhen/distributed_test/colossalAI/test/bug1.py", line 76, in main
[rank2]:     optimizer.step()
[rank2]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/zero/gemini/gemini_optimizer.py", line 288, in step
[rank2]:     ret = self.optim.step(div_scale=combined_scale, *args, **kwargs)
[rank2]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/optim/optimizer.py", line 487, in wrapper
[rank2]:     out = func(*args, **kwargs)
[rank2]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
[rank2]:     return func(*args, **kwargs)
[rank2]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/nn/optimizer/cpu_adam.py", line 193, in step
[rank2]:     assert div_scale == -1, "div_scale should remain default"
[rank2]: AssertionError: div_scale should remain default
[rank1]: Traceback (most recent call last):
[rank1]:   File "/home/yanzhen/distributed_test/colossalAI/test/bug1.py", line 86, in <module>
[rank1]:     main()
[rank1]:   File "/home/yanzhen/distributed_test/colossalAI/test/bug1.py", line 76, in main
[rank1]:     optimizer.step()
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/zero/gemini/gemini_optimizer.py", line 288, in step
[rank1]:     ret = self.optim.step(div_scale=combined_scale, *args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/optim/optimizer.py", line 487, in wrapper
[rank1]:     out = func(*args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
[rank1]:     return func(*args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/nn/optimizer/cpu_adam.py", line 193, in step
[rank1]:     assert div_scale == -1, "div_scale should remain default"
[rank1]: AssertionError: div_scale should remain default
[rank0]: Traceback (most recent call last):
[rank0]:   File "/home/yanzhen/distributed_test/colossalAI/test/bug1.py", line 86, in <module>
[rank0]:     main()
[rank0]:   File "/home/yanzhen/distributed_test/colossalAI/test/bug1.py", line 76, in main
[rank0]:     optimizer.step()
[rank0]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/zero/gemini/gemini_optimizer.py", line 288, in step
[rank0]:     ret = self.optim.step(div_scale=combined_scale, *args, **kwargs)
[rank0]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/optim/optimizer.py", line 487, in wrapper
[rank0]:     out = func(*args, **kwargs)
[rank0]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
[rank0]:     return func(*args, **kwargs)
[rank0]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/nn/optimizer/cpu_adam.py", line 193, in step
[rank0]:     assert div_scale == -1, "div_scale should remain default"
[rank0]: AssertionError: div_scale should remain default
[rank0]:[W1112 09:36:50.405163289 ProcessGroupNCCL.cpp:1250] Warning: WARNING: process group has NOT been destroyed before we destruct ProcessGroupNCCL. On normal program exit, the application should call destroy_process_group to ensure that any pending NCCL operations have finished in this process. In rare cases this process can exit before this point and block the progress of another member of the process group. This constraint has always been present,  but this warning has only been added since PyTorch 2.4 (function operator())
W1112 09:36:51.754890 340847 site-packages/torch/distributed/elastic/multiprocessing/api.py:897] Sending process 340918 closing signal SIGTERM
W1112 09:36:51.758123 340847 site-packages/torch/distributed/elastic/multiprocessing/api.py:897] Sending process 340919 closing signal SIGTERM
E1112 09:36:51.922632 340847 site-packages/torch/distributed/elastic/multiprocessing/api.py:869] failed (exitcode: 1) local_rank: 2 (pid: 340920) of binary: /home/yanzhen/miniconda3/envs/colossal/bin/python3.9
Traceback (most recent call last):
  File "/home/yanzhen/miniconda3/envs/colossal/bin/torchrun", line 7, in <module>
    sys.exit(main())
  File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
    return f(*args, **kwargs)
  File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/distributed/run.py", line 919, in main
    run(args)
  File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/distributed/run.py", line 910, in run
    elastic_launch(
  File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/distributed/launcher/api.py", line 138, in __call__
    return launch_agent(self._config, self._entrypoint, list(args))
  File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/distributed/launcher/api.py", line 269, in launch_agent
    raise ChildFailedError(
torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
============================================================
bug1.py FAILED
------------------------------------------------------------
Failures:
[1]:
  time      : 2025-11-12_09:36:51
  host      : ubuntu
  rank      : 3 (local_rank: 3)
  exitcode  : 1 (pid: 340921)
  error_file: <N/A>
  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
------------------------------------------------------------
Root Cause (first observed failure):
[0]:
  time      : 2025-11-12_09:36:51
  host      : ubuntu
  rank      : 2 (local_rank: 2)
  exitcode  : 1 (pid: 340920)
  error_file: <N/A>
  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
============================================================
Error: failed to run torchrun --nproc_per_node=4 --nnodes=1 --node_rank=0 --master_addr=127.0.0.1 --master_port=29505 bug1.py on 127.0.0.1, is localhost: True, exception: Encountered a bad command exit code!

Command: 'cd /home/yanzhen/distributed_test/colossalAI/test && export SHELL="/bin/bash" COLORTERM="truecolor" VSCODE_DEBUGPY_ADAPTER_ENDPOINTS="/home/yanzhen/.vscode-server/extensions/ms-python.debugpy-2025.14.1/.noConfigDebugAdapterEndpoints/endpoint-8ca95acfe78cb59c.txt" TERM_PROGRAM_VERSION="1.105.1" CONDA_EXE="/home/yanzhen/miniconda3/bin/conda" NCCL_P2P_DISABLE="1" LC_ADDRESS="zh_CN.UTF-8" LC_NAME="zh_CN.UTF-8" PYDEVD_DISABLE_FILE_VALIDATION="1" LC_MONETARY="zh_CN.UTF-8" PWD="/home/yanzhen/distributed_test/colossalAI/test" LOGNAME="yanzhen" XDG_SESSION_TYPE="tty" CONDA_PREFIX="/home/yanzhen/miniconda3/envs/colossal" BUNDLED_DEBUGPY_PATH="/home/yanzhen/.vscode-server/extensions/ms-python.debugpy-2025.14.1/bundled/libs/debugpy" VSCODE_GIT_ASKPASS_NODE="/home/yanzhen/.vscode-server/cli/servers/Stable-7d842fb85a0275a4a8e4d7e040d2625abbf7f084/server/node" MOTD_SHOWN="pam" HOME="/home/yanzhen" LC_PAPER="zh_CN.UTF-8" LANG="en_US.UTF-8" LS_COLORS="rs=0:di=01;34:ln=01;36:mh=00:pi=40;33:so=01;35:do=01;35:bd=40;33;01:cd=40;33;01:or=40;31;01:mi=00:su=37;41:sg=30;43:ca=30;41:tw=30;42:ow=34;42:st=37;44:ex=01;32:*.tar=01;31:*.tgz=01;31:*.arc=01;31:*.arj=01;31:*.taz=01;31:*.lha=01;31:*.lz4=01;31:*.lzh=01;31:*.lzma=01;31:*.tlz=01;31:*.txz=01;31:*.tzo=01;31:*.t7z=01;31:*.zip=01;31:*.z=01;31:*.dz=01;31:*.gz=01;31:*.lrz=01;31:*.lz=01;31:*.lzo=01;31:*.xz=01;31:*.zst=01;31:*.tzst=01;31:*.bz2=01;31:*.bz=01;31:*.tbz=01;31:*.tbz2=01;31:*.tz=01;31:*.deb=01;31:*.rpm=01;31:*.jar=01;31:*.war=01;31:*.ear=01;31:*.sar=01;31:*.rar=01;31:*.alz=01;31:*.ace=01;31:*.zoo=01;31:*.cpio=01;31:*.7z=01;31:*.rz=01;31:*.cab=01;31:*.wim=01;31:*.swm=01;31:*.dwm=01;31:*.esd=01;31:*.jpg=01;35:*.jpeg=01;35:*.mjpg=01;35:*.mjpeg=01;35:*.gif=01;35:*.bmp=01;35:*.pbm=01;35:*.pgm=01;35:*.ppm=01;35:*.tga=01;35:*.xbm=01;35:*.xpm=01;35:*.tif=01;35:*.tiff=01;35:*.png=01;35:*.svg=01;35:*.svgz=01;35:*.mng=01;35:*.pcx=01;35:*.mov=01;35:*.mpg=01;35:*.mpeg=01;35:*.m2v=01;35:*.mkv=01;35:*.webm=01;35:*.webp=01;35:*.ogm=01;35:*.mp4=01;35:*.m4v=01;35:*.mp4v=01;35:*.vob=01;35:*.qt=01;35:*.nuv=01;35:*.wmv=01;35:*.asf=01;35:*.rm=01;35:*.rmvb=01;35:*.flc=01;35:*.avi=01;35:*.fli=01;35:*.flv=01;35:*.gl=01;35:*.dl=01;35:*.xcf=01;35:*.xwd=01;35:*.yuv=01;35:*.cgm=01;35:*.emf=01;35:*.ogv=01;35:*.ogx=01;35:*.aac=00;36:*.au=00;36:*.flac=00;36:*.m4a=00;36:*.mid=00;36:*.midi=00;36:*.mka=00;36:*.mp3=00;36:*.mpc=00;36:*.ogg=00;36:*.ra=00;36:*.wav=00;36:*.oga=00;36:*.opus=00;36:*.spx=00;36:*.xspf=00;36:" PYTHONSTARTUP="/home/yanzhen/.vscode-server/data/User/workspaceStorage/0d3e22743b5008777912953212595ae2/ms-python.python/pythonrc.py" SSL_CERT_DIR="/usr/lib/ssl/certs" CONDA_PROMPT_MODIFIER="(colossal) " GIT_ASKPASS="/home/yanzhen/.vscode-server/cli/servers/Stable-7d842fb85a0275a4a8e4d7e040d2625abbf7f084/server/extensions/git/dist/askpass.sh" SSH_CONNECTION="192.168.1.29 32826 192.168.102.133 18022" USE_MODELSCOPE_HUB="1" VSCODE_PYTHON_AUTOACTIVATE_GUARD="1" _CONDA_EXE="/home/yanzhen/miniconda3/bin/conda" LESSCLOSE="/usr/bin/lesspipe %s %s" _CONDA_ROOT="/home/yanzhen/miniconda3" XDG_SESSION_CLASS="user" TERM="xterm-256color" LC_IDENTIFICATION="zh_CN.UTF-8" PYTHON_BASIC_REPL="1" LESSOPEN="| /usr/bin/lesspipe %s" USER="yanzhen" VSCODE_GIT_IPC_HANDLE="/run/user/1006/vscode-git-760712a092.sock" CONDA_SHLVL="2" SHLVL="1" LC_TELEPHONE="zh_CN.UTF-8" LC_MEASUREMENT="zh_CN.UTF-8" XDG_SESSION_ID="6892" CONDA_PYTHON_EXE="/home/yanzhen/miniconda3/bin/python" LD_LIBRARY_PATH="/home/yanzhen/.tensornvme/lib:/usr/local/cuda-12.4/lib64:/home/yanzhen/.tensornvme/lib:/usr/local/cuda-12.4/lib64:" XDG_RUNTIME_DIR="/run/user/1006" SSL_CERT_FILE="/usr/lib/ssl/cert.pem" SSH_CLIENT="192.168.1.29 32826 18022" CONDA_DEFAULT_ENV="colossal" DEBUGINFOD_URLS="https://debug