NVIDIA-NeMo / NVIDIA-NeMo/RL

Sequence parallelism is not supported for NemotronHForCausalLM

Open
#1,525 3 comments 0 reactions 1 assignee Claimed by @joyang-nv View on GitHub
community-request enhancement external t-pytdensor waiting-on-customer x-sarvam
Dominant language
Python
Stars
2k
Forks
561
Avg merge
4d 5h
Merged PRs (30d)
145

Description

Is there a way use sequence parallelism for NemotronHForCausalLM model?
I am getting:
```
File "/data/TTS/aayush/nemo-rl/nemo_rl/models/policy/dtensor_policy_worker.py", line 352, in __init__
self.model = _parallelize_model(
^^^^^^^^^^^^^^^^^^^
File "/data/TTS/aayush/nemo-rl/nemo_rl/models/dtensor/parallelize.py", line 529, in _parallelize_model
return _parallelize_nm5_h(
^^^^^^^^^^^^^^^^^^^
File "/data/TTS/aayush/nemo-rl/nemo_rl/models/dtensor/parallelize.py", line 415, in _parallelize_nm5_h
assert not sequence_parallel, (
^^^^^^^^^^^^^^^^^^^^^
AssertionError: Sequence parallelism is not supported for NemotronHForCausalLM
```

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.