[Model Support] Qwen3-Next - #1068
Conversation
Signed-off-by: yaoyu-33 <yaoyu.094@gmail.com>
|
torchrun --nproc_per_node=8 examples/conversion/hf_to_megatron_generate_text.py --hf_model_path="Qwen/Qwen3-Next-80B-A3B-Instruct" --prompt="Hello, how are you?" --ep 8 Output non-sense right now |
# Conflicts: # tests/functional_tests/models/test_qwen3_next_conversion.py
Signed-off-by: yaoyu-33 <yaoyu.094@gmail.com>
Signed-off-by: yaoyu-33 <yaoyu.094@gmail.com>
|
/ok to test 0362f36 |
Signed-off-by: yaoyu-33 <yaoyu.094@gmail.com>
|
/ok to test 147ed73 |
|
/ok to test 054120f |
|
/ok to test cbfb04b |
Signed-off-by: yaoyu-33 <yaoyu.094@gmail.com>
|
/ok to test 1e75af2 |
ko3n1g
left a comment
There was a problem hiding this comment.
One discussion regarding dependencies is required from my side
| "Topic :: Utilities", | ||
| ] | ||
| dependencies = [ | ||
| "transformers==4.57.1", |
There was a problem hiding this comment.
Could we just a looser pin than this? If we know that the lower boundary is 4.57.1, I would pledge for >=. Even better would be to have no pin, and to gracefully raise exception on the affected codepaths when transformers is <4.57.1. That way we would be more flexible
ko3n1g
left a comment
There was a problem hiding this comment.
Discussed offline that a follow up PR will follow
Signed-off-by: yaoyu-33 <yaoyu.094@gmail.com> Co-authored-by: Yuzhong Wang <yuzhongw@nvidia.com> Signed-off-by: NeMo Bot <nemo-bot@nvidia.com>
Signed-off-by: yaoyu-33 <yaoyu.094@gmail.com> Co-authored-by: Yuzhong Wang <yuzhongw@nvidia.com> Signed-off-by: Paul Gibbons <pgibbons@nvidia.com>
Signed-off-by: yaoyu-33 <yaoyu.094@gmail.com> Co-authored-by: Yuzhong Wang <yuzhongw@nvidia.com> Signed-off-by: mollys <mollys@mollys.nvidia.com>
part of #786