🚀 The feature, motivation and pitch
Liger's functional doesn't support keyword arguments because it's implemented by replacing torchautograd.Function.apply.
kwargs support is necessary for LigerCrossEntropyLoss to patch transformers model after GA fix
https://github.com/huggingface/transformers/blob/a06a0d12636756352494b99b5b264ac9955bc735/src/transformers/models/llama/modeling_llama.py#L1224
Alternatives
No response
Additional context
No response
🚀 The feature, motivation and pitch
Liger's functional doesn't support keyword arguments because it's implemented by replacing
torchautograd.Function.apply.kwargs support is necessary for LigerCrossEntropyLoss to patch transformers model after GA fix
https://github.com/huggingface/transformers/blob/a06a0d12636756352494b99b5b264ac9955bc735/src/transformers/models/llama/modeling_llama.py#L1224
Alternatives
No response
Additional context
No response