Phi3Transformer does not support an attention implementation through torch.nn.functional.scaled_dot_product_attention yet.

[zxtmzxtm]

Phi3Transformer does not support an attention implementation through torch.nn.functional.scaled_dot_product_attention yet. Please request the support for this architecture: huggingface/transformers#28005. If you believe this error is a bug, please open an issue in Transformers GitHub repository and load your model with the argument attn_implementation="eager" meanwhile. Example: model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="eager")

[1038lab]

its happen because using old version or python, pytorch or CUDA, Update the custom node. it will automatically select between SDPA, eager now.

[123LiVo321]

updated whole comfy, restart, updated all custom nodes, restart, updated just the omnigen node, restart >> still the same error :(

using normal comfy (not the portable), windows10

[Jams63]

@123LiVo321 same here. updated all and getting same error " File "E:\pinokio\api\comfy.git\app\env\lib\site-packages\transformers\modeling_utils.py", line 1565, in _autoset_attn_implementation
config = cls._check_and_enable_sdpa(
File "E:\pinokio\api\comfy.git\app\env\lib\site-packages\transformers\modeling_utils.py", line 1731, in _check_and_enable_sdpa
raise ValueError(
ValueError: Phi3Transformer does not support an attention implementation through torch.nn.functional.scaled_dot_product_attention yet. Please request the support for this architecture: huggingface/transformers#28005. If you believe this error is a bug, please open an issue in Transformers GitHub repository and load your model with the argument attn_implementation="eager" meanwhile. Example: model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="eager") "

[dstults]

In case someone finds this issue, there's a new one with more info including a suggested workaround here:
#23 (comment)