openai

/

whisper-large-v3-turbo

like 1.65k

Follow

OpenAI 3.64k

Hi, trying to use your torch compile example as it is with
accelerate==0.34.2
torch== 2.4.1
transformers==4.45.1

im getting torch compile error related to dynamo backend:

   raise exc.UserError(
torch._dynamo.exc.UserError: Dynamic control flow is not supported at the moment. Please use functorch.experimental.control_flow.cond to explicitly capture the control flow. For more information about this error, see: https://pytorch.org/docs/main/generated/exportdb/index.html#cond-operands

from user code:
  File "/home/jupyter/miniconda3/envs/whisper_turbo/lib/python3.11/site-packages/transformers/models/whisper/modeling_whisper.py", line 1765, in forward
   outputs = self.model(
 File "/home/jupyter/miniconda3/envs/whisper_turbo/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1562, in _call_impl
   return forward_call(*args, **kwargs)
 File "/home/jupyter/miniconda3/envs/whisper_turbo/lib/python3.11/site-packages/transformers/models/whisper/modeling_whisper.py", line 1616, in forward
   encoder_outputs = self.encoder(
 File "/home/jupyter/miniconda3/envs/whisper_turbo/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1562, in _call_impl
   return forward_call(*args, **kwargs)
 File "/home/jupyter/miniconda3/envs/whisper_turbo/lib/python3.11/site-packages/transformers/models/whisper/modeling_whisper.py", line 1121, in forward
   layer_outputs = encoder_layer(
 File "/home/jupyter/miniconda3/envs/whisper_turbo/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1562, in _call_impl
   return forward_call(*args, **kwargs)
 File "/home/jupyter/miniconda3/envs/whisper_turbo/lib/python3.11/site-packages/transformers/models/whisper/modeling_whisper.py", line 689, in forward
   if hidden_states.dtype == torch.float16 and (

Set TORCH_LOGS="+dynamo" and TORCHDYNAMO_VERBOSE=1 for more information


You can suppress this exception and fall back to eager by setting:
   import torch._dynamo
   torch._dynamo.config.suppress_errors = True

adding

torch._dynamo.config.suppress_errors = True

is not helping
So, i'm wondering if it needs more manipulations than provided in your example to run it with torch compile?