-
Notifications
You must be signed in to change notification settings - Fork 52
Open
Description
Quantizing model.layers.0: 0%| | 0/28 [00:00<?, ?it/s]Traceback (most recent call last):
File "/usr/local/bin/auto-round", line 8, in <module>
sys.exit(run())
File "/home/quantize/auto-round_bak/auto_round/__main__.py", line 53, in run
tune(args)
File "/home/quantize/auto-round_bak/auto_round/script/llm.py", line 604, in tune
model, folders = autoround.quantize_and_save(export_dir, format=args.format)
File "/home/quantize/auto-round_bak/auto_round/autoround.py", line 825, in quantize_and_save
model, _ = self.quantize()
File "/home/quantize/auto-round_bak/auto_round/autoround.py", line 1574, in quantize
self._quantize_blocks(
File "/home/quantize/auto-round_bak/auto_round/autoround.py", line 2763, in _quantize_blocks
q_input, input_ids = quantize_block(
File "/home/quantize/auto-round_bak/auto_round/autoround.py", line 2606, in _quantize_block
self._scale_loss_and_backward(scaler, loss)
File "/home/quantize/auto-round_bak/auto_round/autoround.py", line 2978, in _scale_loss_and_backward
scale_loss.backward()
File "/usr/local/lib/python3.10/dist-packages/torch/_tensor.py", line 521, in backward
torch.autograd.backward(
File "/usr/local/lib/python3.10/dist-packages/torch/autograd/__init__.py", line 289, in backward
_engine_run_backward(
File "/usr/local/lib/python3.10/dist-packages/torch/autograd/graph.py", line 769, in _engine_run_backward
return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
RuntimeError: strides() called on an undefined Tensor
auto_round-0.6.1.dev0 has this problem, but auto_round-0.6.0 is normal.
Metadata
Metadata
Assignees
Labels
No labels