Missing 1st page consistently

[tranhoangnguyen03]

I consistently see this message [MISSING_PAGE_FAIL:1] at the top of every pdf I parse. And consistently, the first page is skipped somehow.

I am running the test on Google Colab using this this command !nougat "path/to/file" -o output_directory

[lukas-blecher]

Can you share an example file so I can try and reproduce it?

[smclh]

Missing title and abstract. Below is the article I used. Is the model base or some formulas not recognized MCA.pdf

[evelynmitchell]

When I run nougat on this test file, on colab (A100, high mem), I get a CUDA out of memory error.

return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]
  0% 0/2 [00:17<?, ?it/s]
Traceback (most recent call last):
  File "/usr/local/bin/nougat", line 8, in <module>
    sys.exit(main())
  File "/usr/local/lib/python3.10/dist-packages/predict.py", line 143, in main
    model_output = model.inference(image_tensors=sample, early_stopping=args.skipping)
  File "/usr/local/lib/python3.10/dist-packages/nougat/model.py", line 589, in inference
    decoder_output = self.decoder.model.generate(
  File "/usr/local/lib/python3.10/dist-packages/torch/utils/_contextlib.py", line 115, in decorate_context
    return func(*args, **kwargs)
  File "/usr/local/lib/python3.10/dist-packages/transformers/generation/utils.py", line 1602, in generate
    return self.greedy_search(
  File "/usr/local/lib/python3.10/dist-packages/transformers/generation/utils.py", line 2450, in greedy_search
    outputs = self(
  File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1501, in _call_impl
    return forward_call(*args, **kwargs)
  File "/usr/local/lib/python3.10/dist-packages/transformers/models/mbart/modeling_mbart.py", line 1850, in forward
    outputs = self.model.decoder(
  File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1501, in _call_impl
    return forward_call(*args, **kwargs)
  File "/usr/local/lib/python3.10/dist-packages/transformers/models/mbart/modeling_mbart.py", line 1108, in forward
    layer_outputs = decoder_layer(
  File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1501, in _call_impl
    return forward_call(*args, **kwargs)
  File "/usr/local/lib/python3.10/dist-packages/transformers/models/mbart/modeling_mbart.py", line 424, in forward
    hidden_states, self_attn_weights, present_key_value = self.self_attn(
  File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1501, in _call_impl
    return forward_call(*args, **kwargs)
  File "/usr/local/lib/python3.10/dist-packages/transformers/models/mbart/modeling_mbart.py", line 211, in forward
    value_states = torch.cat([past_key_value[1], value_states], dim=2)
torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 50.00 MiB (GPU 0; 39.56 GiB total capacity; 17.41 GiB already allocated; 30.56 MiB free; 37.69 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation.  See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

[maxbeaudoin]

I'm running into a similar issue. I'm not getting the [MISSING_PAGE_FAIL:1] error, but the first page is missing from the extraction.

EDIT: It works with --model 0.1.0-base

[KartavyaBagga]

@lukas-blecher @maxbeaudoin

nougat example.pdf -o output/ -m 0.1.0-base --no-skipping

Process runs with GPU load but it hangs and stuck What to do ?