Incompatible with the latest version of fairseq?

[NightMachinery]

Here is a Colab notebook where I tried to do inference using this model.

loading archive file /content/models
loading archive file SciTLDR-Data/SciTLDR-AIC/ctrl-bin
Traceback (most recent call last):
  File "/usr/local/lib/python3.7/dist-packages/fairseq/checkpoint_utils.py", line 151, in load_checkpoint_to_cpu
    from fairseq.fb_pathmgr import fb_pathmgr
ModuleNotFoundError: No module named 'fairseq.fb_pathmgr'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "scripts/generate.py", line 99, in <module>
    generate_TLDRs(**vars(args))
  File "scripts/generate.py", line 17, in generate_TLDRs
    task='translation'
  File "/usr/local/lib/python3.7/dist-packages/fairseq/models/bart/model.py", line 104, in from_pretrained
    **kwargs,
  File "/usr/local/lib/python3.7/dist-packages/fairseq/hub_utils.py", line 68, in from_pretrained
    arg_overrides=kwargs,
  File "/usr/local/lib/python3.7/dist-packages/fairseq/checkpoint_utils.py", line 190, in load_model_ensemble_and_task
    state = load_checkpoint_to_cpu(filename, arg_overrides)
  File "/usr/local/lib/python3.7/dist-packages/fairseq/checkpoint_utils.py", line 160, in load_checkpoint_to_cpu
    path, map_location=lambda s, l: default_restore_location(s, "cpu")
  File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 529, in load
    return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
  File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 692, in _legacy_load
    magic_number = pickle_module.load(f, **pickle_load_args)
_pickle.UnpicklingError: invalid load key, '\x0a'.

[aadarshsingh191198]

I was getting the import error while running the generate.py script. I simply used the latest stable version (0.10.2) of fairseq using pip install -U fairseq and the error vanished.
PS: Earlier I had fairseq v0.9.0 installed.

[NightMachinery]

@aadarshsingh191198 commented on Dec 15, 2021, 1:58 PM GMT+3:30:

I was getting the import error while running the generate.py script. I simply used the latest stable version (0.10.2) of fairseq using pip install -U fairseq and the error vanished.
PS: Earlier I had fairseq v0.9.0 installed.

Can you share a working Colab demo? I get this error:

!cd /content/scitldr && python scripts/generate.py /content/models SciTLDR-Data/SciTLDR-AIC/ctrl /content/models --checkpoint_file bart-xsum.tldr-aic.pt --beam 5 --lenpen 0.8 --test_fname test.hypo

Traceback (most recent call last):
  File "scripts/generate.py", line 99, in <module>
    generate_TLDRs(**vars(args))
  File "scripts/generate.py", line 17, in generate_TLDRs
    task='translation'
  File "/usr/local/lib/python3.7/dist-packages/fairseq/models/bart/model.py", line 122, in from_pretrained
    **kwargs,
  File "/usr/local/lib/python3.7/dist-packages/fairseq/hub_utils.py", line 72, in from_pretrained
    arg_overrides=kwargs,
  File "/usr/local/lib/python3.7/dist-packages/fairseq/checkpoint_utils.py", line 279, in load_model_ensemble_and_task
    state = load_checkpoint_to_cpu(filename, arg_overrides)
  File "/usr/local/lib/python3.7/dist-packages/fairseq/checkpoint_utils.py", line 225, in load_checkpoint_to_cpu
    f, map_location=lambda s, l: default_restore_location(s, "cpu")
  File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 529, in load
    return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
  File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 692, in _legacy_load
    magic_number = pickle_module.load(f, **pickle_load_args)
_pickle.UnpicklingError: invalid load key, '\x0a'.

(Which I guess is the same issue as #11 )

[anasilvasd]

Here is a Colab notebook where I tried to do inference using this model.

loading archive file /content/models
loading archive file SciTLDR-Data/SciTLDR-AIC/ctrl-bin
Traceback (most recent call last):
  File "/usr/local/lib/python3.7/dist-packages/fairseq/checkpoint_utils.py", line 151, in load_checkpoint_to_cpu
    from fairseq.fb_pathmgr import fb_pathmgr
ModuleNotFoundError: No module named 'fairseq.fb_pathmgr'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "scripts/generate.py", line 99, in <module>
    generate_TLDRs(**vars(args))
  File "scripts/generate.py", line 17, in generate_TLDRs
    task='translation'
  File "/usr/local/lib/python3.7/dist-packages/fairseq/models/bart/model.py", line 104, in from_pretrained
    **kwargs,
  File "/usr/local/lib/python3.7/dist-packages/fairseq/hub_utils.py", line 68, in from_pretrained
    arg_overrides=kwargs,
  File "/usr/local/lib/python3.7/dist-packages/fairseq/checkpoint_utils.py", line 190, in load_model_ensemble_and_task
    state = load_checkpoint_to_cpu(filename, arg_overrides)
  File "/usr/local/lib/python3.7/dist-packages/fairseq/checkpoint_utils.py", line 160, in load_checkpoint_to_cpu
    path, map_location=lambda s, l: default_restore_location(s, "cpu")
  File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 529, in load
    return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
  File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 692, in _legacy_load
    magic_number = pickle_module.load(f, **pickle_load_args)
_pickle.UnpicklingError: invalid load key, '\x0a'.

The problem happens when the checkpoint is downloaded via wget. If you download it to your machine locally, you will notice that the file is much larger and comes under another name: scitldr_bart-xsum.tldr-aic.pt