Issue of install FastCoref in Colab

[xiongjun926g]

Its been working fantastic till last week in colab.

!pip install fastcoref==2.1.5 -q

!pip install fastcoref==1.6.0 -q

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 474.3/474.3 kB 8.1 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 9.5/9.5 MB 23.7 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 116.3/116.3 kB 5.4 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 207.3/207.3 kB 11.2 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 39.9/39.9 MB 12.9 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 311.4/311.4 kB 7.1 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 134.8/134.8 kB 6.0 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 194.1/194.1 kB 6.8 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 62.7/62.7 kB 845.9 kB/s eta 0:00:00 ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts. cudf-cu12 24.4.1 requires pyarrow<15.0.0a0,>=14.0.1, but you have pyarrow 17.0.0 which is incompatible. ibis-framework 8.0.0 requires pyarrow<16,>=2, but you have pyarrow 17.0.0 which is incompatible.

from fastcoref import FCoref

model = FCoref(device='cuda:0')

model = FCoref(device='cpu')

---

AttributeError Traceback (most recent call last) in <cell line: 1>() ----> 1 from fastcoref import FCoref 2 # model = FCoref(device='cuda:0') 3 model = FCoref(device='cpu')

9 frames /usr/local/lib/python3.10/dist-packages/pyarrow/_compute.pyx in init pyarrow._compute()

AttributeError: module 'pyarrow.lib' has no attribute 'ListViewType'

import pyarrow pyarrow.version

14.0.2

[valerii-delphai]

!pip install -q pyarrow==14.0.1 datasets==2.10.0

[xiongjun926g]

Thanks for help but problems stays the same:

!pip install -q pyarrow==14.0.1 datasets==2.10.0 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38.0/38.0 MB 11.6 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 469.0/469.0 kB 12.7 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 110.5/110.5 kB 6.8 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 134.3/134.3 kB 4.7 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 194.1/194.1 kB 6.7 MB/s eta 0:00:00

!pip install fastcoref -q Preparing metadata (setup.py) ... done Building wheel for fastcoref (setup.py) ... done

from fastcoref import FCoref

model = FCoref(device='cuda:0')

model = FCoref(device='cpu')

/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/_token.py:89: UserWarning: The secret HF_TOKEN does not exist in your Colab secrets. To authenticate with the Hugging Face Hub, create a token in your settings tab (https://huggingface.co/settings/tokens), set it as secret in your Google Colab and restart your session. You will be able to reuse this secret in all of your notebooks. Please note that authentication is recommended but still optional to access public models or datasets. warnings.warn( config.json: 100%  819/819 [00:00<00:00, 14.2kB/s] tokenizer_config.json: 100%  393/393 [00:00<00:00, 4.15kB/s] vocab.json: 100%  798k/798k [00:00<00:00, 2.67MB/s] merges.txt: 100%  456k/456k [00:00<00:00, 3.61MB/s] tokenizer.json: 100%  1.36M/1.36M [00:00<00:00, 3.82MB/s] special_tokens_map.json: 100%  239/239 [00:00<00:00, 4.80kB/s] /usr/local/lib/python3.10/dist-packages/transformers/tokenization_utils_base.py:1601: FutureWarning: clean_up_tokenization_spaces was not set. It will be set to True by default. This behavior will be depracted in transformers v4.45, and will be then set to False by default. For more details check this issue: https://github.com/huggingface/transformers/issues/31884 warnings.warn( pytorch_model.bin: 100%  362M/362M [00:04<00:00, 113MB/s]

ValueError Traceback (most recent call last) in <cell line: 3>() 1 from fastcoref import FCoref 2 # model = FCoref(device='cuda:0') ----> 3 model = FCoref(device='cpu')

7 frames /usr/local/lib/python3.10/dist-packages/transformers/modeling_utils.py in _check_and_enable_sdpa(cls, config, hard_check_only) 1729 if hard_check_only: 1730 if not cls._supports_sdpa: -> 1731 raise ValueError( 1732 f"{cls.name} does not support an attention implementation through torch.nn.functional.scaled_dot_product_attention yet." 1733 " Please request the support for this architecture: https://github.com/huggingface/transformers/issues/28005. If you believe"

ValueError: RobertaModel does not support an attention implementation through torch.nn.functional.scaled_dot_product_attention yet. Please request the support for this architecture: https://github.com/huggingface/transformers/issues/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")

[valerii-delphai]

...problems stays the same:

no, it's not the same :) The new error means transformers library was updated on Google Colab again. So just install an older version supported by fastcoref. I found, for example, transformers==4.31.0 works fine

[FanaticPythoner]

Hey. You can use this to go around the error you've seen. @xiongjun926g . Seems to load and do inference fine on transformers 4.45.1

from fastcoref import LingMessCoref as OriginalLingMessCoref
from fastcoref import FCoref as OriginalFCoref
from transformers import AutoModel
import functools

class PatchedLingMessCoref(OriginalLingMessCoref):
    def __init__(self, *args, **kwargs):
        original_from_config = AutoModel.from_config

        def patched_from_config(config, *args, **kwargs):
            kwargs['attn_implementation'] = 'eager'
            return original_from_config(config, *args, **kwargs)

        try:
            AutoModel.from_config = functools.partial(patched_from_config, attn_implementation='eager')
            super().__init__(*args, **kwargs)
        finally:
            AutoModel.from_config = original_from_config

class PatchedFCoref(OriginalFCoref):
    def __init__(self, *args, **kwargs):
        original_from_config = AutoModel.from_config

        def patched_from_config(config, *args, **kwargs):
            kwargs['attn_implementation'] = 'eager'
            return original_from_config(config, *args, **kwargs)

        try:
            AutoModel.from_config = functools.partial(patched_from_config, attn_implementation='eager')
            super().__init__(*args, **kwargs)
        finally:
            AutoModel.from_config = original_from_config
            
            
            
            
model1 = PatchedLingMessCoref(
    nlp="en_core_web_lg",
    device="cpu"
)

model2 = PatchedFCoref(
    nlp="en_core_web_lg",
    device="cpu"
)

# Run your stuff here

[pphui8]

Hey. You can use this to go around the error you've seen. @xiongjun926g . Seems to load and do inference fine on transformers 4.45.1

from fastcoref import LingMessCoref as OriginalLingMessCoref
from fastcoref import FCoref as OriginalFCoref
from transformers import AutoModel
import functools

class PatchedLingMessCoref(OriginalLingMessCoref):
    def __init__(self, *args, **kwargs):
        original_from_config = AutoModel.from_config

        def patched_from_config(config, *args, **kwargs):
            kwargs['attn_implementation'] = 'eager'
            return original_from_config(config, *args, **kwargs)

        try:
            AutoModel.from_config = functools.partial(patched_from_config, attn_implementation='eager')
            super().__init__(*args, **kwargs)
        finally:
            AutoModel.from_config = original_from_config

class PatchedFCoref(OriginalFCoref):
    def __init__(self, *args, **kwargs):
        original_from_config = AutoModel.from_config

        def patched_from_config(config, *args, **kwargs):
            kwargs['attn_implementation'] = 'eager'
            return original_from_config(config, *args, **kwargs)

        try:
            AutoModel.from_config = functools.partial(patched_from_config, attn_implementation='eager')
            super().__init__(*args, **kwargs)
        finally:
            AutoModel.from_config = original_from_config
            
            
            
            
model1 = PatchedLingMessCoref(
    nlp="en_core_web_lg",
    device="cpu"
)

model2 = PatchedFCoref(
    nlp="en_core_web_lg",
    device="cpu"
)

# Run your stuff here

This works for me, and the error output actually already states how to fix this problem: load your model with the argument attn_implementation="eager" meanwhile.

[FanaticPythoner]

@shon-otmazgin Can you please implement that fix. It's an easy one. Thanks.

[FanaticPythoner]

Any update...?