Fine-tuned DistilBERT won't load in Spaces (AssertionError)

Spaces
1

I fine-tuned a DistilBERT and uploaded it to the hub and my Space in March 2022 and I had not updated the code at all since then. Everything was working just fine until recently my Space was throwing this runtime error:

Runtime error
Space not ready. Reason: Error, exitCode: 1, message: None

Container logs:

Fetching model from: https://huggingface.co/m-newhauser/distilbert-political-tweets
Traceback (most recent call last):
  File "app.py", line 8, in <module>
    interface = gr.Interface.load("huggingface/m-newhauser/distilbert-political-tweets",
  File "/home/user/.local/lib/python3.8/site-packages/gradio/interface.py", line 73, in load
    interface_info = load_interface(name, src, api_key, alias)
  File "/home/user/.local/lib/python3.8/site-packages/gradio/external.py", line 270, in load_interface
    interface_info = repos[src](name, api_key, alias)
  File "/home/user/.local/lib/python3.8/site-packages/gradio/external.py", line 22, in get_huggingface_interface
    assert response.status_code == 200, "Invalid model name or src"
AssertionError: Invalid model name or src

System info
transformers version: v4.16.2
Platform: Linux-3.10.0-1160.62.1.el7.x86_64-x86_64-with-glibc2.17
Python version: 3.7.13
PyTorch version (GPU?): 1.11.0+cu113

Interestingly, when I try to load my model from the hub in a Colab notebook, this code executes perfectly:

from transformers import AutoTokenizer, AutoModelForSequenceClassification

model_base = AutoModelForSequenceClassification.from_pretrained("distilbert-base-uncased", 
                                                           num_labels=2)
model_finetuned = AutoModelForSequenceClassification.from_pretrained("m-newhauser/distilbert-political-tweets", 
                                                           num_labels=2)

I haven’t changed any of the code since I uploaded the model and it was working fine and the inference API on the model page was also returning predictions.

Please let me know if I’ve left out any information!

2

Hello @m-newhauser, sorry for the delay.
There is indeed an issue with your model, you added to the README library_name: pytorch which is not supported, valid values are either nothing (transformers by default), transformers or one of api-inference-community/docker_images at main · huggingface/api-inference-community · GitHub.
As for Colab it might work if the model was previously cached.

3

I changed library_name in my README file to transformers and everything works again. Would have never figured that one out. Thanks, @chris-rannou !