RepositoryNotFoundError: 404 Client Error

Beginners
1

Hi everyone, I am new to NLP and working with HuggingFace. I am working on a text summarization project and trying to fine tune the model. Below is the code I wrote but I am getting the error which I am not able to solve. Any leads would be appreciated.

from huggingface_hub import notebook_login

notebook_login()

I passed here my huggingface token…

from transformers import Seq2SeqTrainingArguments

batch_size = 8
num_train_epochs = 8
# Show the training loss with every epoch
logging_steps = len(tokenized_datasets["article"]) // batch_size
model_name = model_checkpoint.split("/")[-1]

args = Seq2SeqTrainingArguments(
    output_dir="https://huggingface.co/username/mT5",
    evaluation_strategy="epoch",
    learning_rate=5.6e-5,
    per_device_train_batch_size=batch_size,
    per_device_eval_batch_size=batch_size,
    weight_decay=0.01,
    save_total_limit=3,
    num_train_epochs=num_train_epochs,
    predict_with_generate=True,
    logging_steps=logging_steps,
    push_to_hub=True
)

def compute_metrics(eval_pred):
    predictions, labels = eval_pred
    # Decode generated summaries into text
    decoded_preds = tokenizer.batch_decode(predictions, skip_special_tokens=True)
    # Replace -100 in the labels as we can't decode them
    labels = np.where(labels != -100, labels, tokenizer.pad_token_id)
    # Decode reference summaries into text
    decoded_labels = tokenizer.batch_decode(labels, skip_special_tokens=True)
    # ROUGE expects a newline after each sentence
    decoded_preds = ["\n".join(sent_tokenize(pred.strip())) for pred in decoded_preds]
    decoded_labels = ["\n".join(sent_tokenize(label.strip())) for label in decoded_labels]
    # Compute ROUGE scores
    result = rouge.compute(
        predictions=decoded_preds, references=decoded_labels, use_stemmer=True
    )
    # Extract the median scores
    result = {key: value.mid.fmeasure * 100 for key, value in result.items()}
    return {k: round(v, 4) for k, v in result.items()}

from transformers import DataCollatorForSeq2Seq

data_collator = DataCollatorForSeq2Seq(tokenizer, model=model)

from transformers import Seq2SeqTrainer

trainer = Seq2SeqTrainer(
    model = model,
    args = args,
    train_dataset=tokenized_datasets["article"],
    eval_dataset=test_tokenized_datasets["article"],
    data_collator=data_collator,
    tokenizer=tokenizer,
    compute_metrics=compute_metrics,
)

Here is the final error I am getting…

RepositoryNotFoundError: 404 Client Error. (Request ID: Root=1-64f004f3-5195d9d41d468e89023d924f;a7ed83d8-a09f-494e-ae69-8623b7517abb)

Repository Not Found for url: https://huggingface.co/api/models/mT5.
Please make sure you specified the correct repo_id and repo_type.
If you are trying to access a private or gated repo, make sure you are authenticated.

2

One issue which i found in argument output_dir of Seq2SeqTrainingArguments is it should be your local path rather than remote path and you cannot use a remote path over here.

This output directory helps us to save the model checkpoints and other stuffs .
see the docs of Training Arguments

3

Thanks for the suggestion. Along with the solution you mentioned, another issue was that the access token had just read access whereas we need to provide write access.

3 Likes