UKPLab / sentence-transformers

State-of-the-Art Text Embeddings
https://www.sbert.net
Apache License 2.0
14.78k stars 2.43k forks source link

Pretrained model loading timing out #817

Open iprivit opened 3 years ago

iprivit commented 3 years ago

When calling:

model = SentenceTransformer('distilbert-base-nli-stsb-mean-tokens')

I consistently receive timeout errors.

MaxRetryError: HTTPSConnectionPool(host='sbert.net', port=443): Max retries exceeded with url: /models/distilbert-base-nli-stsb-mean-tokens.zip (Caused by NewConnectionError('<urllib3.connection.HTTPSConnection object at 0x7fd33464acc0>: Failed to establish a new connection: [Errno 110] Connection timed out',))

It seems specific to sbert.net because I can load sentence-transformers models without any problems from Huggingface:

model = AutoModel.from_pretrained("sentence-transformers/bert-base-nli-mean-tokens")

nreimers commented 3 years ago

Sadly not sure why it happens. But you can download and unzip the models from www.sbert.net/models

AdesholaAfolabi commented 3 years ago

Hello, I am experiencing the same issue. When I do bertBase = SentenceTransformer('bert-base-nli-stsb-mean-tokens'), I get this error:

MaxRetryError: HTTPConnectionPool(host='sbert.net', port=80): Max retries exceeded with url: /models/bert-base-nli-stsb-mean-tokens.zip (Caused by NewConnectionError('<urllib3.connection.HTTPConnection object at 0x7fec8ef97358>: Failed to establish a new connection: [Errno 110] Connection timed out',))

nreimers commented 3 years ago

As before: Sadly not sure why it happens. But you can download and unzip the models from www.sbert.net/models

Does the issue still happen or was it a temporary problem?

Edit: Do other models than the bert-base-nli-stsb-mean-tokens work for you?

AdesholaAfolabi commented 3 years ago

As before: Sadly not sure why it happens. But you can download and unzip the models from www.sbert.net/models Thanks for this. I downloaded and unzipped it, so I have it working now.

Does the issue still happen or was it a temporary problem? It happened consistently, I thought it was an internet problem, I switched to faster internet and it was the same.

Edit: Do other models than the bert-base-nli-stsb-mean-tokens work for you? Yes, it was the same for other models too. Not sure why.