The download speed is very slow in China

[chncwang]

After I execute bert_score to run examples, it begins to download files, and the first file is downloaded fast, but the second file of size 1.43G, which I suppose is the BERT model file, is very verty slow, and my base is China. Is there any method to work around this problem? for example can I download BERT models elsewhere first and then use bert_score to run them?

[felixgwu]

Hi @chncwang,

It seems to be common that it's very slow to download files from s3 to China.
Since the model downloading part is handled within huggingface's transformers library, it would be better to contact them instead.

To use a downloaded model with bert-score, you can simply set --model path/to/my/roberta-large to specify the folder containing the pretrained model.

Best,
Felix

[chncwang]

Thank you!

[felixgwu]

Hi @chncwang,

I've put a copy of RoBERTa-large on Baidu Pan. I hope this helps.
Link: https://pan.baidu.com/s/1MTmGHsZ3ubn7Vr_W-wyEdQ
Password: dhe5

Best,
Felix

[chncwang]

Hi @chncwang,

I've put a copy of RoBERTa-large on Baidu Pan. I hope this helps.
Link: https://pan.baidu.com/s/1MTmGHsZ3ubn7Vr_W-wyEdQ
Password: dhe5

Best,
Felix

Many thanks! i am just struggling to search where to download bert models satisfying bert_score's format, but have not seen anywhere. I will try it right now!

[chncwang]

Hi @chncwang,

I've put a copy of RoBERTa-large on Baidu Pan. I hope this helps.
Link: https://pan.baidu.com/s/1MTmGHsZ3ubn7Vr_W-wyEdQ
Password: dhe5

Best,
Felix

It works!
Would you mind to facilitate me to download bert-base-chinese please? Since I plan to use bert-score to evaluate conversation models on Chinese. Thank you again!

[felixgwu]

Yes, here is the link: https://pan.baidu.com/s/1THfiCXjWtdGGsCMskQ5svA and password: jvk7
I hope this helps.

[chncwang]

It really helps me! Thank you again!

[chncwang]

Despite various ngram and embedding based automatic metric approaches I have tried, this is the first automatic metric showing our models' priority. It really saves me...