-
-
Save avidale/44cd35bfcdaf8bedf51d97c468cc8001 to your computer and use it in GitHub Desktop.
When you say "arguments for training", where exactly do you use them? Are you using a huggingface trainer or something else?
If you give me a minimal example of code that can reproduce your problem, it would be easier for me to help.
My first guess is that you should replace
tokenizer_name_or_path="https://huggingface.co/yhavinga/t5-base-dutch/blob/main/tokenizer.json"
with simply
tokenizer_name_or_path="yhavinga/t5-base-dutch"
but without more context, I cannot be sure.
Thanks for your response @avidale , Yes I'm using a HF trainer and the arguments are for that. I did made the change as suggested by you but getting a different error as below
-> 310 return _sentencepiece.SentencePieceProcessor_LoadFromFile(self, arg)
311
312 def _EncodeAsIds(self, text, enable_sampling, nbest_size, alpha, add_bos, add_eos, reverse, emit_unk_piece):
TypeError: not a string
I've been following this notebook ("https://colab.research.google.com/github/patil-suraj/exploring-T5/blob/master/t5_fine_tuning.ipynb#scrollTo=hcKmeIGiI582"). Appreciate the help!
Now it looks like a problem with incorrect input.
But again, without knowing the exact code that led to the error, I cannot say for sure.
Would this link of gist help : https://gist.github.com/Sandeep0408/236b164cb09408c920aedb15d5c7e984
If not, I can give you the access for the colab notebook via mail. Thanks!
Hello, I would like to know what version of python you are using, I saved the model as model.safetensors instead of pytorch_model.bin, please do you have any solution, thank you very much
Should this work with XLMRobertaModel, like e5-large? Or is something fundamentally different being used there. It didn't work out for me.
Should this work with XLMRobertaModel, like e5-large? Or is something fundamentally different being used there. It didn't work out for me.
As I can judge from the HF documentation, XLMRobertaTokenizer is based on SentencePiece, just like T5Tokenizer. Thus, in principle, the approach should work; I don't see any fundamental reasons why it wouldn't.
Nevertheless, the specific details, such as model parameter names, tokenizer parameter names, special tokens etc. may differ between T5 and XLMRoberta, so my code will surely need some adaptation to work with E5.
Hi @avidale, I'm trying to run a sentiment classification on a Dutch dataset using the tokenizer as :
tokenizer = T5TokenizerFast.from_pretrained('yhavinga/t5-base-dutch')
and below arguments for training :
model_name_or_path="yhavinga/t5-base-dutch",
tokenizer_name_or_path="https://huggingface.co/yhavinga/t5-base-dutch/blob/main/tokenizer.json"
When trying to train the model , getting an error
170 def LoadFromFile(self, arg):
--> 171 return _sentencepiece.SentencePieceProcessor_LoadFromFile(self, arg)
RuntimeError: Internal: /tmp/pip-install-ilecn6h0/sentencepiece_3c5f89f9146b4090a668d0f42db59389/bundled/sentencepiece/src/sentencepiece_processor.cc(823) [model_proto->ParseFromArray(serialized.data(), serialized.size())].
I'm done following changes but still no luck
Could you help? Thanks in Advance