ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Fix/patch tokenizer (#1171)

Browse source 
* fix: correct tokenizer handling in patch_sft_trainer_tokenizer

* Revert "fix: correct tokenizer handling in patch_sft_trainer_tokenizer"

This reverts commit 7a98e465cbd4f980c8b364b0396d44f2d052090f.

* fix: correct condition for test_text assignment in patch_sft_trainer_tokenizer
This commit is contained in:
Edd 2024-10-23 23:32:33 +04:00 • committed by GitHub
commit f402d945ff
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
unsloth/tokenizer_utils.py
View file

@ -915,7 +915,7 @@ def patch_sft_trainer_tokenizer():
check_text = \
"\n"\
"if 'tokenizer' not in locals(): tokenizer = processing_class\n"\
"test_text = dataset[0][dataset_text_field] if (formatting_func is not None and dataset_text_field is None) else formatting_func(dataset[0])[0]\n"\
"test_text = dataset[0][dataset_text_field] if (formatting_func is None and dataset_text_field is not None) else formatting_func(dataset[0])[0]\n"\
"chat_template = getattr(tokenizer, 'chat_template', None)\n"\
"chat_template = '' if chat_template is None else chat_template\n"\
"has_bos_token_already = (test_text.startswith(tokenizer.bos_token) or tokenizer.bos_token in chat_template) "\