423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-08-22 21:52:51 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix paligemma sft

Browse Source 
Former-commit-id: 7134fb02bbdc9421f6c314ae176d5786a8cd768d
This commit is contained in:
hiyouga 2024-05-21 20:03:09 +08:00
parent a8480baa11
commit 4ddc1c9c16
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/llamafactory/data/preprocess.py
View File

@ -89,7 +89,7 @@ def preprocess_supervised_dataset(
if processor is not None and hasattr(processor, "image_seq_length"): # paligemma case if processor is not None and hasattr(processor, "image_seq_length"): # paligemma case
image_token_id = tokenizer.convert_tokens_to_ids(IMAGE_TOKEN) image_token_id = tokenizer.convert_tokens_to_ids(IMAGE_TOKEN)
input_ids += [image_token_id] * getattr(processor, "image_seq_length") input_ids += [image_token_id] * getattr(processor, "image_seq_length")
labels += [image_token_id] * getattr(processor, "image_seq_length") labels += [IGNORE_INDEX] * getattr(processor, "image_seq_length")
for turn_idx, (source_ids, target_ids) in enumerate( for turn_idx, (source_ids, target_ids) in enumerate(
template.encode_multiturn( template.encode_multiturn(