ros/LLaMA-Factory
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

update readme

Browse Source 
Former-commit-id: b8d0170fe0d094acce85dcb5f91775e4685ee055
This commit is contained in:
hiyouga
2024-05-27 18:14:02 +08:00
parent b0d9966663
commit 97a23e1cbe
10 changed files with 71 additions and 62 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
src/llamafactory/data/processors/feedback.py
View File

@@ -1,6 +1,6 @@
from typing import TYPE_CHECKING, Any, Dict, List, Optional
from ...extras.constants import IGNORE_INDEX, IMAGE_TOKEN
from ...extras.constants import IGNORE_INDEX
from ...extras.logging import get_logger
from .mm_utils import get_paligemma_token_type_ids, get_pixel_values
@@ -46,7 +46,7 @@ def preprocess_feedback_dataset(
continue
if processor is not None and not hasattr(processor, "image_seq_length"): # llava-like models
examples["prompt"][i][0]["content"] = IMAGE_TOKEN + examples["prompt"][i][0]["content"]
examples["prompt"][i][0]["content"] = template.image_token + examples["prompt"][i][0]["content"]
if examples["response"][i][0]["content"]: # desired example
kto_tag = True
@@ -82,7 +82,7 @@ def preprocess_feedback_dataset(
kl_response_ids += [tokenizer.eos_token_id]
if processor is not None and hasattr(processor, "image_seq_length"): # paligemma models
image_token_id = tokenizer.convert_tokens_to_ids(IMAGE_TOKEN)
image_token_id = tokenizer.convert_tokens_to_ids(template.image_token)
prompt_ids = [image_token_id] * getattr(processor, "image_seq_length") + prompt_ids
input_ids = prompt_ids + response_ids