disentangle model from tuner and rename modules
Former-commit-id: 02cbf91e7e424f8379c1fed01b82a5f7a83b6947
This commit is contained in:
27
src/llmtuner/train/rm/collator.py
Normal file
27
src/llmtuner/train/rm/collator.py
Normal file
@@ -0,0 +1,27 @@
|
||||
import torch
|
||||
from dataclasses import dataclass
|
||||
from typing import Any, Dict, Sequence
|
||||
from transformers import DataCollatorWithPadding
|
||||
|
||||
|
||||
@dataclass
|
||||
class PairwiseDataCollatorWithPadding(DataCollatorWithPadding):
|
||||
r"""
|
||||
Data collator for pairwise data.
|
||||
"""
|
||||
|
||||
def __call__(self, features: Sequence[Dict[str, Any]]) -> Dict[str, torch.Tensor]:
|
||||
r"""
|
||||
Pads batched data to the longest sequence in the batch.
|
||||
|
||||
We generate 2 * n examples where the first n examples represent chosen examples and
|
||||
the last n examples represent rejected examples.
|
||||
"""
|
||||
features = [
|
||||
{
|
||||
"input_ids": feature["prompt_ids"] + feature[key],
|
||||
"attention_mask": [1] * (len(feature["prompt_ids"]) + len(feature[key]))
|
||||
}
|
||||
for key in ("chosen_ids", "rejected_ids") for feature in features
|
||||
]
|
||||
return super().__call__(features)
|
||||
Reference in New Issue
Block a user