423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-15 03:10:35 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix reserved label len

Browse Source 
Former-commit-id: db0ab4d601
This commit is contained in:
hiyouga
2024-02-04 17:54:26 +08:00
parent f6b2bcfa16
commit 91b7fd1ec3
3 changed files with 33 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
src/llmtuner/data/preprocess.py
View File

@@ -55,7 +55,12 @@ def preprocess_supervised_dataset(
input_ids, labels = [], []
for turn_idx, (source_ids, target_ids) in enumerate(
template.encode_multiturn(
tokenizer, messages, examples["system"][i], examples["tools"][i], data_args.cutoff_len
tokenizer,
messages,
examples["system"][i],
examples["tools"][i],
data_args.cutoff_len,
data_args.reserved_label_len,
)
):
if data_args.train_on_prompt:
@@ -143,7 +148,12 @@ def preprocess_unsupervised_dataset(
messages = examples["prompt"][i] + [{"role": Role.ASSISTANT, "content": ""}]
input_ids, labels = template.encode_oneturn(
tokenizer, messages, examples["system"][i], examples["tools"][i], data_args.cutoff_len
tokenizer,
messages,
examples["system"][i],
examples["tools"][i],
data_args.cutoff_len,
data_args.reserved_label_len,
)
if template.efficient_eos:
@@ -172,10 +182,20 @@ def preprocess_pairwise_dataset(
rejected_messages = examples["prompt"][i] + [examples["response"][i][1]]
prompt_ids, chosen_ids = template.encode_oneturn(
tokenizer, chosen_messages, examples["system"][i], examples["tools"][i], data_args.cutoff_len
tokenizer,
chosen_messages,
examples["system"][i],
examples["tools"][i],
data_args.cutoff_len,
data_args.reserved_label_len,
)
_, rejected_ids = template.encode_oneturn(
tokenizer, rejected_messages, examples["system"][i], examples["tools"][i], data_args.cutoff_len
tokenizer,
rejected_messages,
examples["system"][i],
examples["tools"][i],
data_args.cutoff_len,
data_args.reserved_label_len,
)
if template.efficient_eos:

12
src/llmtuner/data/template.py
View File

@@ -37,7 +37,7 @@ class Template:
system: Optional[str] = None,
tools: Optional[str] = None,
cutoff_len: Optional[int] = 1_000_000,
reserved_label_len: Optional[int] = 16,
reserved_label_len: Optional[int] = 1,
) -> Tuple[List[int], List[int]]:
r"""
Returns a single pair of token ids representing prompt and response respectively.
@@ -57,7 +57,7 @@ class Template:
system: Optional[str] = None,
tools: Optional[str] = None,
cutoff_len: Optional[int] = 1_000_000,
reserved_label_len: Optional[int] = 16,
reserved_label_len: Optional[int] = 1,
) -> Sequence[Tuple[List[int], List[int]]]:
r"""
Returns multiple pairs of token ids representing prompts and responses respectively.
@@ -144,10 +144,10 @@ class Template:
max_len=(cutoff_len - total_length),
reserved_label_len=reserved_label_len,
)
encoded_messages[i] = encoded_messages[i][:max_source_len]
encoded_messages[i + 1] = encoded_messages[i + 1][:max_target_len]
total_length += len(encoded_messages[i]) + len(encoded_messages[i + 1])
encoded_pairs.append((encoded_messages[i], encoded_messages[i + 1]))
source_ids = encoded_messages[i][:max_source_len]
target_ids = encoded_messages[i + 1][:max_target_len]
total_length += len(source_ids) + len(target_ids)
encoded_pairs.append((source_ids, target_ids))
return encoded_pairs