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

[example] add bash usage (#7794)

Browse Source
This commit is contained in:
hoshi-hiyouga
2025-04-22 00:25:51 +08:00
committed by GitHub
parent 12ada72ed4
commit b07628dea5
13 changed files with 184 additions and 98 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
src/llamafactory/model/model_utils/quantization.py
View File

@@ -124,6 +124,7 @@ def configure_quantization(
try:
from optimum.gptq import utils as gq_utils
if "language_model.model.layers" not in gq_utils.BLOCK_PATTERNS:
gq_utils.BLOCK_PATTERNS.insert(0, "language_model.model.layers")
except ImportError:

26
src/llamafactory/model/patcher.py
View File

@@ -54,26 +54,22 @@ def patch_tokenizer(tokenizer: "PreTrainedTokenizer", model_args: "ModelArgument
if model_args.model_max_length is not None and tokenizer.model_max_length < model_args.model_max_length:
tokenizer.model_max_length = model_args.model_max_length # enlarge the tokenizer max length
if model_args.new_special_tokens is not None:
num_added_special_tokens = tokenizer.add_special_tokens(
dict(additional_special_tokens=model_args.new_special_tokens),
replace_additional_special_tokens=False,
if model_args.add_tokens is not None:
num_added_tokens = tokenizer.add_tokens(new_tokens=model_args.add_tokens, special_tokens=False)
logger.info_rank0("Add tokens {} to tokenizer's vocabulary.".format(",".join(model_args.add_tokens)))
if num_added_tokens > 0 and not model_args.resize_vocab:
model_args.resize_vocab = True
logger.warning_rank0("New tokens have been added, changed `resize_vocab` to True.")
if model_args.add_special_tokens is not None:
num_added_special_tokens = tokenizer.add_tokens(new_tokens=model_args.add_special_tokens, special_tokens=True)
logger.info_rank0(
"Add special tokens {} to tokenizer's vocabulary.".format(",".join(model_args.add_special_tokens))
)
logger.info_rank0("Add special tokens {} to vocab.".format(",".join(model_args.new_special_tokens)))
if num_added_special_tokens > 0 and not model_args.resize_vocab:
model_args.resize_vocab = True
logger.warning_rank0("New special tokens have been added, changed `resize_vocab` to True.")
if model_args.new_normal_tokens is not None:
num_added_normal_tokens = tokenizer.add_tokens(
new_tokens=model_args.new_normal_tokens,
special_tokens=False,
)
logger.info_rank0("Add normal tokens {} to vocab.".format(",".join(model_args.new_normal_tokens)))
if num_added_normal_tokens > 0 and not model_args.resize_vocab:
model_args.resize_vocab = True
logger.warning_rank0("New normal tokens have been added, changed `resize_vocab` to True.")
def patch_processor(
processor: "ProcessorMixin",