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

use pre-commit

Browse Source 
Former-commit-id: 7cfede95df22a9ff236788f04159b6b16b8d04bb
This commit is contained in:
hiyouga
2024-10-29 09:07:46 +00:00
parent 8f5921692e
commit 248d5daaff
66 changed files with 1028 additions and 1044 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
src/llamafactory/data/aligner.py
View File

@@ -161,7 +161,7 @@ def convert_sharegpt(
broken_data = False
for turn_idx, message in enumerate(messages):
if message[dataset_attr.role_tag] not in accept_tags[turn_idx % 2]:
logger.warning("Invalid role tag in {}.".format(messages))
logger.warning(f"Invalid role tag in {messages}.")
broken_data = True
aligned_messages.append(
@@ -171,7 +171,7 @@ def convert_sharegpt(
if (not dataset_attr.ranking and len(aligned_messages) % 2 != 0) or (
dataset_attr.ranking and len(aligned_messages) % 2 == 0
):
logger.warning("Invalid message count in {}.".format(messages))
logger.warning(f"Invalid message count in {messages}.")
broken_data = True
if dataset_attr.kto_tag and isinstance(example[dataset_attr.kto_tag], bool): # kto example
@@ -192,7 +192,7 @@ def convert_sharegpt(
chosen[dataset_attr.role_tag] not in accept_tags[-1]
or rejected[dataset_attr.role_tag] not in accept_tags[-1]
):
logger.warning("Invalid role tag in {}.".format([chosen, rejected]))
logger.warning(f"Invalid role tag in {[chosen, rejected]}.")
broken_data = True
prompt = aligned_messages

6
src/llamafactory/data/collator.py
View File

@@ -137,9 +137,9 @@ class PairwiseDataCollatorWithPadding(MultiModalDataCollatorForSeq2Seq):
for key in ("chosen", "rejected"):
for feature in features:
target_feature = {
"input_ids": feature["{}_input_ids".format(key)],
"attention_mask": feature["{}_attention_mask".format(key)],
"labels": feature["{}_labels".format(key)],
"input_ids": feature[f"{key}_input_ids"],
"attention_mask": feature[f"{key}_attention_mask"],
"labels": feature[f"{key}_labels"],
"images": feature["images"],
"videos": feature["videos"],
}

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

@@ -70,7 +70,7 @@ def merge_dataset(
stopping_strategy="first_exhausted" if data_args.mix_strategy.endswith("under") else "all_exhausted",
)
else:
raise ValueError("Unknown mixing strategy: {}.".format(data_args.mix_strategy))
raise ValueError(f"Unknown mixing strategy: {data_args.mix_strategy}.")
def split_dataset(

10
src/llamafactory/data/formatter.py
View File

@@ -83,14 +83,14 @@ class StringFormatter(Formatter):
if isinstance(slot, str):
for name, value in kwargs.items():
if not isinstance(value, str):
raise RuntimeError("Expected a string, got {}".format(value))
raise RuntimeError(f"Expected a string, got {value}")
slot = slot.replace("{{" + name + "}}", value, 1)
elements.append(slot)
elif isinstance(slot, (dict, set)):
elements.append(slot)
else:
raise RuntimeError("Input must be string, set[str] or dict[str, str], got {}".format(type(slot)))
raise RuntimeError(f"Input must be string, set[str] or dict[str, str], got {type(slot)}")
return elements
@@ -113,7 +113,7 @@ class FunctionFormatter(Formatter):
functions.append((tool_call["name"], json.dumps(tool_call["arguments"], ensure_ascii=False)))
except json.JSONDecodeError:
raise RuntimeError("Invalid JSON format in function message: {}".format(str([content]))) # flat string
raise RuntimeError(f"Invalid JSON format in function message: {str([content])}") # flat string
elements = []
for name, arguments in functions:
@@ -124,7 +124,7 @@ class FunctionFormatter(Formatter):
elif isinstance(slot, (dict, set)):
elements.append(slot)
else:
raise RuntimeError("Input must be string, set[str] or dict[str, str], got {}".format(type(slot)))
raise RuntimeError(f"Input must be string, set[str] or dict[str, str], got {type(slot)}")
return elements
@@ -141,7 +141,7 @@ class ToolFormatter(Formatter):
tools = json.loads(content)
return [self.tool_utils.tool_formatter(tools) if len(tools) != 0 else ""]
except json.JSONDecodeError:
raise RuntimeError("Invalid JSON format in tool description: {}".format(str([content]))) # flat string
raise RuntimeError(f"Invalid JSON format in tool description: {str([content])}") # flat string
@override
def extract(self, content: str) -> Union[str, List["FunctionCall"]]:

14
src/llamafactory/data/loader.py
View File

@@ -51,7 +51,7 @@ def _load_single_dataset(
r"""
Loads a single dataset and aligns it to the standard format.
"""
logger.info("Loading dataset {}...".format(dataset_attr))
logger.info(f"Loading dataset {dataset_attr}...")
data_path, data_name, data_dir, data_files = None, None, None, None
if dataset_attr.load_from in ["hf_hub", "ms_hub", "om_hub"]:
data_path = dataset_attr.dataset_name
@@ -77,12 +77,12 @@ def _load_single_dataset(
data_files.append(local_path)
data_path = FILEEXT2TYPE.get(local_path.split(".")[-1], None)
else:
raise ValueError("File {} not found.".format(local_path))
raise ValueError(f"File {local_path} not found.")
if data_path is None:
raise ValueError("Allowed file types: {}.".format(",".join(FILEEXT2TYPE.keys())))
else:
raise NotImplementedError("Unknown load type: {}.".format(dataset_attr.load_from))
raise NotImplementedError(f"Unknown load type: {dataset_attr.load_from}.")
if dataset_attr.load_from == "ms_hub":
require_version("modelscope>=1.11.0", "To fix: pip install modelscope>=1.11.0")
@@ -145,7 +145,7 @@ def _load_single_dataset(
assert len(indexes) == dataset_attr.num_samples, "Sample num mismatched."
dataset = dataset.select(indexes)
logger.info("Sampled {} examples from dataset {}.".format(dataset_attr.num_samples, dataset_attr))
logger.info(f"Sampled {dataset_attr.num_samples} examples from dataset {dataset_attr}.")
if data_args.max_samples is not None: # truncate dataset
max_samples = min(data_args.max_samples, len(dataset))
@@ -243,7 +243,7 @@ def get_dataset(
if has_tokenized_data(data_args.tokenized_path):
logger.warning("Loading dataset from disk will ignore other data arguments.")
dataset_dict: "DatasetDict" = load_from_disk(data_args.tokenized_path)
logger.info("Loaded tokenized dataset from {}.".format(data_args.tokenized_path))
logger.info(f"Loaded tokenized dataset from {data_args.tokenized_path}.")
dataset_module: Dict[str, "Dataset"] = {}
if "train" in dataset_dict:
@@ -294,8 +294,8 @@ def get_dataset(
if data_args.tokenized_path is not None:
if training_args.should_save:
dataset_dict.save_to_disk(data_args.tokenized_path)
logger.info("Tokenized dataset saved at {}.".format(data_args.tokenized_path))
logger.info("Please restart the training with `tokenized_path: {}`.".format(data_args.tokenized_path))
logger.info(f"Tokenized dataset saved at {data_args.tokenized_path}.")
logger.info(f"Please restart the training with `tokenized_path: {data_args.tokenized_path}`.")
sys.exit(0)

26
src/llamafactory/data/mm_plugin.py
View File

@@ -111,7 +111,7 @@ class BasePlugin:
image = Image.open(image["path"])
if not isinstance(image, ImageObject):
raise ValueError("Expect input is a list of Images, but got {}.".format(type(image)))
raise ValueError(f"Expect input is a list of Images, but got {type(image)}.")
results.append(self._preprocess_image(image, **kwargs))
@@ -253,7 +253,7 @@ class LlavaPlugin(BasePlugin):
message["content"] = content.replace("{{image}}", self.image_token * image_seqlen)
if len(images) != num_image_tokens:
raise ValueError("The number of images does not match the number of {} tokens".format(IMAGE_PLACEHOLDER))
raise ValueError(f"The number of images does not match the number of {IMAGE_PLACEHOLDER} tokens")
return messages
@@ -302,7 +302,7 @@ class LlavaNextPlugin(BasePlugin):
message["content"] = content.replace("{{image}}", self.image_token)
if len(images) != num_image_tokens:
raise ValueError("The number of images does not match the number of {} tokens".format(IMAGE_PLACEHOLDER))
raise ValueError(f"The number of images does not match the number of {IMAGE_PLACEHOLDER} tokens")
return messages
@override
@@ -366,10 +366,10 @@ class LlavaNextVideoPlugin(BasePlugin):
message["content"] = content.replace("{{video}}", self.video_token * video_seqlen)
if len(images) != num_image_tokens:
raise ValueError("The number of images does not match the number of {} tokens".format(IMAGE_PLACEHOLDER))
raise ValueError(f"The number of images does not match the number of {IMAGE_PLACEHOLDER} tokens")
if len(videos) != num_video_tokens:
raise ValueError("The number of videos does not match the number of {} tokens".format(IMAGE_PLACEHOLDER))
raise ValueError(f"The number of videos does not match the number of {IMAGE_PLACEHOLDER} tokens")
return messages
@@ -408,7 +408,7 @@ class PaliGemmaPlugin(BasePlugin):
message["content"] = content.replace("{{image}}", "")
if len(images) != num_image_tokens:
raise ValueError("The number of images does not match the number of {} tokens".format(IMAGE_PLACEHOLDER))
raise ValueError(f"The number of images does not match the number of {IMAGE_PLACEHOLDER} tokens")
return messages
@@ -493,7 +493,7 @@ class Qwen2vlPlugin(BasePlugin):
content = message["content"]
while IMAGE_PLACEHOLDER in content:
if num_image_tokens >= len(image_grid_thw):
raise ValueError("`len(images)` is less than the number of {} tokens.".format(IMAGE_PLACEHOLDER))
raise ValueError(f"`len(images)` is less than the number of {IMAGE_PLACEHOLDER} tokens.")
content = content.replace(
IMAGE_PLACEHOLDER,
@@ -506,7 +506,7 @@ class Qwen2vlPlugin(BasePlugin):
while VIDEO_PLACEHOLDER in content:
if num_video_tokens >= len(video_grid_thw):
raise ValueError("`len(videos)` is less than the number of {} tokens.".format(VIDEO_PLACEHOLDER))
raise ValueError(f"`len(videos)` is less than the number of {VIDEO_PLACEHOLDER} tokens.")
content = content.replace(
VIDEO_PLACEHOLDER,
@@ -520,10 +520,10 @@ class Qwen2vlPlugin(BasePlugin):
message["content"] = content
if len(images) != num_image_tokens:
raise ValueError("The number of images does not match the number of {} tokens".format(IMAGE_PLACEHOLDER))
raise ValueError(f"The number of images does not match the number of {IMAGE_PLACEHOLDER} tokens")
if len(videos) != num_video_tokens:
raise ValueError("The number of videos does not match the number of {} tokens".format(VIDEO_PLACEHOLDER))
raise ValueError(f"The number of videos does not match the number of {VIDEO_PLACEHOLDER} tokens")
return messages
@@ -583,10 +583,10 @@ class VideoLlavaPlugin(BasePlugin):
message["content"] = content.replace("{{video}}", self.video_token * video_seqlen)
if len(images) != num_image_tokens:
raise ValueError("The number of images does not match the number of {} tokens".format(self.image_token))
raise ValueError(f"The number of images does not match the number of {self.image_token} tokens")
if len(videos) != num_video_tokens:
raise ValueError("The number of videos does not match the number of {} tokens".format(self.video_token))
raise ValueError(f"The number of videos does not match the number of {self.video_token} tokens")
return messages
@@ -622,6 +622,6 @@ def get_mm_plugin(
) -> "BasePlugin":
plugin_class = PLUGINS.get(name, None)
if plugin_class is None:
raise ValueError("Multimodal plugin `{}` not found.".format(name))
raise ValueError(f"Multimodal plugin `{name}` not found.")
return plugin_class(image_token, video_token)

6
src/llamafactory/data/parser.py
View File

@@ -87,11 +87,11 @@ def get_dataset_list(dataset_names: Optional[Sequence[str]], dataset_dir: str) -
config_path = os.path.join(dataset_dir, DATA_CONFIG)
try:
with open(config_path, "r") as f:
with open(config_path) as f:
dataset_info = json.load(f)
except Exception as err:
if len(dataset_names) != 0:
raise ValueError("Cannot open {} due to {}.".format(config_path, str(err)))
raise ValueError(f"Cannot open {config_path} due to {str(err)}.")
dataset_info = None
@@ -109,7 +109,7 @@ def get_dataset_list(dataset_names: Optional[Sequence[str]], dataset_dir: str) -
continue
if name not in dataset_info:
raise ValueError("Undefined dataset {} in {}.".format(name, DATA_CONFIG))
raise ValueError(f"Undefined dataset {name} in {DATA_CONFIG}.")
has_hf_url = "hf_hub_url" in dataset_info[name]
has_ms_url = "ms_hub_url" in dataset_info[name]

4
src/llamafactory/data/processors/pairwise.py
View File

@@ -110,8 +110,8 @@ def print_pairwise_dataset_example(example: Dict[str, List[int]], tokenizer: "Pr
print("chosen_input_ids:\n{}".format(example["chosen_input_ids"]))
print("chosen_inputs:\n{}".format(tokenizer.decode(example["chosen_input_ids"], skip_special_tokens=False)))
print("chosen_label_ids:\n{}".format(example["chosen_labels"]))
print("chosen_labels:\n{}".format(tokenizer.decode(valid_chosen_labels, skip_special_tokens=False)))
print(f"chosen_labels:\n{tokenizer.decode(valid_chosen_labels, skip_special_tokens=False)}")
print("rejected_input_ids:\n{}".format(example["rejected_input_ids"]))
print("rejected_inputs:\n{}".format(tokenizer.decode(example["rejected_input_ids"], skip_special_tokens=False)))
print("rejected_label_ids:\n{}".format(example["rejected_labels"]))
print("rejected_labels:\n{}".format(tokenizer.decode(valid_rejected_labels, skip_special_tokens=False)))
print(f"rejected_labels:\n{tokenizer.decode(valid_rejected_labels, skip_special_tokens=False)}")

4
src/llamafactory/data/processors/supervised.py
View File

@@ -160,7 +160,7 @@ def preprocess_packed_supervised_dataset(
)
length = len(input_ids)
if length > data_args.cutoff_len:
logger.warning("Dropped lengthy example with length {} > {}.".format(length, data_args.cutoff_len))
logger.warning(f"Dropped lengthy example with length {length} > {data_args.cutoff_len}.")
else:
lengths.append(length)
length2indexes[length].append(valid_num)
@@ -212,4 +212,4 @@ def print_supervised_dataset_example(example: Dict[str, List[int]], tokenizer: "
print("input_ids:\n{}".format(example["input_ids"]))
print("inputs:\n{}".format(tokenizer.decode(example["input_ids"], skip_special_tokens=False)))
print("label_ids:\n{}".format(example["labels"]))
print("labels:\n{}".format(tokenizer.decode(valid_labels, skip_special_tokens=False)))
print(f"labels:\n{tokenizer.decode(valid_labels, skip_special_tokens=False)}")

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

@@ -147,7 +147,7 @@ class Template:
elif "eos_token" in elem and tokenizer.eos_token_id is not None:
token_ids += [tokenizer.eos_token_id]
else:
raise ValueError("Input must be string, set[str] or dict[str, str], got {}".format(type(elem)))
raise ValueError(f"Input must be string, set[str] or dict[str, str], got {type(elem)}")
return token_ids
@@ -275,9 +275,9 @@ def _add_or_replace_eos_token(tokenizer: "PreTrainedTokenizer", eos_token: str)
num_added_tokens = tokenizer.add_special_tokens({"eos_token": eos_token})
if is_added:
logger.info("Add eos token: {}".format(tokenizer.eos_token))
logger.info(f"Add eos token: {tokenizer.eos_token}")
else:
logger.info("Replace eos token: {}".format(tokenizer.eos_token))
logger.info(f"Replace eos token: {tokenizer.eos_token}")
if num_added_tokens > 0:
logger.warning("New tokens have been added, make sure `resize_vocab` is True.")
@@ -365,13 +365,13 @@ def get_template_and_fix_tokenizer(tokenizer: "PreTrainedTokenizer", data_args:
else:
template = TEMPLATES.get(data_args.template, None)
if template is None:
raise ValueError("Template {} does not exist.".format(data_args.template))
raise ValueError(f"Template {data_args.template} does not exist.")
if data_args.train_on_prompt and template.efficient_eos:
raise ValueError("Current template does not support `train_on_prompt`.")
if data_args.tool_format is not None:
logger.info("Using tool format: {}.".format(data_args.tool_format))
logger.info(f"Using tool format: {data_args.tool_format}.")
eos_slots = [] if template.efficient_eos else [{"eos_token"}]
template.format_function = FunctionFormatter(slots=eos_slots, tool_format=data_args.tool_format)
template.format_tools = ToolFormatter(tool_format=data_args.tool_format)
@@ -389,7 +389,7 @@ def get_template_and_fix_tokenizer(tokenizer: "PreTrainedTokenizer", data_args:
if tokenizer.pad_token_id is None:
tokenizer.pad_token = tokenizer.eos_token
logger.info("Add pad token: {}".format(tokenizer.pad_token))
logger.info(f"Add pad token: {tokenizer.pad_token}")
if stop_words:
num_added_tokens = tokenizer.add_special_tokens(

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

@@ -177,6 +177,6 @@ TOOLS = {
def get_tool_utils(name: str) -> "ToolUtils":
tool_utils = TOOLS.get(name, None)
if tool_utils is None:
raise ValueError("Tool utils `{}` not found.".format(name))
raise ValueError(f"Tool utils `{name}` not found.")
return tool_utils