SYMBOL INDEX (1376 symbols across 213 files)

FILE: docs/_static/js/switcher.js
  function buildSwitcher (line 9) | function buildSwitcher() {
  function hideOtherLanguageToc (line 41) | function hideOtherLanguageToc() {

FILE: scripts/api_example/test_image.py
  function main (line 24) | def main():

FILE: scripts/api_example/test_toolcall.py
  function calculate_gpa (line 25) | def calculate_gpa(grades: list[str], hours: list[int]) -> float:
  function main (line 34) | def main():

FILE: scripts/bench_qwen.py
  class DummyDataset (line 32) | class DummyDataset(Dataset):
    method __init__ (line 33) | def __init__(self, size: int = 1000, seq_length: int = 1024, processor...
    method __len__ (line 54) | def __len__(self):
    method __getitem__ (line 57) | def __getitem__(self, index: int):
  class MultiModalDataCollatorForSeq2Seq (line 79) | class MultiModalDataCollatorForSeq2Seq(DataCollatorForSeq2Seq):
    method __post_init__ (line 80) | def __post_init__(self):
    method __call__ (line 91) | def __call__(self, features: list[dict[str, Any]]) -> dict[str, "torch...
  function bench_qwen (line 119) | def bench_qwen(

FILE: scripts/convert_ckpt/llamafy_baichuan2.py
  function save_weight (line 31) | def save_weight(input_dir: str, output_dir: str, shard_size: str, save_s...
  function save_config (line 76) | def save_config(input_dir: str, output_dir: str):
  function llamafy_baichuan2 (line 91) | def llamafy_baichuan2(

FILE: scripts/convert_ckpt/llamafy_qwen.py
  function save_weight (line 39) | def save_weight(input_dir: str, output_dir: str, shard_size: str, save_s...
  function save_config (line 113) | def save_config(input_dir: str, output_dir: str, torch_dtype: str):
  function llamafy_qwen (line 144) | def llamafy_qwen(

FILE: scripts/eval_bleu_rouge.py
  function compute_metrics (line 35) | def compute_metrics(sample):
  function main (line 61) | def main(filename: str):

FILE: scripts/hf2dcp.py
  function convert (line 31) | def convert(hf_path: str, dcp_path: str) -> None:
  function help (line 49) | def help() -> None:

FILE: scripts/llama_pro.py
  function change_name (line 36) | def change_name(name: str, old_index: int, new_index: int) -> str:
  function block_expansion (line 40) | def block_expansion(

FILE: scripts/loftq_init.py
  function quantize_loftq (line 30) | def quantize_loftq(

FILE: scripts/megatron_merge.py
  function convert_mca_to_hf (line 31) | def convert_mca_to_hf(
  function convert (line 64) | def convert(
  function main (line 125) | def main():

FILE: scripts/pissa_init.py
  function quantize_pissa (line 30) | def quantize_pissa(

FILE: scripts/qwen_omni_merge.py
  function merge_lora (line 36) | def merge_lora(
  function save_full_model (line 90) | def save_full_model(

FILE: scripts/stat_utils/cal_flops.py
  function calculate_flops (line 26) | def calculate_flops(

FILE: scripts/stat_utils/cal_lr.py
  function calculate_lr (line 37) | def calculate_lr(

FILE: scripts/stat_utils/cal_mfu.py
  function compute_model_flops (line 29) | def compute_model_flops(
  function compute_device_flops (line 86) | def compute_device_flops(world_size: int) -> float:
  function calculate_mfu (line 101) | def calculate_mfu(

FILE: scripts/stat_utils/cal_ppl.py
  class PairwiseDataCollatorWithPadding (line 32) | class PairwiseDataCollatorWithPadding(MultiModalDataCollatorForSeq2Seq):
    method __call__ (line 37) | def __call__(self, features: list[dict[str, Any]]) -> dict[str, torch....
  function calculate_ppl (line 55) | def calculate_ppl(

FILE: scripts/stat_utils/length_cdf.py
  function length_cdf (line 25) | def length_cdf(

FILE: scripts/vllm_infer.py
  function _need_video_kwargs (line 39) | def _need_video_kwargs(template):
  function vllm_infer (line 47) | def vllm_infer(

FILE: src/api.py
  function main (line 23) | def main():

FILE: src/llamafactory/api/app.py
  function sweeper (line 54) | async def sweeper() -> None:
  function lifespan (line 61) | async def lifespan(app: "FastAPI", chat_model: "ChatModel"):  # collects...
  function create_app (line 69) | def create_app(chat_model: "ChatModel") -> "FastAPI":
  function run_api (line 127) | def run_api() -> None:

FILE: src/llamafactory/api/chat.py
  function _process_request (line 73) | def _process_request(
  function _create_stream_chat_completion_chunk (line 181) | def _create_stream_chat_completion_chunk(
  function create_chat_completion_response (line 193) | async def create_chat_completion_response(
  function create_stream_chat_completion_response (line 247) | async def create_stream_chat_completion_response(
  function create_score_evaluation_response (line 286) | async def create_score_evaluation_response(

FILE: src/llamafactory/api/common.py
  function dictify (line 38) | def dictify(data: "BaseModel") -> dict[str, Any]:
  function jsonify (line 45) | def jsonify(data: "BaseModel") -> str:
  function check_lfi_path (line 52) | def check_lfi_path(path: str) -> None:
  function check_ssrf_url (line 70) | def check_ssrf_url(url: str) -> None:

FILE: src/llamafactory/api/protocol.py
  class Role (line 23) | class Role(StrEnum):
  class Finish (line 32) | class Finish(StrEnum):
  class ModelCard (line 38) | class ModelCard(BaseModel):
  class ModelList (line 45) | class ModelList(BaseModel):
  class Function (line 50) | class Function(BaseModel):
  class FunctionDefinition (line 55) | class FunctionDefinition(BaseModel):
  class FunctionAvailable (line 61) | class FunctionAvailable(BaseModel):
  class FunctionCall (line 66) | class FunctionCall(BaseModel):
  class URL (line 72) | class URL(BaseModel):
  class MultimodalInputItem (line 77) | class MultimodalInputItem(BaseModel):
  class ChatMessage (line 85) | class ChatMessage(BaseModel):
  class ChatCompletionMessage (line 91) | class ChatCompletionMessage(BaseModel):
  class ChatCompletionRequest (line 97) | class ChatCompletionRequest(BaseModel):
  class ChatCompletionResponseChoice (line 111) | class ChatCompletionResponseChoice(BaseModel):
  class ChatCompletionStreamResponseChoice (line 117) | class ChatCompletionStreamResponseChoice(BaseModel):
  class ChatCompletionResponseUsage (line 123) | class ChatCompletionResponseUsage(BaseModel):
  class ChatCompletionResponse (line 129) | class ChatCompletionResponse(BaseModel):
  class ChatCompletionStreamResponse (line 138) | class ChatCompletionStreamResponse(BaseModel):
  class ScoreEvaluationRequest (line 146) | class ScoreEvaluationRequest(BaseModel):
  class ScoreEvaluationResponse (line 152) | class ScoreEvaluationResponse(BaseModel):

FILE: src/llamafactory/chat/base_engine.py
  class Response (line 32) | class Response:
  class BaseEngine (line 39) | class BaseEngine(ABC):
    method __init__ (line 53) | def __init__(
    method chat (line 64) | async def chat(
    method stream_chat (line 78) | async def stream_chat(
    method get_scores (line 92) | async def get_scores(

FILE: src/llamafactory/chat/chat_model.py
  function _start_background_loop (line 34) | def _start_background_loop(loop: "asyncio.AbstractEventLoop") -> None:
  class ChatModel (line 39) | class ChatModel:
    method __init__ (line 47) | def __init__(self, args: Optional[dict[str, Any]] = None) -> None:
    method chat (line 91) | def chat(
    method achat (line 107) | async def achat(
    method stream_chat (line 120) | def stream_chat(
    method astream_chat (line 139) | async def astream_chat(
    method get_scores (line 155) | def get_scores(
    method aget_scores (line 164) | async def aget_scores(
  function run_chat (line 173) | def run_chat() -> None:

FILE: src/llamafactory/chat/hf_engine.py
  class HuggingfaceEngine (line 44) | class HuggingfaceEngine(BaseEngine):
    method __init__ (line 45) | def __init__(
    method _process_args (line 73) | def _process_args(
    method _chat (line 212) | def _chat(
    method _stream_chat (line 267) | def _stream_chat(
    method _get_scores (line 314) | def _get_scores(
    method chat (line 335) | async def chat(
    method stream_chat (line 366) | async def stream_chat(
    method get_scores (line 402) | async def get_scores(

FILE: src/llamafactory/chat/kt_engine.py
  class KTransformersEngine (line 51) | class KTransformersEngine(BaseEngine):
    method __init__ (line 52) | def __init__(
    method _get_scores (line 88) | def _get_scores(
    method _generate (line 108) | async def _generate(
    method chat (line 220) | async def chat(
    method stream_chat (line 254) | async def stream_chat(
    method get_scores (line 275) | async def get_scores(

FILE: src/llamafactory/chat/sglang_engine.py
  class SGLangEngine (line 46) | class SGLangEngine(BaseEngine):
    method __init__ (line 58) | def __init__(
    method _cleanup_server (line 130) | def _cleanup_server(self):
    method _generate (line 140) | async def _generate(
    method chat (line 232) | async def chat(
    method stream_chat (line 258) | async def stream_chat(
    method get_scores (line 276) | async def get_scores(
    method __del__ (line 283) | def __del__(self):

FILE: src/llamafactory/chat/vllm_engine.py
  class VllmEngine (line 46) | class VllmEngine(BaseEngine):
    method __init__ (line 47) | def __init__(
    method _generate (line 111) | async def _generate(
    method chat (line 216) | async def chat(
    method stream_chat (line 245) | async def stream_chat(
    method get_scores (line 263) | async def get_scores(

FILE: src/llamafactory/cli.py
  function main (line 16) | def main():

FILE: src/llamafactory/data/collator.py
  function prepare_4d_attention_mask (line 42) | def prepare_4d_attention_mask(attention_mask_with_indices: "torch.Tensor...
  class MultiModalDataCollatorForSeq2Seq (line 86) | class MultiModalDataCollatorForSeq2Seq(DataCollatorForSeq2Seq):
    method __post_init__ (line 95) | def __post_init__(self):
    method __call__ (line 109) | def __call__(self, features: list[dict[str, Any]]) -> dict[str, "torch...
  class SFTDataCollatorWith4DAttentionMask (line 258) | class SFTDataCollatorWith4DAttentionMask(MultiModalDataCollatorForSeq2Seq):
    method __call__ (line 265) | def __call__(self, features: list[dict[str, Any]]) -> dict[str, "torch...
  class PairwiseDataCollatorWithPadding (line 278) | class PairwiseDataCollatorWithPadding(MultiModalDataCollatorForSeq2Seq):
    method __call__ (line 281) | def __call__(self, features: list[dict[str, Any]]) -> dict[str, "torch...
  class KTODataCollatorWithPadding (line 304) | class KTODataCollatorWithPadding(MultiModalDataCollatorForSeq2Seq):
    method __call__ (line 307) | def __call__(self, features: list[dict[str, Any]]) -> dict[str, "torch...

FILE: src/llamafactory/data/converter.py
  class DatasetConverter (line 39) | class DatasetConverter:
    method _find_medias (line 43) | def _find_medias(self, medias: Union["MediaType", list["MediaType"], N...
    method __call__ (line 79) | def __call__(self, example: dict[str, Any]) -> dict[str, Any]:
  class AlpacaDatasetConverter (line 85) | class AlpacaDatasetConverter(DatasetConverter):
    method __call__ (line 86) | def __call__(self, example: dict[str, Any]) -> dict[str, Any]:
  class SharegptDatasetConverter (line 135) | class SharegptDatasetConverter(DatasetConverter):
    method __call__ (line 136) | def __call__(self, example: dict[str, Any]) -> dict[str, Any]:
  class OpenAIDatasetConverter (line 231) | class OpenAIDatasetConverter(DatasetConverter):
    method __call__ (line 232) | def __call__(self, example: dict[str, Any]) -> dict[str, Any]:
  function register_dataset_converter (line 377) | def register_dataset_converter(name: str, dataset_converter: type["Datas...
  function get_dataset_converter (line 385) | def get_dataset_converter(name: str, dataset_attr: "DatasetAttr", data_a...
  function align_dataset (line 393) | def align_dataset(

FILE: src/llamafactory/data/data_utils.py
  class Role (line 38) | class Role(StrEnum):
  class DatasetModule (line 46) | class DatasetModule(TypedDict):
  function merge_dataset (line 51) | def merge_dataset(
  function split_dataset (line 85) | def split_dataset(
  function get_dataset_module (line 134) | def get_dataset_module(dataset: Union["Dataset", "DatasetDict"]) -> "Dat...
  function setup_fs (line 158) | def setup_fs(path: str, anon: bool = False) -> "fsspec.AbstractFileSystem":
  function _read_json_with_fs (line 174) | def _read_json_with_fs(fs: "fsspec.AbstractFileSystem", path: str) -> li...
  function read_cloud_json (line 183) | def read_cloud_json(cloud_path: str) -> list[Any]:

FILE: src/llamafactory/data/formatter.py
  class Formatter (line 27) | class Formatter(ABC):
    method apply (line 32) | def apply(self, **kwargs) -> SLOTS:
    method extract (line 36) | def extract(self, content: str) -> str | list["FunctionCall"]:
  class EmptyFormatter (line 45) | class EmptyFormatter(Formatter):
    method __post_init__ (line 46) | def __post_init__(self):
    method apply (line 56) | def apply(self, **kwargs) -> SLOTS:
  class StringFormatter (line 61) | class StringFormatter(Formatter):
    method __post_init__ (line 62) | def __post_init__(self):
    method apply (line 72) | def apply(self, **kwargs) -> SLOTS:
  class FunctionFormatter (line 91) | class FunctionFormatter(StringFormatter):
    method __post_init__ (line 92) | def __post_init__(self):
    method apply (line 97) | def apply(self, **kwargs) -> SLOTS:
  class ToolFormatter (line 144) | class ToolFormatter(Formatter):
    method __post_init__ (line 145) | def __post_init__(self):
    method apply (line 149) | def apply(self, **kwargs) -> SLOTS:
    method extract (line 158) | def extract(self, content: str) -> str | list["FunctionCall"]:

FILE: src/llamafactory/data/loader.py
  function _load_single_dataset (line 51) | def _load_single_dataset(
  function _get_merged_dataset (line 164) | def _get_merged_dataset(
  function _get_dataset_processor (line 189) | def _get_dataset_processor(
  function _get_preprocessed_dataset (line 229) | def _get_preprocessed_dataset(
  function get_dataset (line 276) | def get_dataset(

FILE: src/llamafactory/data/mm_plugin.py
  class EncodedImage (line 65) | class EncodedImage(TypedDict):
  class RegularizedImageOutput (line 73) | class RegularizedImageOutput(TypedDict):
  class RegularizedVideoOutput (line 76) | class RegularizedVideoOutput(TypedDict):
  class RegularizedAudioOutput (line 81) | class RegularizedAudioOutput(TypedDict):
  class MMProcessor (line 85) | class MMProcessor(ProcessorMixin):
    method _get_number_of_features (line 91) | def _get_number_of_features(self, orig_height: int, orig_width: int, h...
  function _get_paligemma_token_type_ids (line 95) | def _get_paligemma_token_type_ids(imglens: list[int], seqlens: list[int]...
  function _get_gemma3_token_type_ids (line 112) | def _get_gemma3_token_type_ids(batch_ids: list[list[int]], processor: "M...
  function _make_batched_images (line 130) | def _make_batched_images(images: list["ImageObject"], imglens: list[int]...
  function _check_video_is_nested_images (line 140) | def _check_video_is_nested_images(video: "VideoInput") -> bool:
  class MMPluginMixin (line 146) | class MMPluginMixin:
    method _validate_input (line 152) | def _validate_input(
    method _validate_messages (line 194) | def _validate_messages(
    method _preprocess_image (line 223) | def _preprocess_image(
    method _get_video_sample_indices (line 242) | def _get_video_sample_indices(
    method _regularize_images (line 254) | def _regularize_images(self, images: list["ImageInput"], **kwargs) -> ...
    method _regularize_videos (line 275) | def _regularize_videos(self, videos: list["VideoInput"], **kwargs) -> ...
    method _regularize_audios (line 306) | def _regularize_audios(
    method _get_mm_inputs (line 327) | def _get_mm_inputs(
  class BasePlugin (line 417) | class BasePlugin(MMPluginMixin):
    method process_messages (line 418) | def process_messages(
    method process_token_ids (line 430) | def process_token_ids(
    method get_mm_inputs (line 444) | def get_mm_inputs(
  class ErnieVLPlugin (line 473) | class ErnieVLPlugin(BasePlugin):
    method process_messages (line 475) | def process_messages(
  class Gemma3Plugin (line 524) | class Gemma3Plugin(BasePlugin):
    method process_messages (line 526) | def process_messages(
    method get_mm_inputs (line 565) | def get_mm_inputs(
  class Gemma3nPlugin (line 583) | class Gemma3nPlugin(Gemma3Plugin):
    method process_messages (line 585) | def process_messages(
  class InternVLPlugin (line 617) | class InternVLPlugin(BasePlugin):
    method _get_mm_inputs (line 619) | def _get_mm_inputs(
    method process_messages (line 703) | def process_messages(
    method get_mm_inputs (line 748) | def get_mm_inputs(
  class KimiVLPlugin (line 767) | class KimiVLPlugin(BasePlugin):
    method process_messages (line 769) | def process_messages(self, messages, images, videos, audios, processor):
  class Llama4Plugin (line 799) | class Llama4Plugin(BasePlugin):
    method process_messages (line 801) | def process_messages(
    method get_mm_inputs (line 848) | def get_mm_inputs(
  class LlavaPlugin (line 866) | class LlavaPlugin(BasePlugin):
    method process_messages (line 868) | def process_messages(
  class LlavaNextPlugin (line 902) | class LlavaNextPlugin(BasePlugin):
    method process_messages (line 904) | def process_messages(
  class LlavaNextVideoPlugin (line 942) | class LlavaNextVideoPlugin(BasePlugin):
    method process_messages (line 944) | def process_messages(
  class MiniCPMVPlugin (line 997) | class MiniCPMVPlugin(BasePlugin):
    method _get_mm_inputs (line 999) | def _get_mm_inputs(
    method process_messages (line 1071) | def process_messages(
    method get_mm_inputs (line 1162) | def get_mm_inputs(
  class MllamaPlugin (line 1231) | class MllamaPlugin(BasePlugin):
    method process_messages (line 1233) | def process_messages(
    method get_mm_inputs (line 1253) | def get_mm_inputs(
  class PaliGemmaPlugin (line 1286) | class PaliGemmaPlugin(BasePlugin):
    method process_messages (line 1288) | def process_messages(
    method process_token_ids (line 1311) | def process_token_ids(
    method get_mm_inputs (line 1332) | def get_mm_inputs(
  class PixtralPlugin (line 1351) | class PixtralPlugin(BasePlugin):
    method process_messages (line 1353) | def process_messages(
    method get_mm_inputs (line 1398) | def get_mm_inputs(
  class Qwen2AudioPlugin (line 1420) | class Qwen2AudioPlugin(BasePlugin):
    method process_messages (line 1422) | def process_messages(
    method get_mm_inputs (line 1459) | def get_mm_inputs(
  class Qwen2VLPlugin (line 1475) | class Qwen2VLPlugin(BasePlugin):
    method _preprocess_image (line 1480) | def _preprocess_image(self, image: "ImageObject", **kwargs) -> "ImageO...
    method _regularize_videos (line 1497) | def _regularize_videos(self, videos: list["VideoInput"], **kwargs) -> ...
    method _get_mm_inputs (line 1534) | def _get_mm_inputs(
    method process_messages (line 1568) | def process_messages(
  class Qwen3VLPlugin (line 1617) | class Qwen3VLPlugin(Qwen2VLPlugin):
    method _get_mm_inputs (line 1619) | def _get_mm_inputs(
    method process_messages (line 1664) | def process_messages(
  class GLM4VPlugin (line 1740) | class GLM4VPlugin(Qwen2VLPlugin):
    method _get_mm_inputs (line 1742) | def _get_mm_inputs(
    method process_messages (line 1778) | def process_messages(
    method get_mm_inputs (line 1853) | def get_mm_inputs(
  class Qwen2OmniPlugin (line 1871) | class Qwen2OmniPlugin(Qwen2VLPlugin):
    method _get_mm_inputs (line 1876) | def _get_mm_inputs(
    method process_messages (line 1930) | def process_messages(
  class VideoLlavaPlugin (line 2052) | class VideoLlavaPlugin(BasePlugin):
    method process_messages (line 2054) | def process_messages(
  class LFMVLPlugin (line 2105) | class LFMVLPlugin(BasePlugin):
    method _get_mm_inputs (line 2114) | def _get_mm_inputs(
    method process_messages (line 2133) | def process_messages(
  class YoutuVLPlugin (line 2172) | class YoutuVLPlugin(BasePlugin):
    method process_messages (line 2179) | def process_messages(
  function register_mm_plugin (line 2231) | def register_mm_plugin(name: str, plugin_class: type["BasePlugin"]) -> N...
  function get_mm_plugin (line 2239) | def get_mm_plugin(

FILE: src/llamafactory/data/parser.py
  class DatasetAttr (line 27) | class DatasetAttr:
    method __repr__ (line 66) | def __repr__(self) -> str:
    method set_attr (line 69) | def set_attr(self, key: str, obj: dict[str, Any], default: Any | None ...
    method join (line 72) | def join(self, attr: dict[str, Any]) -> None:
  function get_dataset_list (line 93) | def get_dataset_list(dataset_names: list[str] | None, dataset_dir: str |...

FILE: src/llamafactory/data/processor/feedback.py
  class FeedbackDatasetProcessor (line 30) | class FeedbackDatasetProcessor(DatasetProcessor):
    method _encode_data_example (line 31) | def _encode_data_example(
    method preprocess_dataset (line 85) | def preprocess_dataset(self, examples: dict[str, list[Any]]) -> dict[s...
    method print_data_example (line 124) | def print_data_example(self, example: dict[str, list[int]]) -> None:

FILE: src/llamafactory/data/processor/pairwise.py
  class PairwiseDatasetProcessor (line 30) | class PairwiseDatasetProcessor(DatasetProcessor):
    method _encode_data_example (line 31) | def _encode_data_example(
    method preprocess_dataset (line 71) | def preprocess_dataset(self, examples: dict[str, list[Any]]) -> dict[s...
    method print_data_example (line 102) | def print_data_example(self, example: dict[str, list[int]]) -> None:

FILE: src/llamafactory/data/processor/pretrain.py
  class PretrainDatasetProcessor (line 26) | class PretrainDatasetProcessor(DatasetProcessor):
    method preprocess_dataset (line 27) | def preprocess_dataset(self, examples: dict[str, list[Any]]) -> dict[s...
    method print_data_example (line 55) | def print_data_example(self, example: dict[str, list[int]]) -> None:

FILE: src/llamafactory/data/processor/processor_utils.py
  class DatasetProcessor (line 29) | class DatasetProcessor(ABC):
    method preprocess_dataset (line 38) | def preprocess_dataset(self, examples: dict[str, list[Any]]) -> dict[s...
    method print_data_example (line 43) | def print_data_example(self, example: dict[str, list[int]]) -> None:
  function search_for_fit (line 48) | def search_for_fit(numbers: list[int], capacity: int) -> int:
  function greedy_knapsack (line 54) | def greedy_knapsack(numbers: list[int], capacity: int) -> list[list[int]]:
  function infer_seqlen (line 76) | def infer_seqlen(source_len: int, target_len: int, cutoff_len: int) -> t...

FILE: src/llamafactory/data/processor/supervised.py
  class SupervisedDatasetProcessor (line 32) | class SupervisedDatasetProcessor(DatasetProcessor):
    method _encode_data_example (line 33) | def _encode_data_example(
    method preprocess_dataset (line 88) | def preprocess_dataset(self, examples: dict[str, list[Any]]) -> dict[s...
    method print_data_example (line 117) | def print_data_example(self, example: dict[str, list[int]]) -> None:
  class PackedSupervisedDatasetProcessor (line 126) | class PackedSupervisedDatasetProcessor(SupervisedDatasetProcessor):
    method preprocess_dataset (line 127) | def preprocess_dataset(self, examples: dict[str, list[Any]]) -> dict[s...

FILE: src/llamafactory/data/processor/unsupervised.py
  class UnsupervisedDatasetProcessor (line 30) | class UnsupervisedDatasetProcessor(DatasetProcessor):
    method _encode_data_example (line 31) | def _encode_data_example(
    method preprocess_dataset (line 59) | def preprocess_dataset(self, examples: dict[str, list[Any]]) -> dict[s...
    method print_data_example (line 87) | def print_data_example(self, example: dict[str, list[int]]) -> None:

FILE: src/llamafactory/data/template.py
  class Template (line 41) | class Template:
    method encode_oneturn (line 59) | def encode_oneturn(
    method encode_multiturn (line 75) | def encode_multiturn(
    method extract_tool (line 86) | def extract_tool(self, content: str) -> Union[str, list["FunctionCall"]]:
    method get_stop_token_ids (line 90) | def get_stop_token_ids(self, tokenizer: "PreTrainedTokenizer") -> list...
    method add_thought (line 98) | def add_thought(self, content: str = "") -> str:
    method remove_thought (line 102) | def remove_thought(self, content: str) -> str:
    method get_thought_word_ids (line 107) | def get_thought_word_ids(self, tokenizer: "PreTrainedTokenizer") -> li...
    method _convert_elements_to_ids (line 111) | def _convert_elements_to_ids(self, tokenizer: "PreTrainedTokenizer", e...
    method _encode (line 130) | def _encode(
    method _add_or_replace_eos_token (line 171) | def _add_or_replace_eos_token(tokenizer: "PreTrainedTokenizer", eos_to...
    method fix_special_tokens (line 187) | def fix_special_tokens(self, tokenizer: "PreTrainedTokenizer") -> None:
    method _jinja_escape (line 216) | def _jinja_escape(content: str) -> str:
    method _convert_slots_to_jinja (line 221) | def _convert_slots_to_jinja(slots: "SLOTS", tokenizer: "PreTrainedToke...
    method _get_jinja_template (line 243) | def _get_jinja_template(self, tokenizer: "PreTrainedTokenizer") -> str:
    method fix_jinja_template (line 271) | def fix_jinja_template(self, tokenizer: "PreTrainedTokenizer") -> None:
    method _convert_slots_to_ollama (line 280) | def _convert_slots_to_ollama(
    method _get_ollama_template (line 304) | def _get_ollama_template(self, tokenizer: "PreTrainedTokenizer") -> str:
    method get_ollama_modelfile (line 316) | def get_ollama_modelfile(self, tokenizer: "PreTrainedTokenizer") -> str:
  class Llama2Template (line 335) | class Llama2Template(Template):
    method _encode (line 339) | def _encode(
    method _get_jinja_template (line 373) | def _get_jinja_template(self, tokenizer: "PreTrainedTokenizer") -> str:
  class ReasoningTemplate (line 405) | class ReasoningTemplate(Template):
    method encode_oneturn (line 409) | def encode_oneturn(
    method encode_multiturn (line 436) | def encode_multiturn(
  class Glm47ReasoningTemplate (line 463) | class Glm47ReasoningTemplate(ReasoningTemplate):
    method add_thought (line 467) | def add_thought(self, content: str = "") -> str:
  function register_template (line 477) | def register_template(
  function parse_template (line 550) | def parse_template(tokenizer: "PreTrainedTokenizer") -> "Template":
  function get_template_and_fix_tokenizer (line 612) | def get_template_and_fix_tokenizer(tokenizer: "PreTrainedTokenizer", dat...

FILE: src/llamafactory/data/tool_utils.py
  class FunctionCall (line 26) | class FunctionCall(NamedTuple):
  class ToolUtils (line 124) | class ToolUtils(ABC):
    method tool_formatter (line 129) | def tool_formatter(tools: list[dict[str, Any]]) -> str:
    method function_formatter (line 135) | def function_formatter(functions: list["FunctionCall"]) -> str:
    method tool_extractor (line 141) | def tool_extractor(content: str) -> Union[str, list["FunctionCall"]]:
  class DefaultToolUtils (line 149) | class DefaultToolUtils(ToolUtils):
    method tool_formatter (line 154) | def tool_formatter(tools: list[dict[str, Any]]) -> str:
    method function_formatter (line 189) | def function_formatter(functions: list["FunctionCall"]) -> str:
    method tool_extractor (line 194) | def tool_extractor(content: str) -> Union[str, list["FunctionCall"]]:
  class GLM4ToolUtils (line 213) | class GLM4ToolUtils(ToolUtils):
    method tool_formatter (line 218) | def tool_formatter(tools: list[dict[str, Any]]) -> str:
    method function_formatter (line 230) | def function_formatter(functions: list["FunctionCall"]) -> str:
    method tool_extractor (line 238) | def tool_extractor(content: str) -> Union[str, list["FunctionCall"]]:
  class Llama3ToolUtils (line 251) | class Llama3ToolUtils(ToolUtils):
    method tool_formatter (line 259) | def tool_formatter(tools: list[dict[str, Any]]) -> str:
    method function_formatter (line 270) | def function_formatter(functions: list["FunctionCall"]) -> str:
    method tool_extractor (line 276) | def tool_extractor(content: str) -> Union[str, list["FunctionCall"]]:
  class MiniMaxM1ToolUtils (line 289) | class MiniMaxM1ToolUtils(ToolUtils):
    method tool_formatter (line 294) | def tool_formatter(tools: list[dict[str, Any]]) -> str:
    method function_formatter (line 304) | def function_formatter(functions: list["FunctionCall"]) -> str:
    method tool_extractor (line 314) | def tool_extractor(content: str) -> Union[str, list["FunctionCall"]]:
  class MiniMaxM2ToolUtils (line 336) | class MiniMaxM2ToolUtils(ToolUtils):
    method tool_formatter (line 341) | def tool_formatter(tools: list[dict[str, Any]]) -> str:
    method function_formatter (line 351) | def function_formatter(functions: list["FunctionCall"]) -> str:
    method tool_extractor (line 366) | def tool_extractor(content: str) -> Union[str, list["FunctionCall"]]:
  class MistralToolUtils (line 392) | class MistralToolUtils(ToolUtils):
    method tool_formatter (line 397) | def tool_formatter(tools: list[dict[str, Any]]) -> str:
    method function_formatter (line 406) | def function_formatter(functions: list["FunctionCall"]) -> str:
    method tool_extractor (line 413) | def tool_extractor(content: str) -> Union[str, list["FunctionCall"]]:
  class QwenToolUtils (line 426) | class QwenToolUtils(ToolUtils):
    method tool_formatter (line 431) | def tool_formatter(tools: list[dict[str, Any]]) -> str:
    method function_formatter (line 441) | def function_formatter(functions: list["FunctionCall"]) -> str:
    method tool_extractor (line 450) | def tool_extractor(content: str) -> Union[str, list["FunctionCall"]]:
  class Qwen35ToolUtils (line 471) | class Qwen35ToolUtils(ToolUtils):
    method tool_formatter (line 476) | def tool_formatter(tools: list[dict[str, Any]]) -> str:
    method function_formatter (line 486) | def function_formatter(functions: list["FunctionCall"]) -> str:
    method tool_extractor (line 503) | def tool_extractor(content: str) -> Union[str, list["FunctionCall"]]:
  class GLM4MOEToolUtils (line 522) | class GLM4MOEToolUtils(QwenToolUtils):
    method tool_formatter (line 527) | def tool_formatter(tools: list[dict[str, Any]]) -> str:
    method function_formatter (line 537) | def function_formatter(functions: list["FunctionCall"]) -> str:
  class SeedToolUtils (line 554) | class SeedToolUtils(ToolUtils):
    method tool_formatter (line 559) | def tool_formatter(tools: list[dict[str, Any]]) -> str:
    method function_formatter (line 564) | def function_formatter(functions: list["FunctionCall"]) -> str:
    method tool_extractor (line 583) | def tool_extractor(content: str) -> Union[str, list["FunctionCall"]]:
  class LingToolUtils (line 604) | class LingToolUtils(QwenToolUtils):
    method tool_formatter (line 609) | def tool_formatter(tools: list[dict[str, Any]]) -> str:
  class LFM2ToolUtils (line 618) | class LFM2ToolUtils(ToolUtils):
    method tool_formatter (line 623) | def tool_formatter(tools: list[dict[str, Any]]) -> str:
    method function_formatter (line 633) | def function_formatter(functions: list["FunctionCall"]) -> str:
    method _ast_to_value (line 649) | def _ast_to_value(node: ast.AST) -> Any:
    method tool_extractor (line 667) | def tool_extractor(content: str) -> Union[str, list["FunctionCall"]]:
  function get_tool_utils (line 738) | def get_tool_utils(name: str) -> "ToolUtils":

FILE: src/llamafactory/eval/evaluator.py
  class Evaluator (line 61) | class Evaluator:
    method __init__ (line 62) | def __init__(self, args: Optional[dict[str, Any]] = None) -> None:
    method batch_inference (line 72) | def batch_inference(self, batch_input: dict[str, "torch.Tensor"]) -> l...
    method eval (line 79) | def eval(self) -> None:
    method _save_results (line 139) | def _save_results(self, category_corrects: dict[str, "NDArray"], resul...
  function run_eval (line 157) | def run_eval() -> None:

FILE: src/llamafactory/eval/template.py
  class EvalTemplate (line 22) | class EvalTemplate:
    method _parse_example (line 27) | def _parse_example(self, example: dict[str, str]) -> tuple[str, str]:
    method format_example (line 36) | def format_example(
  function _register_eval_template (line 56) | def _register_eval_template(name: str, system: str, choice: str, answer:...
  function get_eval_template (line 60) | def get_eval_template(name: str) -> "EvalTemplate":

FILE: src/llamafactory/extras/constants.py
  class AttentionFunction (line 116) | class AttentionFunction(StrEnum):
  class EngineName (line 124) | class EngineName(StrEnum):
  class DownloadSource (line 131) | class DownloadSource(StrEnum):
  class QuantizationMethod (line 138) | class QuantizationMethod(StrEnum):
  class RopeScaling (line 152) | class RopeScaling(StrEnum):
  function register_model_group (line 159) | def register_model_group(

FILE: src/llamafactory/extras/env.py
  function print_env (line 25) | def print_env() -> None:

FILE: src/llamafactory/extras/logging.py
  class LoggerHandler (line 34) | class LoggerHandler(logging.Handler):
    method __init__ (line 37) | def __init__(self, output_dir: str) -> None:
    method _write_log (line 52) | def _write_log(self, log_entry: str) -> None:
    method emit (line 56) | def emit(self, record) -> None:
    method close (line 63) | def close(self) -> None:
  class _Logger (line 68) | class _Logger(logging.Logger):
    method info_rank0 (line 71) | def info_rank0(self, *args, **kwargs) -> None:
    method warning_rank0 (line 74) | def warning_rank0(self, *args, **kwargs) -> None:
    method warning_rank0_once (line 77) | def warning_rank0_once(self, *args, **kwargs) -> None:
  function _get_default_logging_level (line 81) | def _get_default_logging_level() -> "logging._Level":
  function _get_library_name (line 93) | def _get_library_name() -> str:
  function _get_library_root_logger (line 97) | def _get_library_root_logger() -> "_Logger":
  function _configure_library_root_logger (line 101) | def _configure_library_root_logger() -> None:
  function get_logger (line 121) | def get_logger(name: str | None = None) -> "_Logger":
  function add_handler (line 130) | def add_handler(handler: "logging.Handler") -> None:
  function remove_handler (line 136) | def remove_handler(handler: logging.Handler) -> None:
  function info_rank0 (line 142) | def info_rank0(self: "logging.Logger", *args, **kwargs) -> None:
  function warning_rank0 (line 147) | def warning_rank0(self: "logging.Logger", *args, **kwargs) -> None:
  function warning_rank0_once (line 153) | def warning_rank0_once(self: "logging.Logger", *args, **kwargs) -> None:

FILE: src/llamafactory/extras/misc.py
  class AverageMeter (line 57) | class AverageMeter:
    method __init__ (line 60) | def __init__(self):
    method reset (line 63) | def reset(self):
    method update (line 69) | def update(self, val, n=1):
  function check_version (line 76) | def check_version(requirement: str, mandatory: bool = False) -> None:
  function check_dependencies (line 95) | def check_dependencies() -> None:
  function calculate_tps (line 104) | def calculate_tps(dataset: list[dict[str, Any]], metrics: dict[str, floa...
  function count_parameters (line 117) | def count_parameters(model: "torch.nn.Module") -> tuple[int, int]:
  function get_current_device (line 144) | def get_current_device() -> "torch.device":
  function get_device_name (line 160) | def get_device_name() -> str:
  function get_torch_device (line 176) | def get_torch_device():
  function get_device_count (line 187) | def get_device_count() -> int:
  function get_logits_processor (line 201) | def get_logits_processor() -> "LogitsProcessorList":
  function get_current_memory (line 208) | def get_current_memory() -> tuple[int, int]:
  function get_peak_memory (line 222) | def get_peak_memory() -> tuple[int, int]:
  function has_tokenized_data (line 236) | def has_tokenized_data(path: "os.PathLike") -> bool:
  function infer_optim_dtype (line 241) | def infer_optim_dtype(model_dtype: Optional["torch.dtype"]) -> "torch.dt...
  function is_accelerator_available (line 251) | def is_accelerator_available() -> bool:
  function is_env_enabled (line 258) | def is_env_enabled(env_var: str, default: str = "0") -> bool:
  function numpify (line 263) | def numpify(inputs: Union["NDArray", "torch.Tensor"]) -> "NDArray":
  function skip_check_imports (line 275) | def skip_check_imports() -> None:
  function torch_gc (line 281) | def torch_gc() -> None:
  function try_download_model_from_other_hub (line 294) | def try_download_model_from_other_hub(model_args: "ModelArguments") -> str:
  function use_modelscope (line 331) | def use_modelscope() -> bool:
  function use_openmind (line 335) | def use_openmind() -> bool:
  function use_ray (line 339) | def use_ray() -> bool:
  function use_kt (line 343) | def use_kt() -> bool:
  function find_available_port (line 347) | def find_available_port() -> int:
  function fix_proxy (line 356) | def fix_proxy(ipv6_enabled: bool = False) -> None:

FILE: src/llamafactory/extras/packages.py
  function _is_package_available (line 30) | def _is_package_available(name: str) -> bool:
  function _get_package_version (line 34) | def _get_package_version(name: str) -> "Version":
  function is_pyav_available (line 41) | def is_pyav_available():
  function is_librosa_available (line 45) | def is_librosa_available():
  function is_fastapi_available (line 49) | def is_fastapi_available():
  function is_galore_available (line 53) | def is_galore_available():
  function is_apollo_available (line 57) | def is_apollo_available():
  function is_jieba_available (line 61) | def is_jieba_available():
  function is_gradio_available (line 65) | def is_gradio_available():
  function is_matplotlib_available (line 69) | def is_matplotlib_available():
  function is_mcore_adapter_available (line 73) | def is_mcore_adapter_available():
  function is_pillow_available (line 77) | def is_pillow_available():
  function is_ray_available (line 81) | def is_ray_available():
  function is_kt_available (line 85) | def is_kt_available():
  function is_requests_available (line 89) | def is_requests_available():
  function is_rouge_available (line 93) | def is_rouge_available():
  function is_safetensors_available (line 97) | def is_safetensors_available():
  function is_sglang_available (line 101) | def is_sglang_available():
  function is_starlette_available (line 105) | def is_starlette_available():
  function is_transformers_version_greater_than (line 110) | def is_transformers_version_greater_than(content: str):
  function is_torch_version_greater_than (line 115) | def is_torch_version_greater_than(content: str):
  function is_uvicorn_available (line 119) | def is_uvicorn_available():
  function is_vllm_available (line 123) | def is_vllm_available():

FILE: src/llamafactory/extras/ploting.py
  function smooth (line 34) | def smooth(scalars: list[float]) -> list[float]:
  function gen_loss_plot (line 49) | def gen_loss_plot(trainer_log: list[dict[str, Any]]) -> "matplotlib.figu...
  function plot_loss (line 69) | def plot_loss(save_dictionary: str, keys: list[str] = ["loss"]) -> None:

FILE: src/llamafactory/hparams/data_args.py
  class DataArguments (line 23) | class DataArguments:
    method __post_init__ (line 143) | def __post_init__(self):
    method to_dict (line 187) | def to_dict(self) -> dict[str, Any]:

FILE: src/llamafactory/hparams/evaluation_args.py
  class EvaluationArguments (line 23) | class EvaluationArguments:
    method __post_init__ (line 58) | def __post_init__(self):

FILE: src/llamafactory/hparams/finetuning_args.py
  class FreezeArguments (line 20) | class FreezeArguments:
  class LoraArguments (line 56) | class LoraArguments:
  class OFTArguments (line 126) | class OFTArguments:
  class RLHFArguments (line 168) | class RLHFArguments:
  class GaloreArguments (line 263) | class GaloreArguments:
  class ApolloArguments (line 302) | class ApolloArguments:
  class BAdamArgument (line 353) | class BAdamArgument:
  class SwanLabArguments (line 404) | class SwanLabArguments:
  class FinetuningArguments (line 444) | class FinetuningArguments(
    method __post_init__ (line 542) | def __post_init__(self):
    method to_dict (line 599) | def to_dict(self) -> dict[str, Any]:

FILE: src/llamafactory/hparams/generating_args.py
  class GeneratingArguments (line 22) | class GeneratingArguments:
    method to_dict (line 70) | def to_dict(self, obey_generation_config: bool = False) -> dict[str, A...

FILE: src/llamafactory/hparams/model_args.py
  class BaseModelArguments (line 34) | class BaseModelArguments:
    method __post_init__ (line 205) | def __post_init__(self):
  class QuantizationArguments (line 275) | class QuantizationArguments:
  class ProcessorArguments (line 301) | class ProcessorArguments:
    method __post_init__ (line 345) | def __post_init__(self):
  class ExportArguments (line 354) | class ExportArguments:
    method __post_init__ (line 394) | def __post_init__(self):
  class VllmArguments (line 400) | class VllmArguments:
    method __post_init__ (line 424) | def __post_init__(self):
  class SGLangArguments (line 430) | class SGLangArguments:
    method __post_init__ (line 456) | def __post_init__(self):
  class KTransformersArguments (line 462) | class KTransformersArguments:
  class ModelArguments (line 507) | class ModelArguments(
    method __post_init__ (line 542) | def __post_init__(self):
    method copyfrom (line 550) | def copyfrom(cls, source: "Self", **kwargs) -> "Self":
    method to_dict (line 565) | def to_dict(self) -> dict[str, Any]:

FILE: src/llamafactory/hparams/parser.py
  function read_args (line 69) | def read_args(args: dict[str, Any] | list[str] | None = None) -> dict[st...
  function _parse_args (line 86) | def _parse_args(
  function _verify_trackio_args (line 103) | def _verify_trackio_args(training_args: "TrainingArguments") -> None:
  function _set_transformers_logging (line 149) | def _set_transformers_logging() -> None:
  function _set_env_vars (line 156) | def _set_env_vars() -> None:
  function _verify_model_args (line 164) | def _verify_model_args(
  function _check_extra_dependencies (line 189) | def _check_extra_dependencies(
  function _parse_train_args (line 241) | def _parse_train_args(args: dict[str, Any] | list[str] | None = None) ->...
  function _parse_train_mca_args (line 247) | def _parse_train_mca_args(args: dict[str, Any] | list[str] | None = None...
  function _configure_mca_training_args (line 259) | def _configure_mca_training_args(training_args, data_args, finetuning_ar...
  function _parse_infer_args (line 268) | def _parse_infer_args(args: dict[str, Any] | list[str] | None = None) ->...
  function _parse_eval_args (line 274) | def _parse_eval_args(args: dict[str, Any] | list[str] | None = None) -> ...
  function get_ray_args (line 280) | def get_ray_args(args: dict[str, Any] | list[str] | None = None) -> RayA...
  function get_train_args (line 286) | def get_train_args(args: dict[str, Any] | list[str] | None = None) -> _T...
  function get_infer_args (line 519) | def get_infer_args(args: dict[str, Any] | list[str] | None = None) -> _I...
  function get_eval_args (line 554) | def get_eval_args(args: dict[str, Any] | list[str] | None = None) -> _EV...

FILE: src/llamafactory/hparams/training_args.py
  class RayArguments (line 39) | class RayArguments:
    method __post_init__ (line 59) | def __post_init__(self):
  class Fp8Arguments (line 67) | class Fp8Arguments:
  class TrainingArguments (line 90) | class TrainingArguments(Fp8Arguments, RayArguments, BaseTrainingArguments):
    method __post_init__ (line 98) | def __post_init__(self):

FILE: src/llamafactory/launcher.py
  function launch (line 38) | def launch():

FILE: src/llamafactory/model/adapter.py
  function _setup_full_tuning (line 40) | def _setup_full_tuning(
  function _setup_freeze_tuning (line 59) | def _setup_freeze_tuning(
  function _setup_lora_tuning (line 143) | def _setup_lora_tuning(
  function init_adapter (line 321) | def init_adapter(

FILE: src/llamafactory/model/loader.py
  class TokenizerModule (line 52) | class TokenizerModule(TypedDict):
  function _get_init_kwargs (line 57) | def _get_init_kwargs(model_args: "ModelArguments") -> dict[str, Any]:
  function load_tokenizer (line 72) | def load_tokenizer(model_args: "ModelArguments") -> "TokenizerModule":
  function load_config (line 126) | def load_config(model_args: "ModelArguments") -> "PretrainedConfig":
  function load_model (line 132) | def load_model(

FILE: src/llamafactory/model/model_utils/attention.py
  function configure_attn_implementation (line 31) | def configure_attn_implementation(config: "PretrainedConfig", model_args...
  function print_attn_implementation (line 104) | def print_attn_implementation(config: "PretrainedConfig") -> None:

FILE: src/llamafactory/model/model_utils/checkpointing.py
  function get_unsloth_gradient_checkpointing_func (line 43) | def get_unsloth_gradient_checkpointing_func() -> Callable:
  function get_custom_gradient_checkpointing_func (line 80) | def get_custom_gradient_checkpointing_func(gradient_checkpointing_func: ...
  function _gradient_checkpointing_enable (line 106) | def _gradient_checkpointing_enable(
  function _fp32_forward_post_hook (line 137) | def _fp32_forward_post_hook(
  function prepare_model_for_training (line 143) | def prepare_model_for_training(model: "PreTrainedModel", model_args: "Mo...

FILE: src/llamafactory/model/model_utils/embedding.py
  function _noisy_mean_initialization (line 32) | def _noisy_mean_initialization(embed_weight: "torch.Tensor", num_new_tok...
  function _description_based_initialization (line 48) | def _description_based_initialization(
  function _initialize_embeddings (line 114) | def _initialize_embeddings(
  function resize_embedding_layer (line 153) | def resize_embedding_layer(

FILE: src/llamafactory/model/model_utils/ktransformers.py
  function _get_kt_kwargs (line 48) | def _get_kt_kwargs(
  function load_kt_pretrained_model (line 69) | def load_kt_pretrained_model(config: "PretrainedConfig", model_args: "Mo...
  function get_kt_peft_model (line 117) | def get_kt_peft_model(model: "PreTrainedModel", peft_kwargs: dict[str, A...
  function load_kt_peft_model (line 124) | def load_kt_peft_model(model_args: "ModelArguments", model: "PreTrainedM...

FILE: src/llamafactory/model/model_utils/kv_cache.py
  function configure_kv_cache (line 29) | def configure_kv_cache(config: "PretrainedConfig", model_args: "ModelArg...

FILE: src/llamafactory/model/model_utils/liger_kernel.py
  function apply_liger_kernel (line 30) | def apply_liger_kernel(

FILE: src/llamafactory/model/model_utils/longlora.py
  function llama_attention_forward (line 56) | def llama_attention_forward(
  function llama_flash_attention_2_forward (line 141) | def llama_flash_attention_2_forward(
  function llama_sdpa_attention_forward (line 249) | def llama_sdpa_attention_forward(
  function _apply_llama_patch (line 352) | def _apply_llama_patch() -> None:
  function configure_longlora (line 359) | def configure_longlora(config: "PretrainedConfig", model_args: "ModelArg...

FILE: src/llamafactory/model/model_utils/misc.py
  function find_all_linear_modules (line 28) | def find_all_linear_modules(model: "PreTrainedModel", freeze_vision_towe...
  function find_expanded_modules (line 55) | def find_expanded_modules(model: "PreTrainedModel", target_modules: list...
  function register_autoclass (line 80) | def register_autoclass(config: "PretrainedConfig", model: "PreTrainedMod...

FILE: src/llamafactory/model/model_utils/mod.py
  function load_mod_pretrained_model (line 26) | def load_mod_pretrained_model(**init_kwargs) -> "PreTrainedModel":
  function convert_pretrained_model_to_mod (line 32) | def convert_pretrained_model_to_mod(

FILE: src/llamafactory/model/model_utils/moe.py
  function _set_z3_leaf_modules (line 36) | def _set_z3_leaf_modules(model: "PreTrainedModel", leaf_modules: list[Un...
  function add_z3_leaf_module (line 43) | def add_z3_leaf_module(model: "PreTrainedModel") -> None:
  function configure_moe (line 151) | def configure_moe(config: "PretrainedConfig", model_args: "ModelArgument...
  class Qwen3OmniMoeThinkerTextSparseMoeBlock (line 202) | class Qwen3OmniMoeThinkerTextSparseMoeBlock(nn.Module):
    method __init__ (line 203) | def __init__(self, config):
    method forward (line 220) | def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:

FILE: src/llamafactory/model/model_utils/packing.py
  function get_seqlens_in_batch (line 55) | def get_seqlens_in_batch(attention_mask: "torch.Tensor") -> "torch.Tensor":
  function get_unpad_data (line 81) | def get_unpad_data(attention_mask: "torch.Tensor") -> tuple["torch.Tenso...
  function configure_packing (line 110) | def configure_packing(model_args: "ModelArguments", is_trainable: bool) ...

FILE: src/llamafactory/model/model_utils/quantization.py
  function _get_quantization_dataset (line 43) | def _get_quantization_dataset(tokenizer: "PreTrainedTokenizer", model_ar...
  function configure_quantization (line 82) | def configure_quantization(

FILE: src/llamafactory/model/model_utils/rope.py
  function configure_rope (line 35) | def configure_rope(config: "PretrainedConfig", model_args: "ModelArgumen...

FILE: src/llamafactory/model/model_utils/unsloth.py
  function _get_unsloth_kwargs (line 30) | def _get_unsloth_kwargs(
  function load_unsloth_pretrained_model (line 51) | def load_unsloth_pretrained_model(
  function get_unsloth_peft_model (line 68) | def get_unsloth_peft_model(
  function load_unsloth_peft_model (line 82) | def load_unsloth_peft_model(

FILE: src/llamafactory/model/model_utils/valuehead.py
  function load_valuehead_params (line 33) | def load_valuehead_params(path_or_repo_id: str, model_args: "ModelArgume...
  function prepare_valuehead_model (line 61) | def prepare_valuehead_model(model: "PreTrainedModel") -> None:

FILE: src/llamafactory/model/model_utils/visual.py
  class CompositeModel (line 41) | class CompositeModel:
    method get_projector (line 48) | def get_projector(self, module: "torch.nn.Module") -> "torch.nn.Module":
  function _register_composite_model (line 58) | def _register_composite_model(
  class LlavaMultiModalProjectorForYiVL (line 84) | class LlavaMultiModalProjectorForYiVL(torch.nn.Module):
    method __init__ (line 85) | def __init__(self, config: "LlavaConfig") -> None:
    method forward (line 98) | def forward(self, image_features: "torch.Tensor") -> "torch.Tensor":
  class LlavaMultiModalProjectorForYiVLForVLLM (line 118) | class LlavaMultiModalProjectorForYiVLForVLLM(LlavaMultiModalProjectorFor...
    method __init__ (line 119) | def __init__(self, vision_hidden_size: int, text_hidden_size: int, pro...
  function autocast_projector_dtype (line 129) | def autocast_projector_dtype(model: "PreTrainedModel", model_args: "Mode...
  function configure_visual_model (line 148) | def configure_visual_model(config: "PretrainedConfig") -> None:
  function get_forbidden_modules (line 159) | def get_forbidden_modules(config: "PretrainedConfig", finetuning_args: "...
  function patch_target_modules (line 182) | def patch_target_modules(

FILE: src/llamafactory/model/patcher.py
  function patch_qwen3_omni_moe_thinker_text_sparse_moe_block (line 53) | def patch_qwen3_omni_moe_thinker_text_sparse_moe_block():
  function patch_youtu_vl_model (line 64) | def patch_youtu_vl_model(model: "PreTrainedModel") -> None:
  function patch_tokenizer (line 84) | def patch_tokenizer(tokenizer: "PreTrainedTokenizer", model_args: "Model...
  function patch_processor (line 108) | def patch_processor(
  function patch_config (line 126) | def patch_config(
  function patch_model (line 195) | def patch_model(
  function patch_valuehead_model (line 246) | def patch_valuehead_model(model: "AutoModelForCausalLMWithValueHead") ->...

FILE: src/llamafactory/third_party/muon/muon.py
  function zeropower_via_newtonschulz5 (line 48) | def zeropower_via_newtonschulz5(G: "torch.Tensor", steps: int) -> "torch...
  class Muon (line 76) | class Muon(torch.optim.Optimizer):
    method __init__ (line 102) | def __init__(
    method adjust_lr_for_muon (line 137) | def adjust_lr_for_muon(self, lr: float, param_shape: list[int]) -> float:
    method step (line 145) | def step(self, closure=None):

FILE: src/llamafactory/train/callbacks.py
  function fix_valuehead_checkpoint (line 53) | def fix_valuehead_checkpoint(
  class FixValueHeadModelCallback (line 98) | class FixValueHeadModelCallback(TrainerCallback):
    method on_save (line 102) | def on_save(self, args: "TrainingArguments", state: "TrainerState", co...
  class SaveProcessorCallback (line 112) | class SaveProcessorCallback(TrainerCallback):
    method __init__ (line 115) | def __init__(self, processor: "ProcessorMixin") -> None:
    method on_save (line 119) | def on_save(self, args: "TrainingArguments", state: "TrainerState", co...
    method on_train_end (line 125) | def on_train_end(self, args: "TrainingArguments", state: "TrainerState...
  class PissaConvertCallback (line 130) | class PissaConvertCallback(TrainerCallback):
    method on_train_begin (line 134) | def on_train_begin(self, args: "TrainingArguments", state: "TrainerSta...
    method on_train_end (line 146) | def on_train_end(self, args: "TrainingArguments", state: "TrainerState...
  class LogCallback (line 172) | class LogCallback(TrainerCallback):
    method __init__ (line 175) | def __init__(self) -> None:
    method _set_abort (line 194) | def _set_abort(self, signum, frame) -> None:
    method _reset (line 197) | def _reset(self, max_steps: int = 0) -> None:
    method _timing (line 204) | def _timing(self, cur_steps: int) -> None:
    method _write_log (line 213) | def _write_log(self, output_dir: str, logs: dict[str, Any]) -> None:
    method _create_thread_pool (line 217) | def _create_thread_pool(self, output_dir: str) -> None:
    method _close_thread_pool (line 221) | def _close_thread_pool(self) -> None:
    method on_init_end (line 227) | def on_init_end(self, args: "TrainingArguments", state: "TrainerState"...
    method on_train_begin (line 237) | def on_train_begin(self, args: "TrainingArguments", state: "TrainerSta...
    method on_train_end (line 244) | def on_train_end(self, args: "TrainingArguments", state: "TrainerState...
    method on_substep_end (line 248) | def on_substep_end(self, args: "TrainingArguments", state: "TrainerSta...
    method on_step_end (line 254) | def on_step_end(self, args: "TrainingArguments", state: "TrainerState"...
    method on_evaluate (line 260) | def on_evaluate(self, args: "TrainingArguments", state: "TrainerState"...
    method on_predict (line 265) | def on_predict(self, args: "TrainingArguments", state: "TrainerState",...
    method on_log (line 270) | def on_log(self, args: "TrainingArguments", state: "TrainerState", con...
    method on_prediction_step (line 311) | def on_prediction_step(
  class ReporterCallback (line 341) | class ReporterCallback(TrainerCallback):
    method __init__ (line 344) | def __init__(
    method on_train_begin (line 358) | def on_train_begin(self, args: "TrainingArguments", state: "TrainerSta...

FILE: src/llamafactory/train/dpo/ktrainer.py
  class KDPOTrainer (line 32) | class KDPOTrainer(KTrainer, CustomDPOTrainer):
    method concatenated_forward (line 34) | def concatenated_forward(

FILE: src/llamafactory/train/dpo/trainer.py
  class CustomDPOTrainer (line 44) | class CustomDPOTrainer(DPOTrainer):
    method __init__ (line 45) | def __init__(
    method create_optimizer (line 126) | def create_optimizer(self) -> "torch.optim.Optimizer":
    method create_scheduler (line 132) | def create_scheduler(
    method _get_train_sampler (line 139) | def _get_train_sampler(self, *args, **kwargs) -> Optional["torch.utils...
    method get_batch_samples (line 146) | def get_batch_samples(self, *args, **kwargs):
    method odds_ratio_loss (line 150) | def odds_ratio_loss(self, chosen_logps: "torch.Tensor", rejected_logps...
    method simpo_loss (line 160) | def simpo_loss(self, chosen_logps: "torch.Tensor", rejected_logps: "to...
    method bco_loss (line 168) | def bco_loss(
    method compute_preference_loss (line 187) | def compute_preference_loss(
    method concatenated_forward (line 219) | def concatenated_forward(
    method compute_reference_log_probs (line 255) | def compute_reference_log_probs(
    method get_batch_loss_metrics (line 277) | def get_batch_loss_metrics(
    method compute_loss (line 320) | def compute_loss(
    method log (line 327) | def log(self, logs: dict[str, float], *args, **kwargs) -> None:

FILE: src/llamafactory/train/dpo/workflow.py
  function run_dpo (line 35) | def run_dpo(

FILE: src/llamafactory/train/fp8_utils.py
  function create_fp8_kwargs (line 29) | def create_fp8_kwargs(training_args: "TrainingArguments") -> list[Any]:
  function get_fp8_mixed_precision (line 108) | def get_fp8_mixed_precision(training_args: "TrainingArguments") -> Optio...
  function configure_fp8_environment (line 120) | def configure_fp8_environment(training_args: "TrainingArguments") -> None:
  function verify_fp8_status (line 155) | def verify_fp8_status(accelerator, training_args: "TrainingArguments") -...
  function patch_accelerator_for_fp8 (line 185) | def patch_accelerator_for_fp8() -> None:

FILE: src/llamafactory/train/kto/trainer.py
  class CustomKTOTrainer (line 43) | class CustomKTOTrainer(KTOTrainer):
    method __init__ (line 44) | def __init__(
    method create_optimizer (line 123) | def create_optimizer(self) -> "torch.optim.Optimizer":
    method create_scheduler (line 129) | def create_scheduler(
    method _get_train_sampler (line 136) | def _get_train_sampler(self, *args, **kwargs) -> Optional["torch.utils...
    method get_batch_samples (line 144) | def get_batch_samples(self, *args, **kwargs):
    method forward (line 149) | def forward(
    method concatenated_forward (line 184) | def concatenated_forward(
    method compute_reference_log_probs (line 202) | def compute_reference_log_probs(
    method get_batch_loss_metrics (line 221) | def get_batch_loss_metrics(
    method compute_loss (line 271) | def compute_loss(
    method log (line 278) | def log(self, logs: dict[str, float], *args, **kwargs) -> None:

FILE: src/llamafactory/train/kto/workflow.py
  function run_kto (line 35) | def run_kto(

FILE: src/llamafactory/train/mca/workflow.py
  function _data_collator_wrapper (line 60) | def _data_collator_wrapper(data_collator: Any):
  function _check_model_support (line 80) | def _check_model_support(model_args: "ModelArguments"):
  function _freeze_model_parameters (line 98) | def _freeze_model_parameters(model: Any, finetuning_args: "FinetuningArg...
  function run_pt (line 121) | def run_pt(
  function run_sft (line 173) | def run_sft(
  function run_dpo (line 237) | def run_dpo(

FILE: src/llamafactory/train/ppo/ppo_utils.py
  function get_rewards_from_server (line 34) | def get_rewards_from_server(server_url: str, messages: list[str]) -> lis...
  function replace_model (line 43) | def replace_model(model: "AutoModelForCausalLMWithValueHead", target: Li...
  function dump_layernorm (line 65) | def dump_layernorm(model: "PreTrainedModel") -> dict[str, "torch.Tensor"]:
  function restore_layernorm (line 76) | def restore_layernorm(model: "PreTrainedModel", layernorm_params: Option...

FILE: src/llamafactory/train/ppo/trainer.py
  class CustomPPOTrainer (line 64) | class CustomPPOTrainer(PPOTrainer, Trainer):
    method __init__ (line 67) | def __init__(
    method ppo_train (line 200) | def ppo_train(self, resume_from_checkpoint: Optional[str] = None) -> N...
    method create_optimizer (line 311) | def create_optimizer(
    method create_scheduler (line 338) | def create_scheduler(
    method get_inputs (line 351) | def get_inputs(self, batch: dict[str, "torch.Tensor"]) -> tuple[list["...
    method get_rewards (line 389) | def get_rewards(
    method batched_forward_pass (line 422) | def batched_forward_pass(
    method save_model (line 492) | def save_model(self, output_dir: Optional[str] = None) -> None:

FILE: src/llamafactory/train/ppo/workflow.py
  function run_ppo (line 34) | def run_ppo(

FILE: src/llamafactory/train/pt/trainer.py
  class CustomTrainer (line 33) | class CustomTrainer(Trainer):
    method __init__ (line 36) | def __init__(
    method create_optimizer (line 72) | def create_optimizer(self) -> "torch.optim.Optimizer":
    method create_scheduler (line 78) | def create_scheduler(
    method _get_train_sampler (line 85) | def _get_train_sampler(self, *args, **kwargs) -> Optional["torch.utils...
    method compute_loss (line 92) | def compute_loss(self, model, inputs, *args, **kwargs):

FILE: src/llamafactory/train/pt/workflow.py
  function run_pt (line 36) | def run_pt(

FILE: src/llamafactory/train/rm/metric.py
  class ComputeAccuracy (line 28) | class ComputeAccuracy:
    method _dump (line 31) | def _dump(self) -> dict[str, float] | None:
    method __post_init__ (line 39) | def __post_init__(self):
    method __call__ (line 42) | def __call__(self, eval_preds: "EvalPrediction", compute_result: bool ...

FILE: src/llamafactory/train/rm/trainer.py
  class PairwiseTrainer (line 43) | class PairwiseTrainer(Trainer):
    method __init__ (line 46) | def __init__(
    method create_optimizer (line 68) | def create_optimizer(self) -> "torch.optim.Optimizer":
    method create_scheduler (line 74) | def create_scheduler(
    method _get_train_sampler (line 81) | def _get_train_sampler(self, *args, **kwargs) -> Optional["torch.utils...
    method compute_loss (line 88) | def compute_loss(
    method _save (line 113) | def _save(self, output_dir: Optional[str] = None, state_dict=None):
    method save_predictions (line 133) | def save_predictions(self, predict_results: "PredictionOutput") -> None:

FILE: src/llamafactory/train/rm/workflow.py
  function run_rm (line 35) | def run_rm(

FILE: src/llamafactory/train/sft/metric.py
  function eval_logit_processor (line 47) | def eval_logit_processor(logits: "torch.Tensor", labels: "torch.Tensor")...
  class ComputeAccuracy (line 62) | class ComputeAccuracy:
    method _dump (line 65) | def _dump(self) -> Optional[dict[str, float]]:
    method __post_init__ (line 73) | def __post_init__(self):
    method __call__ (line 76) | def __call__(self, eval_preds: "EvalPrediction", compute_result: bool ...
  class ComputeSimilarity (line 88) | class ComputeSimilarity:
    method _dump (line 96) | def _dump(self) -> Optional[dict[str, float]]:
    method __post_init__ (line 104) | def __post_init__(self):
    method __call__ (line 107) | def __call__(self, eval_preds: "EvalPrediction", compute_result: bool ...

FILE: src/llamafactory/train/sft/trainer.py
  class CustomSeq2SeqTrainer (line 47) | class CustomSeq2SeqTrainer(Seq2SeqTrainer):
    method __init__ (line 50) | def __init__(
    method create_optimizer (line 131) | def create_optimizer(self) -> "torch.optim.Optimizer":
    method create_scheduler (line 137) | def create_scheduler(
    method _get_train_sampler (line 144) | def _get_train_sampler(self, *args, **kwargs) -> Optional["torch.utils...
    method compute_loss (line 151) | def compute_loss(self, model, inputs, *args, **kwargs):
    method prediction_step (line 165) | def prediction_step(
    method save_predictions (line 191) | def save_predictions(

FILE: src/llamafactory/train/sft/workflow.py
  function run_sft (line 41) | def run_sft(

FILE: src/llamafactory/train/test_utils.py
  function compare_model (line 34) | def compare_model(model_a: "torch.nn.Module", model_b: "torch.nn.Module"...
  function check_lora_model (line 45) | def check_lora_model(model: "LoraModel") -> tuple[set[str], set[str]]:
  function load_train_model (line 63) | def load_train_model(add_valuehead: bool = False, **kwargs) -> "PreTrain...
  function load_infer_model (line 69) | def load_infer_model(add_valuehead: bool = False, **kwargs) -> "PreTrain...
  function load_reference_model (line 75) | def load_reference_model(
  function load_dataset_module (line 101) | def load_dataset_module(**kwargs) -> "DatasetModule":
  function patch_valuehead_model (line 109) | def patch_valuehead_model() -> None:

FILE: src/llamafactory/train/trainer_utils.py
  class DummyOptimizer (line 68) | class DummyOptimizer(torch.optim.Optimizer):
    method __init__ (line 71) | def __init__(
    method zero_grad (line 79) | def zero_grad(self, set_to_none: bool = True) -> None:
    method step (line 83) | def step(self, closure: Optional[Callable[[], float]] = None) -> Optio...
  function create_modelcard_and_push (line 87) | def create_modelcard_and_push(
  function create_ref_model (line 116) | def create_ref_model(
  function create_reward_model (line 151) | def create_reward_model(
  function _get_decay_parameter_names (line 193) | def _get_decay_parameter_names(model: "PreTrainedModel") -> list[str]:
  function _create_galore_optimizer (line 200) | def _create_galore_optimizer(
  function _create_apollo_optimizer (line 288) | def _create_apollo_optimizer(
  function _create_loraplus_optimizer (line 372) | def _create_loraplus_optimizer(
  function _create_badam_optimizer (line 412) | def _create_badam_optimizer(
  function _create_adam_mini_optimizer (line 473) | def _create_adam_mini_optimizer(
  function _create_muon_optimizer (line 498) | def _create_muon_optimizer(
  function create_custom_optimizer (line 527) | def create_custom_optimizer(
  function create_custom_scheduler (line 551) | def create_custom_scheduler(
  function get_batch_logps (line 592) | def get_batch_logps(
  function dft_loss_func (line 639) | def dft_loss_func(
  function _dft_cross_entropy (line 658) | def _dft_cross_entropy(
  function asft_loss_func (line 686) | def asft_loss_func(
  function _asft_cross_entropy (line 720) | def _asft_cross_entropy(
  function _kl_divergence (line 743) | def _kl_divergence(
  function eaft_loss_func (line 768) | def eaft_loss_func(
  function _eaft_cross_entropy (line 790) | def _eaft_cross_entropy(
  function nested_detach (line 829) | def nested_detach(
  function get_swanlab_callback (line 848) | def get_swanlab_callback(finetuning_args: "FinetuningArguments") -> "Tra...
  function get_placement_group (line 894) | def get_placement_group(num_workers: int) -> tuple["PlacementGroup", dic...
  function get_ray_remote_config_for_worker (line 906) | def get_ray_remote_config_for_worker(
  function get_ray_head_node_ip (line 943) | def get_ray_head_node_ip() -> str:
  function sort_placement_group_by_node_ip (line 949) | def sort_placement_group_by_node_ip(placement_group: "PlacementGroup", m...

FILE: src/llamafactory/train/tuner.py
  function _training_function (line 57) | def _training_function(config: dict[str, Any]) -> None:
  function run_exp (line 115) | def run_exp(args: Optional[dict[str, Any]] = None, callbacks: Optional[l...
  function export_model (line 128) | def export_model(args: Optional[dict[str, Any]] = None) -> None:
  class Worker (line 227) | class Worker:
    method __init__ (line 228) | def __init__(self):
    method _setup_env_visible_devices (line 234) | def _setup_env_visible_devices(self) -> None:
    method _training_function (line 247) | def _training_function(self, config: dict[str, Any]) -> None:
  function _ray_training_function (line 251) | def _ray_training_function(ray_args: "RayArguments", config: dict[str, A...

FILE: src/llamafactory/v1/accelerator/helper.py
  class DeviceType (line 42) | class DeviceType(StrEnum):
  class ReduceOp (line 52) | class ReduceOp(StrEnum):
  function requires_accelerator (line 59) | def requires_accelerator(fn):
  function is_distributed (line 75) | def is_distributed() -> bool:
  function get_rank (line 80) | def get_rank() -> int:
  function get_world_size (line 85) | def get_world_size() -> int:
  function get_local_rank (line 90) | def get_local_rank() -> int:
  function get_local_world_size (line 95) | def get_local_world_size() -> int:
  function get_current_accelerator (line 102) | def get_current_accelerator(check_available: bool = True) -> torch.device:
  function get_device_count (line 110) | def get_device_count() -> int:
  function synchronize (line 116) | def synchronize() -> None:
  function set_device_index (line 122) | def set_device_index() -> None:
  function get_current_device (line 129) | def get_current_device() -> torch.device:
  function is_torch_cuda_available (line 137) | def is_torch_cuda_available():
  function is_torch_mps_available (line 142) | def is_torch_mps_available():
  function is_torch_npu_available (line 147) | def is_torch_npu_available():
  function is_torch_xpu_available (line 152) | def is_torch_xpu_available():
  function operate_tensorlike (line 157) | def operate_tensorlike(fn: Callable[[...], Tensor], data: TensorLike, **...
  function get_process_group_backend (line 183) | def get_process_group_backend() -> str:
  function all_gather (line 193) | def all_gather(tensor: Tensor, group: Optional[ProcessGroup] = None) -> ...
  function all_reduce (line 201) | def all_reduce(tensor: Tensor, op: ReduceOp = ReduceOp.MEAN, group: Opti...
  function broadcast (line 216) | def broadcast(tensor: Tensor, src: int = 0, group: Optional[ProcessGroup...
  function main_process_first (line 223) | def main_process_first(local_only: bool = True) -> None:

FILE: src/llamafactory/v1/accelerator/interface.py
  class Dim (line 45) | class Dim(StrEnum):
  class DistributedStrategy (line 55) | class DistributedStrategy:
    method __post_init__ (line 67) | def __post_init__(self) -> None:
    method model_mesh_shape (line 89) | def model_mesh_shape(self) -> tuple[int, int]:
    method model_mesh_dim_names (line 94) | def model_mesh_dim_names(self) -> tuple[str, str]:
    method data_mesh_shape (line 99) | def data_mesh_shape(self) -> tuple[int, int]:
    method data_mesh_dim_names (line 104) | def data_mesh_dim_names(self) -> tuple[str, str]:
  class DistributedInterface (line 109) | class DistributedInterface:
    method __new__ (line 115) | def __new__(cls, *args: Any, **kwargs: Any) -> "DistributedInterface":
    method __init__ (line 122) | def __init__(self, config: DistributedConfig | None = None) -> None:
    method __str__ (line 166) | def __str__(self) -> str:
    method get_device_mesh (line 173) | def get_device_mesh(self, dim: Dim | None = None) -> DeviceMesh | None:
    method get_group (line 184) | def get_group(self, dim: Dim | None = None) -> Optional[ProcessGroup]:
    method get_rank (line 191) | def get_rank(self, dim: Dim | None = None) -> int:
    method get_world_size (line 200) | def get_world_size(self, dim: Dim | None = None) -> int:
    method get_local_rank (line 209) | def get_local_rank(self) -> int:
    method get_local_world_size (line 213) | def get_local_world_size(self) -> int:
    method all_gather (line 217) | def all_gather(self, data: TensorLike, dim: Dim | None = Dim.DP) -> Te...
    method all_reduce (line 224) | def all_reduce(
    method broadcast (line 233) | def broadcast(self, data: TensorLike, src: int = 0, dim: Dim | None = ...
    method sync (line 240) | def sync(self) -> None:
    method barrier (line 245) | def barrier(self) -> None:
    method destroy (line 250) | def destroy(self) -> None:

FILE: src/llamafactory/v1/config/arg_parser.py
  function get_args (line 34) | def get_args(args: InputArgument = None) -> tuple[ModelArguments, DataAr...

FILE: src/llamafactory/v1/config/arg_utils.py
  class PluginConfig (line 23) | class PluginConfig(dict):
    method name (line 27) | def name(self) -> str:
  class ModelClass (line 39) | class ModelClass(StrEnum):
  class SampleBackend (line 48) | class SampleBackend(StrEnum):
  class BatchingStrategy (line 54) | class BatchingStrategy(StrEnum):
  function _convert_str_dict (line 61) | def _convert_str_dict(data: dict) -> dict:
  function get_plugin_config (line 84) | def get_plugin_config(config: PluginArgument) -> PluginConfig | None:

FILE: src/llamafactory/v1/config/data_args.py
  class DataArguments (line 20) | class DataArguments:

FILE: src/llamafactory/v1/config/model_args.py
  class ModelArguments (line 22) | class ModelArguments:
    method __post_init__ (line 56) | def __post_init__(self) -> None:

FILE: src/llamafactory/v1/config/sample_args.py
  class SampleArguments (line 22) | class SampleArguments:

FILE: src/llamafactory/v1/config/training_args.py
  class TrainingArguments (line 23) | class TrainingArguments:
    method __post_init__ (line 89) | def __post_init__(self) -> None:

FILE: src/llamafactory/v1/core/base_sampler.py
  class BaseSampler (line 23) | class BaseSampler:
    method __init__ (line 33) | def __init__(
    method generate (line 45) | async def generate(self, messages: list[Message], tools: str | None = ...
    method batch_infer (line 58) | async def batch_infer(self, dataset: TorchDataset) -> list[Sample]:

FILE: src/llamafactory/v1/core/base_trainer.py
  class BaseTrainer (line 48) | class BaseTrainer:
    method __init__ (line 49) | def __init__(
    method _create_batch_generator (line 102) | def _create_batch_generator(self) -> None:
    method _shard_model (line 114) | def _shard_model(self) -> None:
    method _init_optimizer (line 132) | def _init_optimizer(self) -> None:
    method _init_lr_scheduler (line 142) | def _init_lr_scheduler(self) -> None:
    method compute_log_probs (line 153) | def compute_log_probs(self, model: HFModel, batch: BatchInput) -> Tensor:
    method compute_loss (line 170) | def compute_loss(self, batch: BatchInput) -> Tensor:
    method fit (line 174) | def fit(self) -> None:
    method save_model (line 224) | def save_model(self) -> None:

FILE: src/llamafactory/v1/core/data_engine.py
  class DataEngine (line 47) | class DataEngine(Dataset):
    method __init__ (line 54) | def __init__(self, dataset_path: str) -> None:
    method _get_dataset_info (line 69) | def _get_dataset_info(self) -> None:
    method _load_dataset (line 82) | def _load_dataset(self) -> None:
    method _build_data_index (line 100) | def _build_data_index(self) -> None:
    method _convert_data_sample (line 117) | def _convert_data_sample(self, raw_sample: dict[str, Any], dataset_nam...
    method __len__ (line 135) | def __len__(self) -> int:
    method __getitem__ (line 146) | def __getitem__(self, index: int | Any) -> Sample | list[Sample]:
    method __iter__ (line 174) | def __iter__(self) -> Iterable[Sample]:

FILE: src/llamafactory/v1/core/model_engine.py
  class ModelEngine (line 47) | class ModelEngine:
    method __init__ (line 55) | def __init__(self, model_args: ModelArguments, is_train: bool = False)...
    method _init_processor (line 69) | def _init_processor(self) -> Processor:
    method _init_model_config (line 80) | def _init_model_config(self) -> HFConfig:
    method _init_model (line 87) | def _init_model(self) -> HFModel:

FILE: src/llamafactory/v1/core/utils/batching.py
  function default_collate_fn (line 46) | def default_collate_fn(buffer: StatefulBuffer, batch_info: BatchInfo) ->...
  class BatchGenerator (line 63) | class BatchGenerator(Iterator):
    method __init__ (line 64) | def __init__(
    method _init_data_provider (line 127) | def _init_data_provider(self) -> None:
    method __len__ (line 162) | def __len__(self) -> int:
    method __iter__ (line 165) | def __iter__(self):
    method __next__ (line 174) | def __next__(self):
    method _fill_buffer (line 182) | def _fill_buffer(self) -> None:
    method _generate_batch (line 196) | def _generate_batch(self) -> list[BatchInput] | None:
    method state_dict (line 204) | def state_dict(self) -> dict[str, Any]:
    method load_state_dict (line 211) | def load_state_dict(self, state: dict[str, Any]) -> None:
    method set_epoch (line 217) | def set_epoch(self, epoch: int) -> None:

FILE: src/llamafactory/v1/core/utils/inference_engine.py
  class BaseEngine (line 31) | class BaseEngine(ABC):
    method __init__ (line 33) | def __init__(
    method generate (line 51) | async def generate(self, messages: list[Message], tools: str | None = ...
    method batch_infer (line 64) | async def batch_infer(self, dataset: TorchDataset) -> list[Sample]:
  class HuggingFaceEngine (line 76) | class HuggingFaceEngine(BaseEngine):
    method __init__ (line 77) | def __init__(
    method generate (line 91) | async def generate(self, messages: list[Message], tools: str | None = ...
    method batch_infer (line 112) | async def batch_infer(self, dataset: TorchDataset) -> list[Sample]:

FILE: src/llamafactory/v1/core/utils/rendering.py
  function render_chatml_messages (line 31) | def render_chatml_messages(
  function parse_chatml_message (line 76) | def parse_chatml_message(generated_text: str) -> Message:
  class Renderer (line 88) | class Renderer:
    method __init__ (line 89) | def __init__(self, template: str, processor: Processor):
    method render_messages (line 93) | def render_messages(
    method parse_message (line 120) | def parse_message(self, generated_text: str) -> Message:
    method process_samples (line 136) | def process_samples(self, samples: list[Sample]) -> list[ModelInput]:

FILE: src/llamafactory/v1/launcher.py
  function launch (line 34) | def launch():
  function main (line 157) | def main():

FILE: src/llamafactory/v1/plugins/data_plugins/converter.py
  class AlpacaSample (line 27) | class AlpacaSample(TypedDict, total=False):
  class SharegptSample (line 40) | class SharegptSample(TypedDict, total=False):
  class OpenaiMessage (line 45) | class OpenaiMessage(TypedDict, total=False):
  class OpenaiSample (line 50) | class OpenaiSample(TypedDict, total=False):
  class PairSample (line 54) | class PairSample(TypedDict, total=False):
  class DataConverterPlugin (line 59) | class DataConverterPlugin(BasePlugin):
    method __call__ (line 62) | def __call__(self, raw_sample: dict[str, Any]) -> Sample:
  function alpaca_converter (line 67) | def alpaca_converter(raw_sample: AlpacaSample) -> SFTSample:
  function sharegpt_converter (line 104) | def sharegpt_converter(raw_sample: SharegptSample) -> SFTSample:
  function pair_converter (line 168) | def pair_converter(raw_sample: PairSample) -> DPOSample:

FILE: src/llamafactory/v1/plugins/data_plugins/loader.py
  class DataLoaderPlugin (line 26) | class DataLoaderPlugin(BasePlugin):
    method load (line 29) | def load(self, dataset_info: DatasetInfo) -> HFDataset:
  function _get_builder_name (line 36) | def _get_builder_name(path: str) -> Literal["arrow", "csv", "json", "par...
  function load_data_from_file (line 53) | def load_data_from_file(filepath: str, split: str, streaming: bool) -> H...
  function adjust_data_index (line 69) | def adjust_data_index(
  function select_data_sample (line 91) | def select_data_sample(

FILE: src/llamafactory/v1/plugins/model_plugins/initialization.py
  class InitPlugin (line 23) | class InitPlugin(BasePlugin):
    method __call__ (line 24) | def __call__(self) -> torch.device:
  function init_on_meta (line 29) | def init_on_meta() -> torch.device:
  function init_on_rank0 (line 34) | def init_on_rank0() -> torch.device:
  function init_on_default (line 42) | def init_on_default() -> torch.device:

FILE: src/llamafactory/v1/plugins/model_plugins/kernels/base.py
  class BaseKernel (line 30) | class BaseKernel(ABC):
    method get_kernel_id (line 40) | def get_kernel_id(cls) -> str:
    method get_device (line 45) | def get_device(cls) -> str:
    method check_deps (line 50) | def check_deps(cls) -> bool:
    method apply (line 67) | def apply(cls, **kwargs) -> HFModel:

FILE: src/llamafactory/v1/plugins/model_plugins/kernels/interface.py
  function scan_all_kernels (line 36) | def scan_all_kernels():
  function get_default_kernels (line 80) | def get_default_kernels():
  function apply_kernel (line 89) | def apply_kernel(kernel_id: str, **kwargs):
  class KernelPlugin (line 107) | class KernelPlugin(BasePlugin):
  function apply_default_kernels (line 114) | def apply_default_kernels(model: HFModel, include_kernels: str = None) -...

FILE: src/llamafactory/v1/plugins/model_plugins/kernels/ops/mlp/npu_fused_moe.py
  class GmmFunction (line 42) | class GmmFunction(torch.autograd.Function):
    method forward (line 46) | def forward(ctx, x, weight, group_list):
    method backward (line 67) | def backward(ctx, grad_output):
  class HybridGmmFunction (line 96) | class HybridGmmFunction(torch.autograd.Function):
    method forward (line 100) | def forward(ctx, num_experts, *args):
    method backward (line 126) | def backward(ctx, *grad_outputs):
  class NpuMoeFused (line 178) | class NpuMoeFused:
    method npu_moe_experts_forward (line 182) | def npu_moe_experts_forward(
    method npu_moe_sparse_block_forward (line 210) | def npu_moe_sparse_block_forward(self, hidden_states: torch.Tensor) ->...
  class Qwen3NpuMoeFused (line 232) | class Qwen3NpuMoeFused:
    method qwen3moe_sparse_moe_block_forward (line 236) | def qwen3moe_sparse_moe_block_forward(self, hidden_states: torch.Tensor):
  class NpuFusedMoEKernel (line 300) | class NpuFusedMoEKernel(BaseKernel):
    method apply (line 307) | def apply(cls, **kwargs) -> HFModel:

FILE: src/llamafactory/v1/plugins/model_plugins/kernels/ops/mlp/npu_swiglu.py
  function npu_swiglu_forward (line 40) | def npu_swiglu_forward(self, hidden_state):
  function _npu_swiglu_glm4_forward (line 55) | def _npu_swiglu_glm4_forward(self, hidden_states):
  function _npu_swiglu_gemma3ntext_forward (line 70) | def _npu_swiglu_gemma3ntext_forward(self, hidden_states):
  class NpuSwiGluKernel (line 90) | class NpuSwiGluKernel(BaseKernel):
    method apply (line 128) | def apply(cls, **kwargs) -> "HFModel":

FILE: src/llamafactory/v1/plugins/model_plugins/kernels/ops/rms_norm/npu_rms_norm.py
  function npu_rms_norm_forward (line 32) | def npu_rms_norm_forward(self, hidden_states):
  class NpuRMSNormKernel (line 48) | class NpuRMSNormKernel(BaseKernel):
    method apply (line 55) | def apply(cls, **kwargs) -> "HFModel":

FILE: src/llamafactory/v1/plugins/model_plugins/kernels/ops/rope/npu_rope.py
  function _apply_rotary_pos_emb (line 42) | def _apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_d...
  function _apply_multimodal_rotary_pos_emb_qwen25_vl (line 63) | def _apply_multimodal_rotary_pos_emb_qwen25_vl(q, k, cos, sin, mrope_sec...
  class NpuRoPEKernel (line 91) | class NpuRoPEKernel(BaseKernel):
    method apply (line 98) | def apply(cls, **kwargs) -> "HFModel":

FILE: src/llamafactory/v1/plugins/model_plugins/kernels/registry.py
  class Registry (line 30) | class Registry:
    method register (line 39) | def register(cls, kernel_cls: type[BaseKernel]) -> type[BaseKernel] | ...
    method get (line 74) | def get(cls, kernel_id: str) -> type[BaseKernel] | None:
    method get_registered_kernels (line 86) | def get_registered_kernels(cls) -> dict[str, type[BaseKernel]]:

FILE: src/llamafactory/v1/plugins/model_plugins/peft.py
  class LoraConfigDict (line 31) | class LoraConfigDict(TypedDict, total=False):
  class FreezeConfigDict (line 62) | class FreezeConfigDict(TypedDict, total=False):
  class PeftPlugin (line 75) | class PeftPlugin(BasePlugin):
    method __call__ (line 76) | def __call__(self, model: HFModel, config: dict, is_train: bool) -> HF...
  function _find_all_linear_modules (line 80) | def _find_all_linear_modules(model: HFModel) -> list[str]:
  function merge_adapters (line 94) | def merge_adapters(model: HFModel, adapter_name_or_path: Union[list[str]...
  function load_adapter (line 106) | def load_adapter(model: HFModel, adapter_name_or_path: Union[list[str], ...
  function get_lora_model (line 152) | def get_lora_model(model: HFModel, config: LoraConfigDict, is_train: boo...
  function get_freeze_model (line 194) | def get_freeze_model(model: HFModel, config: FreezeConfigDict, is_train:...
  function merge_and_export_model (line 287) | def merge_and_export_model(args: InputArgument = None):

FILE: src/llamafactory/v1/plugins/model_plugins/quantization.py
  class QuantizationPlugin (line 36) | class QuantizationPlugin(BasePlugin):
    method __call__ (line 39) | def __call__(
  function quantization_auto (line 53) | def quantization_auto(
  function quantization_with_bnb (line 81) | def quantization_with_bnb(

FILE: src/llamafactory/v1/plugins/model_plugins/rendering.py
  class RenderingPlugin (line 25) | class RenderingPlugin(BasePlugin):
    method _ensure_template_imported (line 28) | def _ensure_template_imported(self) -> None:
    method __getitem__ (line 39) | def __getitem__(self, method_name: str):
    method render_messages (line 43) | def render_messages(
    method parse_messages (line 54) | def parse_messages(self, generated_text: str) -> Message:

FILE: src/llamafactory/v1/plugins/model_plugins/templates/qwen3.py
  function _update_model_input (line 24) | def _update_model_input(
  function _concat_text_content (line 48) | def _concat_text_content(message: Message) -> str:
  function _get_last_query_index (line 60) | def _get_last_query_index(messages: list[Message]) -> int:
  function _split_assistant_content (line 86) | def _split_assistant_content(message: Message) -> tuple[str, str, list[T...
  function render_qwen3_messages (line 111) | def render_qwen3_messages(
  function parse_qwen3_message (line 220) | def parse_qwen3_message(generated_text: str) -> Message:

FILE: src/llamafactory/v1/plugins/model_plugins/templates/qwen3_nothink.py
  function _update_model_input (line 24) | def _update_model_input(
  function _concat_text_content (line 48) | def _concat_text_content(message: Message) -> str:
  function render_qwen3_nothink_messages (line 61) | def render_qwen3_nothink_messages(
  function parse_qwen3_nothink_message (line 170) | def parse_qwen3_nothink_message(generated_text: str) -> Message:

FILE: src/llamafactory/v1/plugins/trainer_plugins/batching.py
  class BatchingPlugin (line 20) | class BatchingPlugin(BasePlugin):
    method compute_length (line 21) | def compute_length(self, data_provider: DataLoader) -> int:
    method fill_buffer (line 28) | def fill_buffer(self, buffer: StatefulBuffer, batch_info: BatchInfo) -...
    method generate_batch (line 32) | def generate_batch(self, buffer: StatefulBuffer, batch_info: BatchInfo...

FILE: src/llamafactory/v1/plugins/trainer_plugins/distributed/deepspeed.py
  class DeepSpeedEngine (line 35) | class DeepSpeedEngine:
    method __init__ (line 47) | def __init__(self, dist_config: dict[str, Any], num_micro_batch: int =...
    method shard_model (line 67) | def shard_model(self, model: HFModel) -> "DeepSpeedEngine":
    method prepare (line 74) | def prepare(
    method backward (line 94) | def backward(self, loss: torch.Tensor) -> None:
    method get_grad_norm (line 104) | def get_grad_norm(self) -> float:
  function save_model (line 112) | def save_model(model: HFModel, output_dir: str, processor: Processor) ->...

FILE: src/llamafactory/v1/plugins/trainer_plugins/distributed/fsdp2.py
  function get_transformer_layer_cls (line 38) | def get_transformer_layer_cls(model: HFModel) -> type[nn.Module] | None:
  function save_model (line 54) | def save_model(model: HFModel, output_dir: str, processor: Processor) ->...
  class FSDP2Engine (line 68) | class FSDP2Engine:
    method __init__ (line 69) | def __init__(self, dist_config: dict):
    method get_mp_policy (line 96) | def get_mp_policy(self) -> MixedPrecisionPolicy:
    method is_lora_module_wrap (line 113) | def is_lora_module_wrap(self, model) -> bool:
    method prepare_model (line 116) | def prepare_model(self, model: HFModel) -> HFModel:
    method materialize_and_load (line 196) | def materialize_and_load(self, model: HFModel, hf_model_path: str, dcp...
    method _save_non_persistent_buffers (line 216) | def _save_non_persistent_buffers(self, model: HFModel) -> dict:
    method _restore_non_persistent_buffers (line 229) | def _restore_non_persistent_buffers(self, model: HFModel, saved_buffer...
    method shard_model (line 246) | def shard_model(self, model: HFModel) -> HFModel:
    method _load_from_dcp (line 266) | def _load_from_dcp(self, model: HFModel, dcp_path: str):
    method _load_weights_from_hf_checkpoint (line 285) | def _load_weights_from_hf_checkpoint(self, model: HFModel, hf_model_pa...
    method _resolve_hf_checkpoint_dir (line 348) | def _resolve_hf_checkpoint_dir(self, hf_model_path: str) -> str:
    method _copy_weights (line 433) | def _copy_weights(self, param, loaded_tensor):

FILE: src/llamafactory/v1/plugins/trainer_plugins/distributed/hub.py
  class DistributedPlugin (line 27) | class DistributedPlugin(BasePlugin):
    method __call__ (line 28) | def __call__(self, model: HFModel, dist_config: PluginConfig, **kwargs...
  function shard_model_fsdp2 (line 33) | def shard_model_fsdp2(model: HFModel, dist_config: PluginConfig, **kwarg...
  function save_model_fsdp2 (line 40) | def save_model_fsdp2(model: HFModel, output_dir: str, processor: Process...
  function shard_model_deepspeed (line 47) | def shard_model_deepspeed(model: HFModel, dist_config: PluginConfig, **k...
  function save_model_deepspeed (line 58) | def save_model_deepspeed(model: HFModel, output_dir: str, processor: Pro...

FILE: src/llamafactory/v1/plugins/trainer_plugins/lr_scheduler.py
  class LRSchedulerPlugin (line 18) | class LRSchedulerPlugin(BasePlugin):

FILE: src/llamafactory/v1/plugins/trainer_plugins/optimizer.py
  class OptimizerPlugin (line 18) | class OptimizerPlugin(BasePlugin):

FILE: src/llamafactory/v1/samplers/cli_sampler.py
  class SyncSampler (line 28) | class SyncSampler(BaseSampler):
    method __init__ (line 29) | def __init__(
    method generate (line 45) | def generate(self, messages: list[Message], tools: str | None = None) ...
    method batch_infer (line 63) | def batch_infer(self, dataset: TorchDataset) -> list[Sample]:
  function run_chat (line 75) | def run_chat(args: InputArgument = None):

FILE: src/llamafactory/v1/trainers/sft_trainer.py
  class SFTTrainer (line 24) | class SFTTrainer(BaseTrainer):
    method compute_loss (line 25) | def compute_loss(self, batch: BatchInput) -> Tensor:
  function run_sft (line 32) | def run_sft(args: InputArgument = None):

FILE: src/llamafactory/v1/utils/dtype.py
  class DtypeRegistry (line 26) | class DtypeRegistry:
  class DtypeInterface (line 32) | class DtypeInterface:
    method is_available (line 40) | def is_available(precision: str | torch.dtype) -> bool:
    method is_fp16 (line 51) | def is_fp16(precision: str | torch.dtype) -> bool:
    method is_fp32 (line 55) | def is_fp32(precision: str | torch.dtype) -> bool:
    method is_bf16 (line 59) | def is_bf16(precision: str | torch.dtype) -> bool:
    method to_dtype (line 63) | def to_dtype(precision: str | torch.dtype) -> torch.dtype:
    method to_str (line 74) | def to_str(precision: torch.dtype) -> str:
    method set_dtype (line 85) | def set_dtype(self, precision: str | torch.dtype):

FILE: src/llamafactory/v1/utils/env.py
  function find_available_port (line 19) | def find_available_port() -> int:
  function is_env_enabled (line 28) | def is_env_enabled(env_var: str, default: str = "0") -> bool:
  function use_ray (line 33) | def use_ray() -> bool:
  function use_kt (line 37) | def use_kt() -> bool:

FILE: src/llamafactory/v1/utils/helper.py
  function set_seed (line 25) | def set_seed(seed: int) -> None:
  function is_tokenizer (line 34) | def is_tokenizer(processor: Processor) -> bool:
  function get_tokenizer (line 46) | def get_tokenizer(processor: Processor) -> PreTrainedTokenizer:
  function _pad_and_truncate (line 58) | def _pad_and_truncate(tensor: Tensor, max_seqlen: int, pad_value: int = ...
  function pad_and_truncate (line 68) | def pad_and_truncate(samples: list[ModelInput], max_seqlen: int) -> list...
  function compute_valid_tokens (line 89) | def compute_valid_tokens(batches: list[BatchInput]) -> int:

FILE: src/llamafactory/v1/utils/logging.py
  class _Logger (line 31) | class _Logger(logging.Logger):
    method info_rank0 (line 34) | def info_rank0(self, *args, **kwargs) -> None:
    method warning_rank0 (line 37) | def warning_rank0(self, *args, **kwargs) -> None:
    method warning_rank0_once (line 40) | def warning_rank0_once(self, *args, **kwargs) -> None:
  function _get_default_logging_level (line 44) | def _get_default_logging_level() -> "logging._Level":
  function _get_library_name (line 56) | def _get_library_name() -> str:
  function _get_library_root_logger (line 60) | def _get_library_root_logger() -> "_Logger":
  function _configure_library_root_logger (line 64) | def _configure_library_root_logger() -> None:
  function get_logger (line 84) | def get_logger(name: str | None = None) -> "_Logger":
  function add_handler (line 93) | def add_handler(handler: "logging.Handler") -> None:
  function remove_handler (line 99) | def remove_handler(handler: logging.Handler) -> None:
  function info_rank0 (line 105) | def info_rank0(self: "logging.Logger", *args, **kwargs) -> None:
  function warning_rank0 (line 110) | def warning_rank0(self: "logging.Logger", *args, **kwargs) -> None:
  function warning_rank0_once (line 116) | def warning_rank0_once(self: "logging.Logger", *args, **kwargs) -> None:

FILE: src/llamafactory/v1/utils/objects.py
  class StatefulBuffer (line 21) | class StatefulBuffer:
    method __init__ (line 24) | def __init__(self, max_buffer_size: int = 1_000_000_000) -> None:
    method __len__ (line 29) | def __len__(self) -> int:
    method size (line 33) | def size(self) -> int:
    method put (line 36) | def put(self, samples: list[ModelInput]) -> None:
    method get (line 45) | def get(self, value: int) -> list[ModelInput]:
    method clear (line 52) | def clear(self) -> None:
    method state_dict (line 57) | def state_dict(self) -> dict:
    method load_state_dict (line 64) | def load_state_dict(self, state_dict: dict) -> None:

FILE: src/llamafactory/v1/utils/packages.py
  function _is_package_available (line 37) | def _is_package_available(name: str) -> bool:
  function _get_package_version (line 41) | def _get_package_version(name: str) -> "Version":
  function is_transformers_version_greater_than (line 49) | def is_transformers_version_greater_than(content: str):
  function check_version (line 53) | def check_version(requirement: str, mandatory: bool = False) -> None:

FILE: src/llamafactory/v1/utils/plugin.py
  class BasePlugin (line 26) | class BasePlugin:
    method __init__ (line 55) | def __init__(self, name: str | None = None) -> None:
    method register (line 59) | def register(self, method_name: str = "__call__") -> Callable:
    method __call__ (line 73) | def __call__(self, *args, **kwargs) -> Any:
    method __getattr__ (line 77) | def __getattr__(self, method_name: str) -> Callable:
    method __getitem__ (line 81) | def __getitem__(self, method_name: str) -> Callable:
  class PrintPlugin (line 94) | class PrintPlugin(BasePlugin):
    method again (line 95) | def again(self):  # optional
  function print_hello (line 99) | def print_hello():
  function print_hello_again (line 103) | def print_hello_again():

FILE: src/llamafactory/v1/utils/pytest.py
  function dist_env (line 20) | def dist_env(local_rank: int = 0, world_size: int = 1, master_port: int ...

FILE: src/llamafactory/v1/utils/types.py
  class DatasetInfo (line 57) | class DatasetInfo(TypedDict, total=False):
  class DistributedConfig (line 74) | class DistributedConfig(TypedDict, total=False):
  class Content (line 87) | class Content(TypedDict):
  class Message (line 94) | class Message(TypedDict):
  class SFTSample (line 103) | class SFTSample(TypedDict):
  class DPOSample (line 114) | class DPOSample(TypedDict):
  class ToolCall (line 130) | class ToolCall(TypedDict):
  class ModelInput (line 137) | class ModelInput(TypedDict, total=False):
  class BatchInput (line 152) | class BatchInput(TypedDict, total=False):
  class BatchInfo (line 167) | class BatchInfo(TypedDict):
  class ModelOutput (line 178) | class ModelOutput(NamedTuple):

FILE: src/llamafactory/webui/chatter.py
  function _escape_html (line 41) | def _escape_html(text: str) -> str:
  function _format_response (line 46) | def _format_response(text: str, lang: str, escape_html: bool, thought_wo...
  function update_attr (line 73) | def update_attr(obj: Any, name: str, value: Any):
  class WebChatModel (line 80) | class WebChatModel(ChatModel):
    method __init__ (line 81) | def __init__(self, manager: "Manager", demo_mode: bool = False, lazy_i...
    method loaded (line 98) | def loaded(self) -> bool:
    method load_model (line 101) | def load_model(self, data) -> Generator[str, None, None]:
    method unload_model (line 161) | def unload_model(self, data) -> Generator[str, None, None]:
    method append (line 175) | def append(
    method stream (line 193) | def stream(

FILE: src/llamafactory/webui/common.py
  function abort_process (line 46) | def abort_process(pid: int) -> None:
  function get_save_dir (line 59) | def get_save_dir(*paths: str) -> os.PathLike:
  function _get_config_path (line 69) | def _get_config_path() -> os.PathLike:
  function load_config (line 74) | def load_config() -> dict[str, str | dict[str, Any]]:
  function save_config (line 83) | def save_config(
  function get_model_path (line 103) | def get_model_path(model_name: str) -> str:
  function get_template (line 125) | def get_template(model_name: str) -> str:
  function get_time (line 130) | def get_time() -> str:
  function is_multimodal (line 135) | def is_multimodal(model_name: str) -> bool:
  function load_dataset_info (line 140) | def load_dataset_info(dataset_dir: str) -> dict[str, dict[str, Any]]:
  function load_args (line 154) | def load_args(config_path: str) -> dict[str, Any] | None:
  function save_args (line 163) | def save_args(config_path: str, config_dict: dict[str, Any]) -> None:
  function _clean_cmd (line 169) | def _clean_cmd(args: dict[str, Any]) -> dict[str, Any]:
  function gen_cmd (line 182) | def gen_cmd(args: dict[str, Any]) -> str:
  function save_cmd (line 202) | def save_cmd(args: dict[str, Any]) -> str:
  function load_eval_results (line 212) | def load_eval_results(path: os.PathLike) -> str:
  function calculate_pixels (line 220) | def calculate_pixels(pixels: str) -> int:
  function create_ds_config (line 228) | def create_ds_config() -> None:

FILE: src/llamafactory/webui/components/chatbot.py
  function check_json_schema (line 34) | def check_json_schema(text: str, lang: str) -> None:
  function create_chat_box (line 49) | def create_chat_box(

FILE: src/llamafactory/webui/components/data.py
  function prev_page (line 34) | def prev_page(page_index: int) -> int:
  function next_page (line 38) | def next_page(page_index: int, total_num: int) -> int:
  function can_preview (line 42) | def can_preview(dataset_dir: str, dataset: list) -> "gr.Button":
  function _load_data_file (line 60) | def _load_data_file(file_path: str) -> list[Any]:
  function get_preview (line 70) | def get_preview(dataset_dir: str, dataset: list, page_index: int) -> tup...
  function create_preview_box (line 86) | def create_preview_box(dataset_dir: "gr.Textbox", dataset: "gr.Dropdown"...

FILE: src/llamafactory/webui/components/eval.py
  function create_eval_tab (line 33) | def create_eval_tab(engine: "Engine") -> dict[str, "Component"]:

FILE: src/llamafactory/webui/components/export.py
  function can_quantize (line 40) | def can_quantize(checkpoint_path: str | list[str]) -> "gr.Dropdown":
  function save_model (line 47) | def save_model(
  function create_export_tab (line 118) | def create_export_tab(engine: "Engine") -> dict[str, "Component"]:

FILE: src/llamafactory/webui/components/footer.py
  function get_device_memory (line 29) | def get_device_memory() -> "gr.Slider":
  function create_footer (line 39) | def create_footer() -> dict[str, "Component"]:

FILE: src/llamafactory/webui/components/infer.py
  function create_infer_tab (line 32) | def create_infer_tab(engine: "Engine") -> dict[str, "Component"]:

FILE: src/llamafactory/webui/components/top.py
  function create_top (line 33) | def create_top() -> dict[str, "Component"]:

FILE: src/llamafactory/webui/components/train.py
  function create_train_tab (line 37) | def create_train_tab(engine: "Engine") -> dict[str, "Component"]:

FILE: src/llamafactory/webui/control.py
  function switch_hub (line 41) | def switch_hub(hub_name: str) -> None:
  function can_quantize (line 50) | def can_quantize(finetuning_type: str) -> "gr.Dropdown":
  function can_quantize_to (line 62) | def can_quantize_to(quantization_method: str) -> "gr.Dropdown":
  function change_stage (line 78) | def change_stage(training_stage: str = list(TRAINING_STAGES.keys())[0]) ...
  function get_model_info (line 87) | def get_model_info(model_name: str) -> tuple[str, str]:
  function check_template (line 96) | def check_template(lang: str, template: str) -> None:
  function get_trainer_info (line 107) | def get_trainer_info(lang: str, output_path: os.PathLike, do_train: bool...
  function list_checkpoints (line 160) | def list_checkpoints(model_name: str, finetuning_type: str) -> "gr.Dropd...
  function list_config_paths (line 182) | def list_config_paths(current_time: str) -> "gr.Dropdown":
  function list_datasets (line 197) | def list_datasets(dataset_dir: str = None, training_stage: str = list(TR...
  function list_output_dirs (line 209) | def list_output_dirs(model_name: str | None, finetuning_type: str, curre...

FILE: src/llamafactory/webui/engine.py
  class Engine (line 28) | class Engine:
    method __init__ (line 31) | def __init__(self, demo_mode: bool = False, pure_chat: bool = False) -...
    method _update_component (line 40) | def _update_component(self, input_dict: dict[str, dict[str, Any]]) -> ...
    method resume (line 49) | def resume(self):
    method change_lang (line 77) | def change_lang(self, lang: str):

FILE: src/llamafactory/webui/interface.py
  function create_ui (line 38) | def create_ui(demo_mode: bool = False) -> "gr.Blocks":
  function create_web_demo (line 73) | def create_web_demo() -> "gr.Blocks":
  function run_web_ui (line 91) | def run_web_ui() -> None:
  function run_web_demo (line 100) | def run_web_demo() -> None:

FILE: src/llamafactory/webui/manager.py
  class Manager (line 23) | class Manager:
    method __init__ (line 26) | def __init__(self) -> None:
    method add_elems (line 30) | def add_elems(self, tab_name: str, elem_dict: dict[str, "Component"]) ...
    method get_elem_list (line 37) | def get_elem_list(self) -> list["Component"]:
    method get_elem_iter (line 41) | def get_elem_iter(self) -> Generator[tuple[str, "Component"], None, No...
    method get_elem_by_id (line 46) | def get_elem_by_id(self, elem_id: str) -> "Component":
    method get_id_by_elem (line 53) | def get_id_by_elem(self, elem: "Component") -> str:
    method get_base_elems (line 57) | def get_base_elems(self) -> set["Component"]:

FILE: src/llamafactory/webui/runner.py
  class Runner (line 54) | class Runner:
    method __init__ (line 57) | def __init__(self, manager: "Manager", demo_mode: bool = False) -> None:
    method set_abort (line 69) | def set_abort(self) -> None:
    method _initialize (line 74) | def _initialize(self, data: dict["Component", Any], do_train: bool, fr...
    method _finalize (line 116) | def _finalize(self, lang: str, finish_info: str) -> None:
    method _parse_train_args (line 126) | def _parse_train_args(self, data: dict["Component", Any]) -> dict[str,...
    method _parse_eval_args (line 292) | def _parse_eval_args(self, data: dict["Component", Any]) -> dict[str, ...
    method _preview (line 346) | def _preview(self, data: dict["Component", Any], do_train: bool) -> Ge...
    method _launch (line 357) | def _launch(self, data: dict["Component", Any], do_train: bool) -> Gen...
    method _build_config_dict (line 381) | def _build_config_dict(self, data: dict["Component", Any]) -> dict[str...
    method preview_train (line 392) | def preview_train(self, data):
    method preview_eval (line 395) | def preview_eval(self, data):
    method run_train (line 398) | def run_train(self, data):
    method run_eval (line 401) | def run_eval(self, data):
    method monitor (line 404) | def monitor(self):
    method save_args (line 462) | def save_args(self, data):
    method load_args (line 478) | def load_args(self, lang: str, config_path: str):
    method check_output_dir (line 492) | def check_output_dir(self, lang: str, model_name: str, finetuning_type...

FILE: src/train.py
  function main (line 18) | def main():
  function _mp_fn (line 22) | def _mp_fn(index):

FILE: src/webui.py
  function main (line 21) | def main():

FILE: tests/check_license.py
  function main (line 22) | def main():

FILE: tests/conftest.py
  function pytest_configure (line 35) | def pytest_configure(config: Config):
  function _handle_runs_on (line 51) | def _handle_runs_on(items: list[Item]):
  function _handle_slow_tests (line 66) | def _handle_slow_tests(items: list[Item]):
  function _get_visible_devices_env (line 75) | def _get_visible_devices_env() -> str | None:
  function _handle_device_visibility (line 85) | def _handle_device_visibility(items: list[Item]):
  function pytest_collection_modifyitems (line 109) | def pytest_collection_modifyitems(config: Config, items: list[Item]):
  function _cleanup_distributed_state (line 123) | def _cleanup_distributed_state():
  function _manage_distributed_env (line 131) | def _manage_distributed_env(request: FixtureRequest, monkeypatch: Monkey...
  function fix_valuehead_cpu_loading (line 166) | def fix_valuehead_cpu_loading():
  function bypass_mistral_regex_check (line 172) | def bypass_mistral_regex_check():

FILE: tests/data/processor/test_feedback.py
  function test_feedback_data (line 48) | def test_feedback_data(num_samples: int):

FILE: tests/data/processor/test_pairwise.py
  function _convert_sharegpt_to_openai (line 46) | def _convert_sharegpt_to_openai(messages: list[dict[str, str]]) -> list[...
  function test_pairwise_data (line 57) | def test_pairwise_data(num_samples: int):

FILE: tests/data/processor/test_processor_utils.py
  function test_infer_seqlen (line 34) | def test_infer_seqlen(test_input: tuple[int, int, int], test_output: tup...

FILE: tests/data/processor/test_supervised.py
  function test_supervised_single_turn (line 48) | def test_supervised_single_turn(num_samples: int):
  function test_supervised_multi_turn (line 77) | def test_supervised_multi_turn(num_samples: int):
  function test_supervised_train_on_prompt (line 95) | def test_supervised_train_on_prompt(num_samples: int):
  function test_supervised_mask_history (line 113) | def test_supervised_mask_history(num_samples: int):

FILE: tests/data/processor/test_unsupervised.py
  function test_unsupervised_data (line 52) | def test_unsupervised_data(num_samples: int):

FILE: tests/data/test_collator.py
  function test_base_collator (line 34) | def test_base_collator():
  function test_multimodal_collator (line 78) | def test_multimodal_collator():
  function test_4d_attention_mask (line 136) | def test_4d_attention_mask():

FILE: tests/data/test_converter.py
  function test_alpaca_converter (line 24) | def test_alpaca_converter():
  function test_sharegpt_converter (line 45) | def test_sharegpt_converter():

FILE: tests/data/test_formatter.py
  function test_empty_formatter (line 42) | def test_empty_formatter():
  function test_string_formatter (line 48) | def test_string_formatter():
  function test_function_formatter (line 54) | def test_function_formatter():
  function test_multi_function_formatter (line 64) | def test_multi_function_formatter():
  function test_default_tool_formatter (line 75) | def test_default_tool_formatter():
  function test_default_tool_extractor (line 94) | def test_default_tool_extractor():
  function test_default_multi_tool_extractor (line 101) | def test_default_multi_tool_extractor():
  function test_glm4_function_formatter (line 114) | def test_glm4_function_formatter():
  function test_glm4_tool_formatter (line 121) | def test_glm4_tool_formatter():
  function test_glm4_tool_extractor (line 132) | def test_glm4_tool_extractor():
  function test_llama3_function_formatter (line 139) | def test_llama3_function_formatter():
  function test_llama3_multi_function_formatter (line 148) | def test_llama3_multi_function_formatter():
  function test_llama3_tool_formatter (line 159) | def test_llama3_tool_formatter():
  function test_llama3_tool_extractor (line 173) | def test_llama3_tool_extractor():
  function test_llama3_multi_tool_extractor (line 180) | def test_llama3_multi_tool_extractor():
  function test_mistral_function_formatter (line 193) | def test_mistral_function_formatter():
  function test_mistral_multi_function_formatter (line 203) | def test_mistral_multi_function_formatter():
  function test_mistral_tool_formatter (line 215) | def test_mistral_tool_formatter():
  function test_mistral_tool_extractor (line 224) | def test_mistral_tool_extractor():
  function test_mistral_multi_tool_extractor (line 231) | def test_mistral_multi_tool_extractor():
  function test_qwen_function_formatter (line 244) | def test_qwen_function_formatter():
  function test_qwen_multi_function_formatter (line 253) | def test_qwen_multi_function_formatter():
  function test_qwen_tool_formatter (line 264) | def test_qwen_tool_formatter():
  function test_qwen_tool_extractor (line 278) | def test_qwen_tool_extractor():
  function test_qwen_multi_tool_extractor (line 285) | def test_qwen_multi_tool_extractor():
  function test_lfm2_function_formatter (line 298) | def test_lfm2_function_formatter():
  function test_lfm2_multi_function_formatter (line 307) | def test_lfm2_multi_function_formatter():
  function test_lfm2_tool_formatter (line 317) | def test_lfm2_tool_formatter():
  function test_lfm2_tool_extractor (line 325) | def test_lfm2_tool_extractor():
  function test_lfm2_multi_tool_extractor (line 332) | def test_lfm2_multi_tool_extractor():
  function test_lfm2_tool_extractor_with_nested_dict (line 342) | def test_lfm2_tool_extractor_with_nested_dict():
  function test_lfm2_tool_extractor_with_list_arg (line 354) | def test_lfm2_tool_extractor_with_list_arg():
  function test_lfm2_tool_extractor_no_match (line 366) | def test_lfm2_tool_extractor_no_match():
  function test_lfm2_tool_round_trip (line 374) | def test_lfm2_tool_round_trip():

FILE: tests/data/test_loader.py
  function test_load_train_only (line 44) | def test_load_train_only():
  function test_load_val_size (line 51) | def test_load_val_size():
  function test_load_eval_data (line 58) | def test_load_eval_data():

FILE: tests/data/test_mm_plugin.py
  function _get_mm_inputs (line 93) | def _get_mm_inputs(processor: "ProcessorMixin") -> dict[str, "torch.Tens...
  function _get_omni_inputs (line 98) | def _get_omni_inputs(processor: "ProcessorMixin") -> dict[str, "torch.Te...
  function _is_close (line 117) | def _is_close(batch_a: dict[str, Any], batch_b: dict[str, Any]) -> None:
  function _load_tokenizer_module (line 130) | def _load_tokenizer_module(model_name_or_path: str) -> "TokenizerModule":
  function _check_plugin (line 135) | def _check_plugin(
  function test_base_plugin (line 183) | def test_base_plugin():
  function test_gemma3_plugin (line 193) | def test_gemma3_plugin():
  function test_internvl_plugin (line 215) | def test_internvl_plugin():
  function test_llama4_plugin (line 234) | def test_llama4_plugin():
  function test_llava_plugin (line 255) | def test_llava_plugin():
  function test_llava_next_plugin (line 269) | def test_llava_next_plugin():
  function test_llava_next_video_plugin (line 283) | def test_llava_next_video_plugin():
  function test_paligemma_plugin (line 298) | def test_paligemma_plugin():
  function test_pixtral_plugin (line 318) | def test_pixtral_plugin():
  function test_qwen2_omni_plugin (line 341) | def test_qwen2_omni_plugin():
  function test_qwen2_vl_plugin (line 371) | def test_qwen2_vl_plugin():
  function test_qwen3_vl_plugin (line 389) | def test_qwen3_vl_plugin():
  function test_video_llava_plugin (line 411) | def test_video_llava_plugin():
  function test_lfm2_vl_plugin (line 425) | def test_lfm2_vl_plugin():

FILE: tests/data/test_template.py
  function _check_tokenization (line 51) | def _check_tokenization(
  function _check_template (line 64) | def _check_template(
  function test_encode_oneturn (line 95) | def test_encode_oneturn():
  function test_encode_multiturn (line 110) | def test_encode_multiturn():
  function test_reasoning_encode_oneturn (line 133) | def test_reasoning_encode_oneturn(cot_messages: bool, enable_thinking: b...
  function test_reasoning_encode_multiturn (line 159) | def test_reasoning_encode_multiturn(cot_messages: bool, enable_thinking:...
  function test_jinja_template (line 186) | def test_jinja_template():
  function test_ollama_modelfile (line 196) | def test_ollama_modelfile():
  function test_get_stop_token_ids (line 214) | def test_get_stop_token_ids():
  function test_gemma_template (line 222) | def test_gemma_template():
  function test_gemma2_template (line 235) | def test_gemma2_template():
  function test_llama3_template (line 248) | def test_llama3_template():
  function test_llama4_template (line 260) | def test_llama4_template():
  function test_phi4_template (line 272) | def test_phi4_template():
  function test_qwen2_5_template (line 285) | def test_qwen2_5_template():
  function test_qwen3_template (line 299) | def test_qwen3_template(cot_messages: bool):
  function test_parse_llama3_template (line 317) | def test_parse_llama3_template():
  function test_parse_qwen_template (line 332) | def test_parse_qwen_template():
  function test_parse_qwen3_template (line 345) | def test_parse_qwen3_template():

FILE: tests/e2e/test_chat.py
  function test_chat (line 41) | def test_chat():
  function test_stream_chat (line 47) | def test_stream_chat():

FILE: tests/e2e/test_sglang.py
  function test_chat (line 44) | def test_chat():
  function test_stream_chat (line 54) | def test_stream_chat():

FILE: tests/e2e/test_train.py
  function test_run_exp (line 63) | def test_run_exp(stage: str, dataset: str):
  function test_export (line 70) | def test_export():

FILE: tests/eval/test_eval_template.py
  function test_eval_template_en (line 21) | def test_eval_template_en():
  function test_eval_template_zh (line 60) | def test_eval_template_zh():

FILE: tests/model/model_utils/test_add_tokens.py
  function test_add_tokens (line 29) | def test_add_tokens(special_tokens: bool):

FILE: tests/model/model_utils/test_attention.py
  function is_torch_sdpa_available (line 26) | def is_torch_sdpa_available():
  function test_attention (line 43) | def test_attention():

FILE: tests/model/model_utils/test_checkpointing.py
  function test_vanilla_checkpointing (line 43) | def test_vanilla_checkpointing(disable_gradient_checkpointing: bool):
  function test_unsloth_gradient_checkpointing (line 49) | def test_unsloth_gradient_checkpointing():
  function test_upcast_layernorm (line 55) | def test_upcast_layernorm():
  function test_upcast_lmhead_output (line 62) | def test_upcast_lmhead_output():

FILE: tests/model/model_utils/test_misc.py
  function test_expanded_modules (line 28) | def test_expanded_modules():

FILE: tests/model/model_utils/test_packing.py
  function test_get_seqlens_in_batch (line 37) | def test_get_seqlens_in_batch(attention_mask, golden_seq_lens):
  function test_get_unpad_data (line 63) | def test_get_unpad_data(attention_mask, golden_indices, golden_cu_seqlen...

FILE: tests/model/model_utils/test_visual.py
  function test_visual_full (line 30) | def test_visual_full(freeze_vision_tower: bool, freeze_multi_modal_proje...
  function test_visual_lora (line 53) | def test_visual_lora(freeze_vision_tower: bool, freeze_language_model: b...
  function test_visual_model_save_load (line 84) | def test_visual_model_save_load():

FILE: tests/model/test_base.py
  function test_base (line 33) | def test_base():
  function test_valuehead (line 40) | def test_valuehead():

FILE: tests/model/test_freeze.py
  function test_freeze_train_all_modules (line 46) | def test_freeze_train_all_modules():
  function test_freeze_train_extra_modules (line 57) | def test_freeze_train_extra_modules():
  function test_freeze_inference (line 68) | def test_freeze_inference():

FILE: tests/model/test_full.py
  function test_full_train (line 46) | def test_full_train():
  function test_full_inference (line 53) | def test_full_inference():

FILE: tests/model/test_lora.py
  function test_lora_train_qv_modules (line 58) | def test_lora_train_qv_modules():
  function test_lora_train_all_modules (line 64) | def test_lora_train_all_modules():
  function test_lora_train_extra_modules (line 70) | def test_lora_train_extra_modules():
  function test_lora_train_old_adapters (line 76) | def test_lora_train_old_adapters():
  function test_lora_train_new_adapters (line 82) | def test_lora_train_new_adapters():
  function test_lora_train_valuehead (line 91) | def test_lora_train_valuehead():
  function test_lora_inference (line 100) | def test_lora_inference():

FILE: tests/model/test_pissa.py
  function test_pissa_train (line 53) | def test_pissa_train():
  function test_pissa_inference (line 60) | def test_pissa_inference():

FILE: tests/train/test_sft_trainer.py
  class DataCollatorWithVerbose (line 49) | class DataCollatorWithVerbose(DataCollatorWithPadding):
    method __call__ (line 52) | def __call__(self, features: list[dict[str, Any]]) -> dict[str, Any]:
  function test_shuffle (line 63) | def test_shuffle(disable_shuffling: bool):

FILE: tests_v1/accelerator/test_interface.py
  function _all_reduce_tests (line 26) | def _all_reduce_tests(local_rank: int, world_size: int, master_port: int):
  function test_all_device (line 48) | def test_all_device():
  function test_multi_device (line 57) | def test_multi_device():

FILE: tests_v1/config/test_args_parser.py
  function test_get_args_from_yaml (line 22) | def test_get_args_from_yaml(tmp_path: Path):

FILE: tests_v1/conftest.py
  function pytest_configure (line 36) | def pytest_configure(config: Config):
  function _handle_runs_on (line 52) | def _handle_runs_on(items: list[Item]):
  function _handle_slow_tests (line 67) | def _handle_slow_tests(items: list[Item]):
  function _get_visible_devices_env (line 76) | def _get_visible_devices_env() -> str | None:
  function _handle_device_visibility (line 86) | def _handle_device_visibility(items: list[Item]):
  function pytest_collection_modifyitems (line 110) | def pytest_collection_modifyitems(config: Config, items: list[Item]):
  function _set_env (line 124) | def _set_env():
  function _cleanup_distributed_state (line 135) | def _cleanup_distributed_state():
  function _manage_distributed_env (line 143) | def _manage_distributed_env(request: FixtureRequest, monkeypatch: Monkey...
  function bypass_mistral_regex_check (line 178) | def bypass_mistral_regex_check():

FILE: tests_v1/core/test_data_engine.py
  function test_map_dataset (line 25) | def test_map_dataset(num_samples: int):

FILE: tests_v1/core/test_model_loader.py
  function test_tiny_qwen (line 21) | def test_tiny_qwen():
  function test_tiny_qwen_with_kernel_plugin (line 30) | def test_tiny_qwen_with_kernel_plugin():

FILE: tests_v1/core/utils/test_batching.py
  function test_normal_batching (line 21) | def test_normal_batching():

FILE: tests_v1/core/utils/test_rendering.py
  function _get_input_ids (line 26) | def _get_input_ids(inputs: list | dict) -> list:
  function test_chatml_rendering (line 87) | def test_chatml_rendering():
  function test_chatml_parse (line 109) | def test_chatml_parse():
  function test_chatml_rendering_remote (line 118) | def test_chatml_rendering_remote(num_samples: int):
  function test_qwen3_nothink_rendering (line 130) | def test_qwen3_nothink_rendering():
  function test_qwen3_nothink_parse (line 158) | def test_qwen3_nothink_parse():
  function test_qwen3_nothink_rendering_remote (line 178) | def test_qwen3_nothink_rendering_remote(num_samples: int):
  function test_process_sft_samples (line 198) | def test_process_sft_samples():
  function test_process_dpo_samples (line 211) | def test_process_dpo_samples():

FILE: tests_v1/plugins/data_plugins/test_converter.py
  function test_alpaca_converter (line 26) | def test_alpaca_converter(num_samples: int):
  function test_sharegpt_converter (line 52) | def test_sharegpt_converter():
  function test_pair_converter (line 75) | def test_pair_converter(num_samples: int):

FILE: tests_v1/plugins/model_plugins/test_init_plugin.py
  function test_init_on_meta (line 21) | def test_init_on_meta():
  function test_init_on_rank0 (line 32) | def test_init_on_rank0():
  function test_init_on_default (line 46) | def test_init_on_default():

FILE: tests_v1/plugins/model_plugins/test_kernel_plugin.py
  function _apply_kernel (line 22) | def _apply_kernel(rank) -> None:
  function _apply_all_kernels (line 45) | def _apply_all_kernels(rank) -> None:
  function test_apply_kernel (line 68) | def test_apply_kernel():
  function test_apply_all_kernels (line 72) | def test_apply_all_kernels():

FILE: tests_v1/plugins/model_plugins/test_peft.py
  function model_path (line 27) | def model_path():
  function model (line 32) | def model(model_path):
  function tokenizer (line 37) | def tokenizer(model_path):
  function adapter_path (line 42) | def adapter_path(tmp_path):
  function test_find_all_linear_modules (line 60) | def test_find_all_linear_modules(model):
  function test_get_lora_model (line 67) | def test_get_lora_model(model):
  function test_get_freeze_model_layers (line 76) | def test_get_freeze_model_layers(model):
  function test_get_freeze_model_modules (line 94) | def test_get_freeze_model_modules(model):
  function test_load_adapter_single_for_inference (line 109) | def test_load_adapter_single_for_inference(model, adapter_path):
  function test_load_adapter_resume_train (line 116) | def test_load_adapter_resume_train(model, adapter_path):
  function test_load_adapter_train_multiple_disallowed (line 123) | def test_load_adapter_train_multiple_disallowed(model, adapter_path):
  function test_load_adapter_infer_multiple_merges (line 129) | def test_load_adapter_infer_multiple_merges(model, adapter_path):
  function test_merge_and_export_model (line 136) | def test_merge_and_export_model(tmp_path, adapter_path):

FILE: tests_v1/plugins/model_plugins/test_quantization_plugin.py
  function check_quantization_status (line 24) | def check_quantization_status(model):
  function test_quantization_plugin (line 39) | def test_quantization_plugin(name, quantization_bit):

FILE: tests_v1/plugins/trainer_plugins/distributed/test_fsdp2.py
  function collect_non_persistent_buffers (line 33) | def collect_non_persistent_buffers(model):
  function test_fsdp2_meta_loading_buffers_and_tied_weights (line 45) | def test_fsdp2_meta_loading_buffers_and_tied_weights():

FILE: tests_v1/sampler/test_cli_sampler.py
  function test_sync_sampler (line 23) | def test_sync_sampler():

FILE: tests_v1/trainers/test_fsdp2_sft_trainer.py
  function test_fsdp2_sft_trainer (line 24) | def test_fsdp2_sft_trainer(tmp_path: Path):