微调参数
基本参数
FinetuningArguments
LoRA
LoraArguments
r
all
λ = ηB/ηA
ηA, ηB
-1
RLHF
RLHF训练参数介绍
Freeze
FreezeArguments
Apollo
ApolloArguments
BAdam
BAdamArgument
GaLore
GaLoreArguments
数据参数
DataArguments
模型参数
ModelArguments
多模态模型
ProcessorArguments
vllm 推理
VllmArguments
模型量化
QuantizationArguments
模型导出
ExportArguments
评估参数
EvalArguments
生成参数
GeneratingArguments
SwanLab 参数
SwanLabArguments
训练参数
RAY
RayArguments
环境变量
Environment Variables
API_HOST
API_PORT
API_KEY
API_MODEL_NAME
API_VERBOSE
FASTAPI_ROOT_PATH
MAX_CONCURRENT
DISABLE_VERSION_CHECK
FORCE_CHECK_IMPORTS
ALLOW_EXTRA_ARGS
LLAMAFACTORY_VERBOSITY
USE_MODELSCOPE_HUB
USE_OPENMIND_HUB
USE_RAY
RECORD_VRAM
OPTIM_TORCH
NPU_JIT_COMPILE
CUDA_VISIBLE_DEVICES
ASCEND_RT_VISIBLE_DEVICES
FORCE_TORCHRUN
torchrun
MASTER_ADDR
MASTER_PORT
NNODES
NODE_RANK
NNODES-1
NPROC_PER_NODE
WANDB_DISABLED
WANDB_PROJECT
WANDB_API_KEY
GRADIO_SHARE
GRADIO_SERVER_NAME
0.0.0.0
GRADIO_SERVER_PORT
GRADIO_ROOT_PATH
GRADIO_IPV6
ENABLE_SHORT_CONSOLE
lmf
llamafactory-cli