35 lines
754 B
YAML
35 lines
754 B
YAML
bf16: true
|
|
cutoff_len: 2048
|
|
dataset: gpu_aiAgent_zh,gpu_aiAgent_others
|
|
dataset_dir: data
|
|
ddp_timeout: 180000000
|
|
do_train: true
|
|
enable_thinking: true
|
|
finetuning_type: lora
|
|
flash_attn: auto
|
|
gradient_accumulation_steps: 8
|
|
include_num_input_tokens_seen: true
|
|
learning_rate: 5.0e-05
|
|
logging_steps: 5
|
|
lora_alpha: 128
|
|
lora_dropout: 0
|
|
lora_rank: 64
|
|
lora_target: all
|
|
lr_scheduler_type: cosine
|
|
max_grad_norm: 1.0
|
|
max_samples: 100000
|
|
model_name_or_path: Qwen/Qwen3-1.7B-Base
|
|
num_train_epochs: 3.0
|
|
optim: adamw_torch
|
|
output_dir: saves/Qwen3-1.7B-Base/lora/train_2026-01-03
|
|
packing: false
|
|
per_device_train_batch_size: 2
|
|
plot_loss: true
|
|
preprocessing_num_workers: 16
|
|
report_to: none
|
|
save_steps: 100
|
|
stage: sft
|
|
template: default
|
|
trust_remote_code: true
|
|
warmup_steps: 0
|