单卡 A100
Distill_data_17k-train.arrow
"Distill": {
"file_name": "/data/jupyterfile/wei/distill/Distill_data_17k-train.arrow",
"formatting": "sharegpt",
"columns": {
"messages": "conversations",
"system": "system"
},
"tags": {
"role_tag": "from",
"content_tag": "value",
"user_tag": "user",
"assistant_tag": "assistant"
}
}
### model
model_name_or_path: /data/jupyterfile/wei/distill/Qwen2.5-1.5B-Instruct
### method
stage: sft
do_train: true
finetuning_type: full
deepspeed: examples/deepspeed/ds_z3_offload_config.json # choices: [ds_z0_config.json, ds_z2_config.json, ds_z3_config.json]
### dataset
dataset: Distill
template: qwen
cutoff_len: 8192
max_samples: 100000
overwrite_cache: true
preprocessing_num_workers: 16
### output
output_dir: ./saves/MiniDeepSeekR1/full/original
logging_steps: 1
save_steps: 100
plot_loss: true
### train
per_device_train_batch_size: 1
gradient_accumulation_steps: 12
learning_rate: 1.0e-5
num_train_epochs: 1.0 #3.0
lr_scheduler_type: cosine
warmup_ratio: 0.1
bf16: true
ddp_timeout: 180000000
report_to: wandb
run_name: Distill