bf16: true
cutoff_len: 2048
dataset: limo
dataset_dir: data
ddp_timeout: 180000000
do_train: true
finetuning_type: full
flash_attn: auto
gradient_accumulation_steps: 8
include_num_input_tokens_seen: true
learning_rate: 2.0e-05
logging_steps: 5
lr_scheduler_type: cosine
max_grad_norm: 1.0
max_samples: 100000
model_name_or_path: Qwen/Qwen2.5-1.5B
num_train_epochs: 15.0
optim: adamw_torch
output_dir: saves/Qwen2.5-1.5B/full/train_2025-02-17-12-11-10
packing: false
per_device_train_batch_size: 2
plot_loss: true
preprocessing_num_workers: 16
report_to:
- wandb
save_steps: 100
stage: sft
template: default
trust_remote_code: true
warmup_steps: 0