news_finetune.yaml

### model
model_name_or_path: Qwen/Qwen2.5-1.5B-Instruct
trust_remote_code: true

### method
stage: sft
do_train: true
finetuning_type: lora
lora_rank: 64
lora_target: all

### dataset
dataset: news_finetune_train
eval_dataset: news_finetune_val
template: qwen
cutoff_len: 3500
# max_samples: 50
overwrite_cache: true
preprocessing_num_workers: 16

### output
# resume_from_checkpoint: /gdrive/MyDrive/youtube-resources/llm-finetuning/models/checkpoint-1500
output_dir: /gdrive/MyDrive/ara-finetune/models/
logging_steps: 10
save_steps: 500
plot_loss: true
# overwrite_output_dir: true

### train
per_device_train_batch_size: 1
gradient_accumulation_steps: 4
learning_rate: 1.0e-4
num_train_epochs: 3.0
lr_scheduler_type: cosine
warmup_ratio: 0.1
bf16: true
ddp_timeout: 180000000

### eval
# val_size: 0.1
per_device_eval_batch_size: 1
eval_strategy: steps
eval_steps: 100

report_to: wandb
run_name: newsx-finetune-llamafactory

push_to_hub: false
# When the setup script generates the runnable YAML, these are filled from .env.
hub_model_id: ""
export_hub_model_id: ""
hub_private_repo: true
hub_strategy: checkpoint