42 lines
1.0 KiB
YAML
42 lines
1.0 KiB
YAML
assistant_tag: gpt
|
|
bf16: true
|
|
content_tag: value
|
|
cutoff_len: 16384
|
|
dataset: mlfoundations-dev/open-thoughts-science-claude
|
|
dataset_dir: ONLINE
|
|
ddp_timeout: 180000000
|
|
deepspeed: dcft/train/zero3.json
|
|
do_train: true
|
|
enable_liger_kernel: true
|
|
finetuning_type: full
|
|
formatting: sharegpt
|
|
global_batch_size: 512
|
|
gradient_accumulation_steps: 32
|
|
hub_model_id: mlfoundations-dev/qwen_openthoughts_science_claude
|
|
learning_rate: 8.0e-05
|
|
logging_steps: 1
|
|
lr_scheduler_type: cosine
|
|
max_samples: 1000000
|
|
messages: conversations
|
|
model_name_or_path: Qwen/Qwen2.5-7B-Instruct
|
|
neat_packing: true
|
|
num_train_epochs: 3.0
|
|
output_dir: /data/horse/ws/rehe951g-p_finetuning/checkpoints/qwen_openthoughts_science_claude
|
|
overwrite_cache: true
|
|
overwrite_output_dir: true
|
|
packing: true
|
|
per_device_eval_batch_size: 1
|
|
per_device_train_batch_size: 1
|
|
plot_loss: true
|
|
preprocessing_num_workers: 16
|
|
push_to_db: true
|
|
push_to_hub: true
|
|
report_to: wandb
|
|
role_tag: from
|
|
run_name: qwen_openthoughts_science_claude
|
|
save_steps: 10
|
|
stage: sft
|
|
template: qwen25
|
|
user_tag: human
|
|
warmup_ratio: 0.1
|