Files
ModelHub XC 285056620b 初始化项目,由ModelHub XC社区提供模型
Model: mlfoundations-dev/oh-dcft-v3.1-gpt-4o-mini-qwen
Source: Original Platform
2026-04-26 21:01:02 +08:00

42 lines
1014 B
YAML

adam_beta1: '0.9'
adam_beta2: '0.999'
bf16: 'True'
cutoff_len: '2048'
dataset: mlfoundations-dev/oh-dcft-v3.1-gpt-4o-mini
dataset_dir: ONLINE
ddp_timeout: '180000000'
deepspeed: /opt/ml/code/zero3.json
do_train: 'True'
enable_liger_kernel: 'False'
eval_strategy: epoch
finetuning_type: full
formatting: sharegpt
global_batch_size: '128'
gradient_accumulation_steps: '2'
gradient_checkpointing: 'True'
hub_model_id: mlfoundations-dev/oh-dcft-v3.1-gpt-4o-mini-qwen
learning_rate: 5e-06
logging_steps: '10'
lr_scheduler_type: constant
max_grad_norm: '1'
messages: conversations
model_name_or_path: Qwen/Qwen2.5-7B
neat_packing: 'True'
num_train_epochs: '3.0'
output_dir: /opt/ml/model
overwrite_cache: 'True'
overwrite_output_dir: 'True'
packing: 'True'
per_device_train_batch_size: '8'
plot_loss: 'True'
preprocessing_num_workers: '16'
push_to_db: 'True'
push_to_hub: 'True'
report_to: wandb
run_name: oh-dcft-v3.1-gpt-4o-mini-qwen
save_strategy: epoch
stage: sft
template: qwen
val_size: '0.05'
weight_decay: '0.1'