adam_beta1: '0.9'
adam_beta2: '0.999'
assistant_tag: gpt
bf16: 'True'
content_tag: value
cutoff_len: '2048'
dataset: mlfoundations-dev/oh-dcft-v3-llama3.1-nemotron-70b_shareGPT_format
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: '512'
gradient_accumulation_steps: '8'
gradient_checkpointing: 'True'
hub_model_id: mlfoundations-dev/hp_ablations_grid_mistral_bsz512_lr5e-6_scheduler-cosine-warmup0.05-minlr5e-7-mistralv0.3
learning_rate: 5e-06
logging_steps: '10'
lr_scheduler_kwargs: min_lr=5e-07
lr_scheduler_type: cosine_with_min_lr
max_grad_norm: '1'
messages: conversations
model_name_or_path: mistralai/Mistral-7B-v0.1
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
role_tag: from
run_name: hp_ablations_grid_mistral_bsz512_lr5e-6_scheduler-cosine-warmup0.05-minlr5e-7-mistralv0.3
save_strategy: epoch
stage: sft
template: mistral
user_tag: human
val_size: '0.05'
warmup_ratio: '0.05'
weight_decay: '0.1'