sft-4e-exp2

This model is an experimental fine-tuned version of TencentARC/Mistral_Pro_8B_v0.1.

This model is intended for safety research only.

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-06
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• total_eval_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• num_epochs: 4

Training results

arc (25)	hellaswag (10)	mmlu (5)	truthfulqa_mc (20)	winogrande	gsm8k (exact)	avg (exact)
0.6305	0.8413	0.6041	0.5535	0.7624	0.5754	0.6612

Framework versions

• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.0

License

This version is to be used for non-commercial research purposes only. We will release open models after we have finished training and evaluations.