prm_version2_subsample_hf

This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B-Instruct on the prm_conversations_prm_version2_math+webinstructsub-mcq+webinstructsub-oe+apps_mix_ref_subsample_hf dataset. It achieves the following results on the evaluation set:

• Loss: 0.1349

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: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 8
• total_train_batch_size: 64
• total_eval_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
0.1845	0.1303	500	0.2062
0.1762	0.2606	1000	0.1863
0.1704	0.3910	1500	0.1710
0.1529	0.5213	2000	0.1579
0.1393	0.6516	2500	0.1471
0.1442	0.7819	3000	0.1387
0.13	0.9123	3500	0.1353

Framework versions

• Transformers 4.45.0
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.20.3