llama-3-8b-base-new-dpo-hh-harmless-4xh200-batch-64-q_t-0.5-s_star-1.0/README.md

library_name, base_model, tags, datasets, model-index

library_name	base_model	tags	datasets	model-index
transformers	W-61/llama-3-8b-base-sft-hh-harmless-4xh200	alignment-handbook new-dpo generated_from_trainer	Anthropic/hh-rlhf	name results llama-3-8b-base-new-dpo-hh-harmless-4xh200-batch-64-q_t-0.5-s_star-1.0

llama-3-8b-base-new-dpo-hh-harmless-4xh200-batch-64-q_t-0.5-s_star-1.0

This model is a fine-tuned version of W-61/llama-3-8b-base-sft-hh-harmless-4xh200 on the Anthropic/hh-rlhf dataset. It achieves the following results on the evaluation set:

• Loss: 0.5680
• Fcm Dpo/beta: 0.6569
• Margin Dpo/margin Mean: 1.3983
• Margin Dpo/margin Std: 2.4811
• Logps/chosen: -77.8234
• Logps/rejected: -83.9112
• Logps/ref Chosen: -74.8595
• Logps/ref Rejected: -79.5490
• Logits/chosen: 0.1962
• Logits/rejected: 0.1585

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-07
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• total_eval_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Fcm Dpo/beta	Margin Dpo/margin Mean	Margin Dpo/margin Std	Logps/chosen	Logps/rejected	Logps/ref Chosen	Logps/ref Rejected	Logits/chosen	Logits/rejected
1.0309	0.3023	200	0.6114	1.5800	0.5045	1.0288	-75.8633	-81.0573	-74.8595	-79.5490	0.1685	0.1337
1.0799	0.6047	400	0.5736	0.7420	1.1067	2.0391	-77.2471	-83.0433	-74.8595	-79.5490	0.2053	0.1675
1.0999	0.9070	600	0.5680	0.6569	1.3983	2.4811	-77.8234	-83.9112	-74.8595	-79.5490	0.1962	0.1585

Framework versions

• Transformers 4.51.0
• Pytorch 2.3.1+cu121
• Datasets 2.21.0
• Tokenizers 0.21.4