Samhita/OrpoLlama-3-8B-Instruct-Copy

OrpoLlama-3-8B-Instruct

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B-Instruct on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.1648
• Rewards/chosen: -0.0603
• Rewards/rejected: -0.0824
• Rewards/accuracies: 0.5
• Rewards/margins: 0.0221
• Logps/rejected: -0.8240
• Logps/chosen: -0.6033
• Logits/rejected: -0.1024
• Logits/chosen: -0.2381
• Nll Loss: 1.1016
• Log Odds Ratio: -0.6324
• Log Odds Chosen: 0.4547

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: 8e-06
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 10
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen	Nll Loss	Log Odds Ratio	Log Odds Chosen
0.95	0.5980	74	1.2436	-0.0675	-0.0830	0.4000	0.0155	-0.8295	-0.6746	-0.2051	-0.3444	1.1740	-0.6961	0.2761
0.9613	1.1960	148	1.1952	-0.0621	-0.0799	0.4000	0.0179	-0.7994	-0.6209	-0.1256	-0.2699	1.1280	-0.6717	0.3516
1.5258	1.7939	222	1.1740	-0.0609	-0.0818	0.5	0.0209	-0.8183	-0.6094	-0.1255	-0.2648	1.1099	-0.6414	0.4267
1.1971	2.3919	296	1.1648	-0.0603	-0.0824	0.5	0.0221	-0.8240	-0.6033	-0.1024	-0.2381	1.1016	-0.6324	0.4547

Framework versions

• PEFT 0.11.1
• Transformers 4.41.2
• Pytorch 2.3.1+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1