norllm-ai-normistral-7b-align-scan

This model is a fine-tuned version of data/norllm-ai-normistral-7b-sft-qlora on the hugodk-sch/aftonposten_title_prefs dataset. It achieves the following results on the evaluation set:

• Loss: 0.9294
• Rewards/chosen: -0.0900
• Rewards/rejected: -0.1614
• Rewards/accuracies: 0.6009
• Rewards/margins: 0.0715
• Logps/rejected: -35.5053
• Logps/chosen: -31.7323
• Logits/rejected: -2.8259
• Logits/chosen: -2.8279

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: 4
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.9239	0.26	100	0.9661	-0.0010	-0.0351	0.6013	0.0342	-34.8738	-31.2872	-2.8027	-2.8055
0.8146	0.52	200	0.9363	-0.0747	-0.1389	0.6184	0.0641	-35.3925	-31.6561	-2.8206	-2.8233
0.7173	0.78	300	0.9279	-0.0837	-0.1567	0.6125	0.0730	-35.4817	-31.7010	-2.8247	-2.8267

Framework versions

• PEFT 0.10.0
• Transformers 4.39.0.dev0
• Pytorch 2.1.2+cu121
• Datasets 2.14.6
• Tokenizers 0.15.1