grounded-ai/phi3.5-rag-relevance-judge

grounded-ai-rag-3

This model is a fine-tuned version of microsoft/Phi-3.5-mini-instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5557
• Rouge1: 1.0
• Rouge2: 0.0
• Rougel: 1.0
• Rougelsum: 1.0

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: 7e-05
• train_batch_size: 2
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 15
• training_steps: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
1.8222	5.0	5	1.9460	1.0	0.0	1.0	1.0
1.7609	10.0	10	1.6547	1.0	0.0	1.0	1.0
1.4433	15.0	15	1.3821	1.0	0.0	1.0	1.0
1.2307	20.0	20	1.1176	1.0	0.0	1.0	1.0
0.9889	25.0	25	0.7975	1.0	0.0	1.0	1.0
0.6934	30.0	30	0.6240	1.0	0.0	1.0	1.0
0.5838	35.0	35	0.5633	1.0	0.0	1.0	1.0
0.5625	40.0	40	0.5557	1.0	0.0	1.0	1.0

Framework versions

• PEFT 0.13.2
• Transformers 4.45.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.1