llama3-2M-MedEV

This model is a fine-tuned version of unsloth/llama-3-8b-Instruct-bnb-4bit on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.4249
• Bleu: 47.7973

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: 0.0002
• train_batch_size: 32
• eval_batch_size: 16
• seed: 3407
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 5
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
1.534	0.1200	320	1.4902
1.3171	0.2399	640	1.4705
1.29	0.3599	960	1.4644
1.2699	0.4798	1280	1.4287
1.2567	0.5998	1600	1.4576
1.2448	0.7197	1920	1.4196
1.2353	0.8397	2240	1.4249
1.2274	0.9596	2560	1.4172
1.1635	1.0796	2880	1.4180
1.1337	1.1995	3200	1.4219
1.1346	1.3195	3520	1.3954
1.131	1.4394	3840	1.3714
1.1325	1.5594	4160	1.3923
1.1269	1.6793	4480	1.4118
1.1221	1.7993	4800	1.4251
1.1226	1.9192	5120	1.3970
1.0898	2.0392	5440	1.4198
1.0372	2.1591	5760	1.4310
1.0325	2.2791	6080	1.4209
1.0334	2.3990	6400	1.4205
1.0328	2.5190	6720	1.4306
1.0303	2.6389	7040	1.4222
1.0283	2.7589	7360	1.4266
1.0273	2.8788	7680	1.4251
1.0295	2.9988	8000	1.4249

Framework versions

• PEFT 0.10.0
• Transformers 4.40.2
• Pytorch 2.3.0
• Datasets 2.19.1
• Tokenizers 0.19.1