geshijoker/HealthCareMagic_sft_llama3_instruct_full

sft_full

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

• Loss: 1.7460

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: 1e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 8
• total_train_batch_size: 512
• total_eval_batch_size: 64
• optimizer: Use 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: 0.05
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss
1.2093	2.8429	500	1.7462

Evaluation results

Name	Checkpoint	Rouge1	RougeL	Meteor	Bert Score
baseline instruct model	Meta-Llama-3-8B-Instruct	0.254	0.128	0.222	0.747
full fientune from inst	sft_llama3_instruct_full	0.315	0.189	0.238	0.782
lora sft from inst	sft_llama3_instruct_lora_all	0.271	0.143	0.194	0.774
lora sft from base	sft_llama3_lora_all	0.239	0.113	0.211	0.735
qlora sft from inst	sft_llama3_instruct_qlora_all	0.137	0.071	0.102	0.679
qlora sft from base	sft_llama3_qlora_all	0.192	0.090	0.159	0.718

Framework versions

• Transformers 4.46.1
• Pytorch 2.5.1
• Datasets 3.1.0
• Tokenizers 0.20.3