Beta
Explore
Marketplace
Neural Labs
Chat
Wallet
Docs
medllm-pubmedqa-qlora – AI Model by jesjoah | AlphaNeural AI
You can deploy this model and start earning money today!
jesjoah
/
medllm-pubmedqa-qlora
like
0
peft
safetensors
medical
qlora
llama
fine-tuned
en
qiaojin/PubMedQA
openlifescienceai/medmcqa
meta-llama/Llama-3.2-3B-Instruct
adapter
us
Views
No views yet
Model card
Files and Versions
Community
API
Deploy
MedLLM — Medical QA via QLoRA Fine-tuning
Fine-tuned LLaMA 3.2 3B Instruct on PubMedQA + MedMCQA using QLoRA (4-bit quantization) for domain-specific medical question answering.
Model Details
Base model:
LLaMA 3.2 3B Instruct
Technique:
QLoRA (4-bit NF4 quantization)
LoRA rank:
r=16, alpha=32
Trainable parameters:
9.17M (0.28% of total)
Training data:
900 PubMedQA + 5000 MedMCQA examples
GPU:
RTX 4070 8GB VRAM
Training time:
~4 hours
Evaluation Results
Dataset
ROUGE-L
BERTScore
PubMedQA
0.0812
0.8318
MedMCQA
0.0780
0.8177
Combined
0.0801
0.8248
