AsclepiusLM / README.md
Chilliwiddit's picture
Create README.md
dfe7db9 verified
metadata
base_model: unsloth/Meta-Llama-3.1-8B-bnb-4bit
library_name: peft
tags:
  - base_model:adapter:unsloth/Meta-Llama-3.1-8B-bnb-4bit
  - lora
  - transformers
language:
  - en
metrics:
  - bleu
  - bertscore
  - rouge
pipeline_tag: summarization

Model Card for Model ID

Basically paired the unsloth/Meta-Llama-3.1-8B-bnb-4bit base model with the fine-tuned Chilliwiddit/Openi-llama3.1-8B-WeightedLoss-small2 adapter.

Training Details

Training Data

I used the Open-i dataset

Training Hyperparameters

  • Training regime: [More Information Needed]

  • 16 Mixed Precision

  • LR of 0.0-1

  • 5 Epochs

  • lambda medical weight of 20 and lambda negation weight of 20

  • Used 2nd iteration of summary medical concepts file