README.md

metadata

language: en
tags:
  - medical
  - deepseek
  - fine-tuned
  - clinical-reasoning
license: apache-2.0
datasets:
  - FreedomIntelligence/medical-o1-reasoning-SFT

Fine-tuned DeepSeek R1 Model for Medical Reasoning

This model is a fine-tuned version of DeepSeek R1 specialized for medical reasoning and clinical decision-making.

Training Details

• Base Model: unsloth/DeepSeek-R1-Distill-Llama-8B
• Training Data: Medical reasoning dataset (FreedomIntelligence/medical-o1-reasoning-SFT)
• Fine-tuning Method: LoRA (Low-Rank Adaptation)
• Training Parameters:
  • Batch Size: 2
  • Learning Rate: 2e-4
  • Epochs: 1
  • Max Sequence Length: 2048

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("Vedant101/fine-tune-deep-seek-r1")
tokenizer = AutoTokenizer.from_pretrained("Vedant101/fine-tune-deep-seek-r1")