Model Card: Imran1/Med-R1-v1

Model Details

• Name: Imran1/Med-R1-v1
• Size: 1B parameters
• Dataset: FreedomIntelligence/medical-o1-reasoning-SFT
• Max Sequence Length: 2048 tokens
• Quantization: Supports 4-bit inference

Description

Imran1/Med-R1-v1 is a fine-tuned language model for medical reasoning, trained on structured medical datasets to provide contextually relevant and accurate responses for healthcare-related queries.

Intended Use

• Use Cases: Medical Q&A, reasoning for diagnoses, and treatment suggestions
• Users: Healthcare professionals, researchers, and developers

Example

from unsloth import FastLanguageModel
from transformers import TextStreamer

model_name = "Imran1/Med-R1-v1" 
max_seq_length = 2048
model, tokenizer = FastLanguageModel.from_pretrained(
    model_name=model_name,
    max_seq_length=max_seq_length,
    dtype=None,
    load_in_4bit=True,
)

FastLanguageModel.for_inference(model)  # Enable native 2x faster inference

prompt = "i have pain in my kidney, what should i need to do?"

messages = [
    {"role": "system", "content": "You are AI medical assistant."},
    {"role": "user", "content": prompt}
]

text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)

inputs = tokenizer([text], return_tensors="pt").to(model.device)

text_streamer = TextStreamer(tokenizer)

generated_ids = model.generate(
    **inputs,
    streamer=text_streamer,
    max_new_tokens=2000
)

generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]

output

<｜begin▁of▁sentence｜>You are AI medical assistant.<｜User｜>i have pain in my kidney, what should i need to do?<｜Assistant｜><|thinking|> Okay, let's think about this. When we have pain in the kidney, it's definitely something to take seriously. The kidneys are the main filters for blood, so there's definitely a risk of kidney issues.

First, I should consider what the pain might mean. Could it be a urinary tract infection? That's a big possibility, but it could also be something like a kidney stones or a urinary tract blockage. And what about the rest of the body? Could there be a urinary tract infection or some kind of renal or urinary blockage?

Now, I should think about what to do next. If it's a urinary tract infection, I'd want to go for an emergency treatment. This could mean antibiotics or maybe even surgery if it's severe. But if it's not severe, maybe a simple follow-up with a urinalysis could help.

If it's not a kidney blockage, but there's pain, it might mean a urinary tract issue, like a kidney stones. In this case, I'd need to go for a urinalysis to see what's going on.

Also, I should think about other potential issues. Like, could it be something like a renal vein or glomerulonephritis? Those would have different treatment plans.

Hmm, let's not forget about the possibility of something like a kidney transplant. If that's the case, it's important to make sure the patient is healthy and there's no other risk to consider.

So, in the end, the best next step is to check for what's causing the pain. If it's something like a kidney stone, a urinalysis is the way to go. But if it's a urinary tract infection, I'd need to get it treated fast.

Alright, let's wrap it up. If it's a urinary tract infection, I need to act quickly. If it's a kidney stone, I'll need to do a urinalysis. And just in case, a kidney transplant is something to think about, but only if it seems like the best option given the situation. <|thinking_end|>  

 <|Final|> When you experience pain in your kidney, it's important to approach the situation with a clear mind and consider the possible causes of the pain. Here's a structured approach to what you should do next:

1. **Assess the Pain Type**: Determine if the pain is likely due to a urinary tract infection (UTI), a kidney stone, or some other condition. This will guide the treatment approach.

2. **If Urethral Infection**: 
   - If the pain is severe and worsens with exertion, it's likely a urinary tract infection. In this case, you should:
     - **Conduct a Urinalysis**: This will help identify any stones, infections, or other kidney issues.
     - **Immediate Treatment**: Depending on the severity, you may need antibiotics, surgery, or more invasive treatment like a kidney transplant if the infection is severe and advanced.

3. **If Kidney Stones**:
   - If pain is noticeable but not severe, it might indicate a kidney stone. In this situation, a urinalysis is recommended to assess the stone's size and shape.

4. **Other Considerations**:
   - **Renal Vein**: If the pain is in the urinary tract and there's no clear diagnosis of a kidney stone or infection, consider a renal vein. This may involve a lumbar puncture to check for blood flow and assess the condition.

5. **Other Potential Conditions**:
   - **Renal Transplant**: If you suspect renal vein involvement, a renal transplant could be considered, but only if there's no other risk or evidence of another renal issue.

By following this structured approach, you can manage the pain effectively and take appropriate steps to address the underlying cause. <|Final_end|><｜end▁of▁sentence｜>

Limitations

• Not a substitute for medical advice: This model provides general reasoning and suggestions but should not replace professional medical consultation.
• Dataset Bias: The model’s outputs are influenced by the FreedomIntelligence/medical-o1-reasoning-SFT dataset and may reflect its limitations.
• Sensitive Use: Ensure appropriate use in non-critical scenarios and always verify with certified professionals.

Deployment

The model supports efficient deployment with 4-bit quantization for inference. It is optimized for use in applications requiring medical reasoning and is compatible with Hugging Face and unsloth frameworks.

Citation

If you use this model, please cite:

@model{imran1_med_r1_v1,
  title={Imran1/Med-R1-v1: A Medical Reasoning Language Model},
  author={Imran1},
  year={2025},
  url={https://huggingface.co/Imran1/Med-R1-v1}
}