GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
AmareshHebbar
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
AmareshHebbar
Model Tree
unsloth/Qwen2.5-3B-Instruct
Input Modalities
Output Modalities
Supported Functionality
| Base model | unsloth/Qwen2.5-3B-Instruct |
| Fine-tuning method | QLoRA (4-bit NF4), rank 16 |
| Training framework | Unsloth + TRL SFTTrainer |
| Training data | AmareshHebbar/radiology-coder-sft — 25,090 rows |
| Hardware | NVIDIA A40 (48GB), single GPU |
| License | Apache 2.0 |
Automate radiology coding workflows, integrate into RIS/PACS systems, or build a report-to-code pipeline for teleradiology platforms.
This model is not a substitute for a certified medical professional's judgment. Output should be reviewed by a qualified person before being used in a clinical or billing decision. The model can make mistakes, especially on rare or compound cases.
python
from transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModelimport torch base_model = "unsloth/Qwen2.5-3B-Instruct"adapter = "AmareshHebbar/radiology-coder-qwen25-3b" tokenizer = AutoTokenizer.from_pretrained(base_model)model = AutoModelForCausalLM.from_pretrained( base_model, torch_dtype=torch.bfloat16, device_map="auto",)model = PeftModel.from_pretrained(model, adapter) messages = [ {"role": "system", "content": "You are a radiology coding specialist. Given a radiology report impression, return the ICD-10-CM codes for all findings."}, {"role": "user", "content": "IMPRESSION: 1.8cm hypoechoic nodule right thyroid lobe, TIRADS 4. Recommend FNA."},]inputs = tokenizer.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True).to(model.device)outputs = model.generate(inputs, max_new_tokens=128, temperature=0.1, do_sample=True)print(tokenizer.decode(outputs[0][inputs.shape[1]:], skip_special_tokens=True))Expected output:
text
ICD-10 Codes:\n1. E04.1 — Non-toxic single thyroid nodule (primary finding)\nRecommend: FNA if TIRADS ≥4. Add malignancy code post-biopsy.python
from unsloth import FastLanguageModel model, tokenizer = FastLanguageModel.from_pretrained( model_name="AmareshHebbar/radiology-coder-qwen25-3b", max_seq_length=512, load_in_4bit=True,)FastLanguageModel.for_inference(model) messages = [ {"role": "system", "content": "You are a radiology coding specialist. Given a radiology report impression, return the ICD-10-CM codes for all findings."}, {"role": "user", "content": "IMPRESSION: Acute pulmonary embolism involving bilateral main pulmonary arteries. No right heart strain."},]prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)inputs = tokenizer(prompt, return_tensors="pt").to("cuda")outputs = model.generate(**inputs, max_new_tokens=128, temperature=0.1, do_sample=True)print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))bash
vllm serve unsloth/Qwen2.5-3B-Instruct \ --enable-lora \ --lora-modules radiology-coder-qwen25-3b=AmareshHebbar/radiology-coder-qwen25-3b \ --host 0.0.0.0 --port 8000 --dtype bfloat16python
from openai import OpenAI client = OpenAI(base_url="http://localhost:8000/v1", api_key="not-needed")response = client.chat.completions.create( model="radiology-coder-qwen25-3b", messages=[ {"role": "system", "content": "You are a radiology coding specialist. Given a radiology report impression, return the ICD-10-CM codes for all findings."}, {"role": "user", "content": "IMPRESSION: 3.2cm hypodense lesion right hepatic lobe, arterial enhancement with washout, consistent with hepatocellular carcinoma."}, ], temperature=0.1,)print(response.choices[0].message.content)Trained on 25,090 examples extracted from 25k radiology-relevant clinical notes filtered by imaging keywords. No synthetic or LLM-generated training data — every example pairs real-world input with its authoritative output.
See the dataset card for the full extraction pipeline.
| Parameter | Value |
|---|---|
| LoRA rank | 16 |
| LoRA alpha | 32 |
| LoRA dropout | 0 |
| Target modules | q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj |
| Quantization | 4-bit NF4 (QLoRA) |
| Max sequence length | 512 |
| Optimizer | paged_adamw_8bit |
| Learning rate | 2e-4, cosine schedule |
Fine-tuned with Unsloth for 2x faster training and reduced VRAM, using TRL's SFTTrainer. Training run on a single NVIDIA A40 GPU. Experiment tracking via Weights & Biases.
| Model | Task | Size |
|---|---|---|
| icd10-coder-qwen25-7b | ICD-10-CM medical coding | 7B |
| snomed-mapper-qwen25-7b | Clinical concept mapping | 7B |
| icd10-to-drg-qwen25-1b | ICD-10 → DRG reimbursement | 1.5B |
| pmjay-classifier-qwen25-3b |
Full collection: [link your HF collection here]
bibtex
@misc{medicalai2026, author = {Hebbar, Amaresh}, title = {Medical AI Fine-tuning Suite}, year = {2026}, publisher = {HuggingFace}, url = {https://huggingface.co/AmareshHebbar}}vllm serve unsloth/Qwen2.5-3B-Instruct \ --enable-lora \ --lora-modules radiology-coder-qwen25-3b=AmareshHebbar/radiology-coder-qwen25-3b \ --host 0.0.0.0 --port 8000 --dtype bfloat16@misc{medicalai2026, author = {Hebbar, Amaresh}, title = {Medical AI Fine-tuning Suite}, year = {2026}, publisher = {HuggingFace}, url = {https://huggingface.co/AmareshHebbar}}| India PM-JAY classification |
| 3B |