GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
convaiinnovations
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
convaiinnovations
Model Tree
google/medgemma-4b-it
Input Modalities
Output Modalities
Supported Functionality
This is a fully merged fine-tuned version of google/medgemma-4b-it trained on the PULSE-ECG/ECGInstruct dataset containing 1.15M ECG instruction-following examples. The LoRA adapter has been merged into the base model for easier deployment.
Developed by: ConvAI Innovations
Base Model: google/medgemma-4b-it
Training Infrastructure: AIRAWAT (C-DAC) - 8x NVIDIA A100 40GB GPUs
Training Duration: 72 hours (3 days)
Final Token Accuracy: 86.83%
Final Training Loss: 0.6188
GPU-Hours: 576
Model Size: ~8.5 GB
Hardware:
Training Configuration:
Training Metrics:
bash
pip install transformers pillow torchpython
from transformers import AutoModelForImageTextToText, AutoProcessorfrom PIL import Image # Load model and processormodel_id = "convaiinnovations/medgemma-4b-ecginstruct"model = AutoModelForImageTextToText.from_pretrained( model_id, torch_dtype="auto", device_map="auto")processor = AutoProcessor.from_pretrained(model_id)python
# Load ECG imageimage = Image.open("ecg_image.png").convert("RGB") # Prepare prompt using chat templatemessages = [ { "role": "user", "content": [ {"type": "image"}, {"type": "text", "text": "Analyze this ECG and provide a detailed interpretation."} ] }] # Process inputtext = processor.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)inputs = processor(text=[text], images=[[image]], return_tensors="pt", padding=True)inputs = {k: v.to(model.device) for k, v in inputs.items()} # Generate interpretationoutputs = model.generate( **inputs, max_new_tokens=512, do_sample=False) # Decode and printresponse = processor.decode(outputs[0], skip_special_tokens=True)print(response)python
# Detailed interpretation"Analyze this ECG and provide a detailed interpretation." # Specific abnormality detection"What abnormalities are present in this ECG?" # Diagnosis suggestion"Based on this ECG, what is the most likely diagnosis?" # Question answering"Does this ECG show signs of atrial fibrillation?" # Rate and rhythm"What is the heart rate and rhythm in this ECG?"This model can:
Training Metrics:
| Metric | Value |
|---|---|
| Token Accuracy | 86.83% |
| Final Loss | 0.6188 |
| Training Time | 72 hours |
| GPU-Hours | 576 |
Inference Metrics (A100 GPU):
| Metric | Value |
|---|---|
| TTFT (Time to First Token) | ~150ms |
| ISL (Input Sequence Length) | 2048 tokens |
| OSL (Output Sequence Length) | 512 tokens |
| End-to-End Latency | 2-3 seconds |
| Throughput | ~45 tokens/sec |
⚠️ MEDICAL DISCLAIMER
This model is for RESEARCH AND EDUCATIONAL PURPOSES ONLY.
- ❌ NOT validated for clinical use
- ❌ NOT FDA/CE approved
- ❌ NOT a substitute for professional medical diagnosis
- ❌ Should NOT be used for patient care decisions
Always consult qualified healthcare professionals for medical decisions.
Important Notes:
Appropriate Uses:
Inappropriate Uses:
If you use this model in your research, please cite:
bibtex
@misc{medgemma-ecginstruct, author = {convaiinnovations}, title = {MedGemma-4B ECGInstruct: Fine-tuned ECG Interpretation Model}, year = {2025}, publisher = {HuggingFace}, howpublished = {\url{https://huggingface.co/convaiinnovations/medgemma-4b-ecginstruct}}}Apache 2.0 (following base model license)
For questions or issues, please open an issue on the model repository or contact the maintainers.
from transformers import AutoModelForImageTextToText, AutoProcessorfrom PIL import Image # Load model and processormodel_id = "convaiinnovations/medgemma-4b-ecginstruct"model = AutoModelForImageTextToText.from_pretrained( model_id, torch_dtype="auto", device_map="auto")processor = AutoProcessor.from_pretrained(model_id)# Load ECG imageimage = Image.open("ecg_image.png").convert("RGB") # Prepare prompt using chat templatemessages = [ { "role": "user", "content": [ {"type": "image"}, {"type": "text", "text": "Analyze this ECG and provide a detailed interpretation."} ] }] # Process inputtext = processor.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)inputs = processor(text=[text], images=[[image]], return_tensors="pt", padding=True)inputs = {k: v.to(model.device) for k, v in inputs.items()} # Generate interpretationoutputs = model.generate( **inputs, max_new_tokens=512, do_sample=False) # Decode and printresponse = processor.decode(outputs[0], skip_special_tokens=True)print(response)# Detailed interpretation"Analyze this ECG and provide a detailed interpretation." # Specific abnormality detection"What abnormalities are present in this ECG?" # Diagnosis suggestion"Based on this ECG, what is the most likely diagnosis?" # Question answering"Does this ECG show signs of atrial fibrillation?" # Rate and rhythm"What is the heart rate and rhythm in this ECG?"@misc{medgemma-ecginstruct, author = {convaiinnovations}, title = {MedGemma-4B ECGInstruct: Fine-tuned ECG Interpretation Model}, year = {2025}, publisher = {HuggingFace}, howpublished = {\url{https://huggingface.co/convaiinnovations/medgemma-4b-ecginstruct}}}