Instructions to use BlacqTangent/medgemma-4b-it-sft-lora-diff with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use BlacqTangent/medgemma-4b-it-sft-lora-diff with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("unsloth/medgemma-4b-it-unsloth-bnb-4bit")
model = PeftModel.from_pretrained(base_model, "BlacqTangent/medgemma-4b-it-sft-lora-diff")

Transformers

How to use BlacqTangent/medgemma-4b-it-sft-lora-diff with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="BlacqTangent/medgemma-4b-it-sft-lora-diff")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("BlacqTangent/medgemma-4b-it-sft-lora-diff", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use BlacqTangent/medgemma-4b-it-sft-lora-diff with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "BlacqTangent/medgemma-4b-it-sft-lora-diff"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "BlacqTangent/medgemma-4b-it-sft-lora-diff",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/BlacqTangent/medgemma-4b-it-sft-lora-diff

SGLang

How to use BlacqTangent/medgemma-4b-it-sft-lora-diff with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "BlacqTangent/medgemma-4b-it-sft-lora-diff" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "BlacqTangent/medgemma-4b-it-sft-lora-diff",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "BlacqTangent/medgemma-4b-it-sft-lora-diff" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "BlacqTangent/medgemma-4b-it-sft-lora-diff",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Unsloth Studio new

How to use BlacqTangent/medgemma-4b-it-sft-lora-diff with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for BlacqTangent/medgemma-4b-it-sft-lora-diff to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for BlacqTangent/medgemma-4b-it-sft-lora-diff to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for BlacqTangent/medgemma-4b-it-sft-lora-diff to start chatting

Load model with FastModel

pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
    model_name="BlacqTangent/medgemma-4b-it-sft-lora-diff",
    max_seq_length=2048,
)

Docker Model Runner
How to use BlacqTangent/medgemma-4b-it-sft-lora-diff with Docker Model Runner:
```
docker model run hf.co/BlacqTangent/medgemma-4b-it-sft-lora-diff
```

medgemma-4b-it-sft-lora-diff

This model is a fine-tuned version of unsloth/medgemma-4b-it-unsloth-bnb-4bit on the imagefolder dataset. It achieves the following results on the evaluation set:

Loss: 0.1144

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0002
train_batch_size: 2
eval_batch_size: 2
seed: 42
gradient_accumulation_steps: 16
total_train_batch_size: 32
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.03
num_epochs: 7
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.4505	0.1333	2	6.9109
0.413	0.2667	4	5.8505
0.3524	0.4	6	5.0933
0.3067	0.5333	8	4.4616
0.268	0.6667	10	3.9022
0.2333	0.8	12	3.3732
0.2006	0.9333	14	2.8276
0.1669	1.0667	16	2.3019
0.1349	1.2	18	1.8597
0.1078	1.3333	20	1.5445
0.0907	1.4667	22	1.3199
0.0786	1.6	24	1.1433
0.0679	1.7333	26	0.9538
0.0559	1.8667	28	0.8040
0.0488	2.0	30	0.7037
0.0431	2.1333	32	0.6440
0.0394	2.2667	34	0.6086
0.0372	2.4	36	0.5814
0.036	2.5333	38	0.5506
0.0335	2.6667	40	0.5160
0.0315	2.8	42	0.4795
0.0293	2.9333	44	0.4408
0.0267	3.0667	46	0.3978
0.0241	3.2	48	0.3642
0.0226	3.3333	50	0.3685
0.0234	3.4667	52	0.3769
0.0232	3.6	54	0.3629
0.0223	3.7333	56	0.3403
0.0209	3.8667	58	0.3231
0.0199	4.0	60	0.3149
0.0194	4.1333	62	0.3050
0.0187	4.2667	64	0.2897
0.0177	4.4	66	0.2734
0.0167	4.5333	68	0.2596
0.0158	4.6667	70	0.2453
0.015	4.8	72	0.2285
0.0139	4.9333	74	0.2105
0.0127	5.0667	76	0.1926
0.0116	5.2	78	0.1769
0.0107	5.3333	80	0.1634
0.0099	5.4667	82	0.1505
0.0091	5.6	84	0.1393
0.0083	5.7333	86	0.1307
0.008	5.8667	88	0.1248
0.0077	6.0	90	0.1208
0.0074	6.1333	92	0.1186
0.0073	6.2667	94	0.1171
0.0071	6.4	96	0.1160
0.0071	6.5333	98	0.1154
0.007	6.6667	100	0.1147
0.0071	6.8	102	0.1146
0.007	6.9333	104	0.1144

Framework versions

PEFT 0.16.0
Transformers 4.57.1
Pytorch 2.6.0+cu124
Datasets 4.3.0
Tokenizers 0.22.1

Downloads last month: -

Model tree for BlacqTangent/medgemma-4b-it-sft-lora-diff

Base model

google/gemma-3-4b-pt

Finetuned

google/medgemma-4b-pt

Finetuned

google/medgemma-4b-it

Quantized

unsloth/medgemma-4b-it-unsloth-bnb-4bit

Adapter

(19)

this model