JSL-MedMX-7X
This model is developed by John Snow Labs. Performance on biomedical benchmarks: Open Medical LLM Leaderboard.
This model is available under a CC-BY-NC-ND license and must also conform to this Acceptable Use Policy. If you need to license this model for commercial use, please contact us at [email protected].
π» Usage
!pip install -qU transformers accelerate
from transformers import AutoTokenizer
import transformers
import torch
model = "johnsnowlabs/JSL-MedMX-7X"
messages = [{"role": "user", "content": "What is a large language model?"}]
tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
pipeline = transformers.pipeline(
"text-generation",
model=model,
torch_dtype=torch.float16,
device_map="auto",
)
outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])
π Evaluation
Tasks | Version | Filter | n-shot | Metric | Value | Stderr | |
---|---|---|---|---|---|---|---|
stem | N/A | none | 0 | acc_norm | 0.5783 | Β± | 0.0067 |
none | 0 | acc | 0.6177 | Β± | 0.0057 | ||
- medmcqa | Yaml | none | 0 | acc | 0.5668 | Β± | 0.0077 |
none | 0 | acc_norm | 0.5668 | Β± | 0.0077 | ||
- medqa_4options | Yaml | none | 0 | acc | 0.6159 | Β± | 0.0136 |
none | 0 | acc_norm | 0.6159 | Β± | 0.0136 | ||
- anatomy (mmlu) | 0 | none | 0 | acc | 0.7111 | Β± | 0.0392 |
- clinical_knowledge (mmlu) | 0 | none | 0 | acc | 0.7396 | Β± | 0.0270 |
- college_biology (mmlu) | 0 | none | 0 | acc | 0.7778 | Β± | 0.0348 |
- college_medicine (mmlu) | 0 | none | 0 | acc | 0.6647 | Β± | 0.0360 |
- medical_genetics (mmlu) | 0 | none | 0 | acc | 0.7200 | Β± | 0.0451 |
- professional_medicine (mmlu) | 0 | none | 0 | acc | 0.7868 | Β± | 0.0249 |
- pubmedqa | 1 | none | 0 | acc | 0.7840 | Β± | 0.0184 |
- Downloads last month
- 24
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.