!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "Reverb/MedLLaMA-3"
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"])

🏆 Evaluation

Tasks	Version	Filter	Metric	Value		Stderr
stem	N/A	none	acc	0.6466	±	0.0056
		none	acc_norm	0.6124	±	0.0066
- medmcqa	Yaml	none	acc	0.6118	±	0.0075
		none	acc_norm	0.6118	±	0.0075
- medqa_4options	Yaml	none	acc	0.6143	±	0.0136
		none	acc_norm	0.6143	±	0.0136
- anatomy (mmlu)	0	none	acc	0.7185	±	0.0389
- clinical_knowledge (mmlu)	0	none	acc	0.7811	±	0.0254
- college_biology (mmlu)	0	none	acc	0.8264	±	0.0317
- college_medicine (mmlu)	0	none	acc	0.7110	±	0.0346
- medical_genetics (mmlu)	0	none	acc	0.8300	±	0.0378
- professional_medicine (mmlu)	0	none	acc	0.7868	±	0.0249
- pubmedqa	1	none	acc	0.7420	±	0.0196

Groups	Version	Filter	n-shot	Metric	Value		Stderr
stem	N/A	none	0	acc	0.6466	±	0.0056
		none	0	acc_norm	0.6124	±	0.0066