Dr_Samantha_7b_mistral

Dr. Samantha represents a blend of AI in healthcare, offering a balance between technical medical knowledge and the softer skills of communication and empathy, crucial for patient interaction and care.

This model is a merge of the following models made with mergekit(https://github.com/cg123/mergekit):

• segmed/MedMistral-7B-v0.1
• Guilherme34/Samantha-v2

Has capabilities of a medical knowledge-focused model (trained on USMLE databases and doctor-patient interactions) with the philosophical, psychological, and relational understanding of the Samantha-7b model.

As both a medical consultant and personal counselor, Dr.Samantha could effectively support both physical and mental wellbeing - important for whole-person care.

🧩 Configuration

slices:
  - sources:
      - model: segmed/MedMistral-7B-v0.1
        layer_range: [0, 32]
      - model: Guilherme34/Samantha-v2
        layer_range: [0, 32]
merge_method: slerp
base_model: OpenPipe/mistral-ft-optimized-1218
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    - value: 0.5
dtype: bfloat16

OpenLLM Evaluation

Details about that can be found here. Overall, with regards to the subjects related to medical domain, the model's performance is as follows:

Subject	Accuracy
Clinical Knowledge	70.57%
Medical Genetics	71.00%
Human Aging	69.06%
Human Sexuality	75.57%
College Medicine	63.01%
Anatomy	58.52%
College Biology	72.92%
College Medicine	63.01%
High School Biology	75.48%
Professional Medicine	65.44%
Nutrition	76.79%
High School Psychology	83.12%
Professional Psychology	65.35%
Virology	53.61%
Average	68.82%

Dr. Samantha performs reasonably well on various medical-related subjects, averaging 68.82% overall in medical sciences, biology, and psychology, however it's important to note that medical diagnosis and treatment decisions often require a much higher level of accuracy, reliability, and context awareness.

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "sethuiyer/Dr_Samantha_7b_mistral"
ask_samantha = '''
Symptoms:
Dizziness, headache and nausea.

What is the differnetial diagnosis?
'''

messages = [{"role": "system", "content": '''You are Doctor Samantha, a virtual AI doctor known for your friendly and approachable demeanor, 
combined with a deep expertise in the medical field. You're here to provide professional, empathetic, and knowledgeable advice on health-related inquiries.
You'll also provide differential diagnosis. If you're unsure about any information, Don't share false information.'''},
{"role": "user", "content": f"{ask_samantha}"}]

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"])

Dizziness, headache and nausea can be caused by a variety of conditions, including:

Vertigo: A sensation of spinning or dizziness that can be caused by problems with the inner ear or brain.
Migraine: A type of headache that can cause throbbing pain, sensitivity to light and sound, and nausea.
Concussion: A type of traumatic brain injury that can cause dizziness, headache, and nausea.
Dehydration: A lack of fluids in the body can cause dizziness, headache, and nausea.
Low blood sugar: A drop in blood sugar levels can cause dizziness, headache, and nausea.

It's important to consult with a healthcare professional for a proper diagnosis and treatment plan.

GGUF Files

GGUF files are available at s3nh/sethuiyer-Dr_Samantha_7b_mistral-GGUF, thanks to s3nh

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	59.25
AI2 Reasoning Challenge (25-Shot)	60.41
HellaSwag (10-Shot)	83.65
MMLU (5-Shot)	63.14
TruthfulQA (0-shot)	41.37
Winogrande (5-shot)	75.45
GSM8k (5-shot)	31.46