README.md

---
license: apache-2.0
tags:
- moe
- merge
- AdaptLLM/medicine-chat
- microsoft/Orca-2-7b
datasets:
- open-llm-leaderboard/details_Technoculture__Medchator-2x7b
model-index:
- name: Medchator-2x7b
  results:
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: AI2 Reasoning Challenge (25-Shot)
      type: ai2_arc
      config: ARC-Challenge
      split: test
      args:
        num_few_shot: 25
    metrics:
    - type: acc_norm
      value: 57.59
      name: normalized accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Technoculture/Medchator-2x7b
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: HellaSwag (10-Shot)
      type: hellaswag
      split: validation
      args:
        num_few_shot: 10
    metrics:
    - type: acc_norm
      value: 78.14
      name: normalized accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Technoculture/Medchator-2x7b
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: MMLU (5-Shot)
      type: cais/mmlu
      config: all
      split: test
      args:
        num_few_shot: 5
    metrics:
    - type: acc
      value: 56.13
      name: accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Technoculture/Medchator-2x7b
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: TruthfulQA (0-shot)
      type: truthful_qa
      config: multiple_choice
      split: validation
      args:
        num_few_shot: 0
    metrics:
    - type: mc2
      value: 48.77
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Technoculture/Medchator-2x7b
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: Winogrande (5-shot)
      type: winogrande
      config: winogrande_xl
      split: validation
      args:
        num_few_shot: 5
    metrics:
    - type: acc
      value: 75.3
      name: accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Technoculture/Medchator-2x7b
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: GSM8k (5-shot)
      type: gsm8k
      config: main
      split: test
      args:
        num_few_shot: 5
    metrics:
    - type: acc
      value: 32.83
      name: accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Technoculture/Medchator-2x7b
      name: Open LLM Leaderboard
---

# Medchator-2x7b

Medchator-2x7b is a Mixure of Experts (MoE) made with the following models:
* [AdaptLLM/medicine-chat](https://huggingface.co/AdaptLLM/medicine-chat)
* [microsoft/Orca-2-7b](https://huggingface.co/microsoft/Orca-2-7b)

## Evaluations

# Open LLM Leaderboard

![image/png](https://cdn-uploads.huggingface.co/production/uploads/63486df1f8f01fcc4b23e97d/ZSMRhGuLrE-K1WNlfbDAG.png)

| Model Name         | ARC      | HellaSwag | MMLU     | TruthfulQA | Winogrande | GSM8K    |
| ------------------ | -------- | --------- | -------- | ---------- | ---------- | -------- |
| Orca-2-7b          | **78.4** | 76.1      | 53.7     | **52.4**   | 74.2       | **47.2** |
| LLAMA-2-7b         | 43.2     | 77.1      | 44.4     | 38.7       | 69.5       | 16       |
| MT7Bi-sft          | 54.1     | 75.11     | -        | 43.08      | 72.14      | 15.54    |
| MT7bi-dpo	         | 54.69    | 75.89     | 52.82    | 45.48      | 71.58      | 25.93    |
| Medorca-2x7b       | 54.1     | 76.04     | 54.1     | 48.04      | 74.51      | 20.64    |
| Medchator-2x7b     | **57.59**| **78.14** | **56.13**| **48.77**  | **75.3**   | **32.83**|

## Medical Performance

Clinical Camel demonstrates competitive performance on medical benchmarks.

**Table: Five-Shot Performance of GPT3.5, llama-2-7b and Llama-2-70b on Various Medical Datasets**

| Dataset                    | Medchator-2x7b | GPT3.5 | Llama-2 7b | Llama-2 70b |
|----------------------------|----------------|--------|------------|-------------|
| MMLU Anatomy               | 56.3           | 60.7   | 48.9       | 62.9        |
| MMLU Clinical Knowledge    | 63.0           | 68.7   | 46.0       | 71.7        |
| MMLU College Biology       | 63.8           | 72.9   | 47.2       | 84.7        |
| MMLU College Medicine      | 50.9           | 63.6   | 42.8       | 64.2        |
| MMLU Medical Genetics      | 67.0           | 68.0   | 55.0       | 74.0        |
| MMLU Professional Medicine | 55.1           | 69.8   | 53.6       | 75.0        |

## 🧩 Configuration

```yaml
base_model: microsoft/Orca-2-7b
gate_mode: hidden
dtype: bfloat16
experts:
  - source_model: AdaptLLM/medicine-chat
    positive_prompts: 
      - "How does sleep affect cardiovascular health?"
      - "Could a plant-based diet improve arthritis symptoms?"
      - "A patient comes in with symptoms of dizziness and nausea"
      - "When discussing diabetes management, the key factors to consider are"
      - "The differential diagnosis for a headache with visual aura could include"
    negative_prompts:
      - "Recommend a good recipe for a vegetarian lasagna."
      - "Give an overview of the French Revolution."
      - "Explain how a digital camera captures an image."
      - "What are the environmental impacts of deforestation?"
      - "The recent advancements in artificial intelligence have led to developments in"
      - "The fundamental concepts in economics include ideas like supply and demand, which explain"
  - source_model: microsoft/Orca-2-7b
    positive_prompts:
      - "Here is a funny joke for you -"
      - "When considering the ethical implications of artificial intelligence, one must take into account"
      - "In strategic planning, a company must analyze its strengths and weaknesses, which involves"
      - "Understanding consumer behavior in marketing requires considering factors like"
      - "The debate on climate change solutions hinges on arguments that"
    negative_prompts:
      - "In discussing dietary adjustments for managing hypertension, it's crucial to emphasize"
      - "For early detection of melanoma, dermatologists recommend that patients regularly check their skin for"
      - "Explaining the importance of vaccination, a healthcare professional should highlight"
```

## 💻 Usage

```python
!pip install -qU transformers bitsandbytes accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "Technoculture/Medchator-2x7b"

tokenizer = AutoTokenizer.from_pretrained(model)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    model_kwargs={"torch_dtype": torch.float16, "load_in_4bit": True},
)

messages = [{"role": "user", "content": "Explain what a Mixture of Experts is in less than 100 words."}]
prompt = pipeline.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
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"])
```
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_Technoculture__Medchator-2x7b)

|             Metric              |Value|
|---------------------------------|----:|
|Avg.                             |58.13|
|AI2 Reasoning Challenge (25-Shot)|57.59|
|HellaSwag (10-Shot)              |78.14|
|MMLU (5-Shot)                    |56.13|
|TruthfulQA (0-shot)              |48.77|
|Winogrande (5-shot)              |75.30|
|GSM8k (5-shot)                   |32.83|