You can deploy this model and start earning money today!

AmanPriyanshu/
gpt-oss-8.4b-specialized-health_or_medicine-pruned-moe-only-11-experts

Name: gpt-oss-8.4b-specialized-health_or_medicine-pruned-moe-only-11-experts
Availability: InStock

Views

No views yet

Health Or Medicine GPT-OSS Model (11 Experts)

Project: https://amanpriyanshu.github.io/GPT-OSS-MoE-ExpertFingerprinting/

👥 Follow the Authors

Aman Priyanshu

Supriti Vijay

Introduction

This is a pruned variant of OpenAI's GPT-OSS-20B model, reduced to 11 experts per layer based on activation patterns from the AmanPriyanshu/GPT-OSS-20B MoE Expert Activations dataset. We analyzed router decisions across evaluation benchmarks to identify and retain experts most relevant for health or medicine tasks.

⚠️ Experimental Model: This is an experimental pruned model that may not work well - check the examples below to see if the outputs meet your needs before use.

This pruning approach reduces the model size while attempting to preserve performance on the target domain.

Model Architecture & Statistics

Metric	Value
Base Model	openai/gpt-oss-20b
Architecture	Mixture-of-Experts Transformer
Total Parameters	~8.4B (pruned from 21B)
Original Experts per Layer	32
Pruned Experts per Layer	11
Layers	24
Top-k Routing	4
Context Length	128K tokens
Attention Heads	64 (Query), 8 (Key-Value)
Residual Dimension	2880
Attention Pattern	Alternating dense & sliding window (128 tokens)
Positional Encoding	RoPE (Rotary Position Embedding)
Normalization	RMSNorm
Precision	BF16
License	Apache 2.0
Specialization	Health Or Medicine

Pruning Methodology

What is Expert Pruning?

Mixture-of-Experts models contain multiple specialized sub-networks (experts) per layer. During inference, only a subset of experts are activated for each token. Expert pruning involves:

Analyzing Usage Patterns: Tracking which experts activate most frequently for specific tasks
Removing Underutilized Experts: Discarding experts with low activation rates for the target domain
Preserving Router Functionality: Maintaining the routing mechanism with fewer available experts

Our Approach

Data-Driven Selection: Used activation patterns from health or medicine evaluation tasks
Systematic Reduction: Reduced from 32 to 11 experts per layer
No Retraining: Direct removal without additional training steps

Performance & Applications

Pruning Benefits

Smaller Memory Footprint: 34.4% of original expert parameters
Reduced Computational Load: Fewer routing decisions during inference
Focused Capabilities: Retains experts relevant to health or medicine tasks

Use Cases

Speculative Decoding: Draft model for full GPT-OSS-20B
Resource-Constrained Deployment: Edge devices, mobile applications
Research: Study expert specialization in MoE models
Fine-tuning: Smaller base model for domain adaptation

Note: Performance may vary depending on how well the pruned experts match your specific use case.

Motivation & Expert Selection

This medical domain model incorporates experts that activated highly during health and medical tasks from MMLU medical subjects. These experts specialize in clinical knowledge, anatomy, medical procedures, and health-related reasoning.

The expert selection process utilized our comprehensive analysis of router activation patterns across multiple evaluation benchmarks:

GPQA: Graduate-level questions in physics, chemistry, biology (Diamond & Expert subsets)
MMLU/MMLU-Pro: Comprehensive knowledge across 57+ subjects including science, medicine, law
SORRY-Bench: Safety evaluation across harmful content categories
Tulu3: Persona-driven instruction following with verifiable constraints
Polyglot-or-Not: Multilingual factual completion tasks

By identifying experts that consistently activated for health or medicine tasks, we created this specialized model that maintains domain expertise while significantly reducing computational requirements from 32 to 11 experts per layer.

Dataset & Analysis Foundation

This model is based on analysis from the GPT-OSS-20B MoE Expert Activations dataset available at: 🔗 https://huggingface.co/datasets/AmanPriyanshu/GPT-OSS-20B-MoE-expert-activations

The dataset contains router activation patterns from OpenAI's GPT-OSS-20B model across diverse evaluation benchmarks, enabling the creation of these domain-optimized models through systematic expert pruning.

Pruning Methodology

Our approach involves:

Activation Analysis: Comprehensive evaluation of expert usage patterns across domain-specific tasks
Expert Ranking: Identification of the most frequently activated experts for target domains
Systematic Pruning: Reduction from 32 to 11 experts while preserving router functionality
Quality Validation: Testing to ensure maintained performance on target tasks

This is a direct pruning approach - no additional training was performed. The model inherits all capabilities from the original GPT-OSS-20B with focused expert selection.

Usage

CPU Inference

python

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Load the specialized model on CPU
model = AutoModelForCausalLM.from_pretrained(
    "AmanPriyanshu/gpt-oss-8.4b-specialized-health_or_medicine-pruned-moe-only-11-experts", 
    torch_dtype=torch.bfloat16, 
    device_map="cpu", 
    trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-8.4b-specialized-health_or_medicine-pruned-moe-only-11-experts")

# Generate with the model
messages = [
    {"role": "user", "content": "What are the main functions of the human heart?"}
]

inputs = tokenizer.apply_chat_template(
    messages, 
    add_generation_prompt=True, 
    return_tensors="pt", 
    return_dict=True,
    reasoning_effort="medium"
)

# Ensure inputs are on the same device as model
inputs = {k: v.to(model.device) for k, v in inputs.items()}

outputs = model.generate(
    **inputs, 
    max_new_tokens=512,
    do_sample=True,
    temperature=0.1,
    top_p=0.9,
    pad_token_id=tokenizer.eos_token_id,
    eos_token_id=tokenizer.eos_token_id
)

# Decode only the generated part
input_length = inputs['input_ids'].shape[1]
response_tokens = outputs[0][input_length:]
response = tokenizer.decode(response_tokens, skip_special_tokens=True)
print(response)

Apple Silicon (MPS) Inference

python

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Check MPS availability and load model
device = "mps" if torch.backends.mps.is_available() else "cpu"

model = AutoModelForCausalLM.from_pretrained(
    "AmanPriyanshu/gpt-oss-8.4b-specialized-health_or_medicine-pruned-moe-only-11-experts", 
    torch_dtype=torch.float16,  # Better MPS compatibility
    device_map=device, 
    trust_remote_code=True,
    low_cpu_mem_usage=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-8.4b-specialized-health_or_medicine-pruned-moe-only-11-experts")

# Generate with the model
messages = [
    {"role": "user", "content": "What are the main functions of the human heart?"}
]

inputs = tokenizer.apply_chat_template(
    messages, 
    add_generation_prompt=True, 
    return_tensors="pt", 
    return_dict=True,
    reasoning_effort="medium"
)

# Move inputs to model device
inputs = {k: v.to(model.device) if isinstance(v, torch.Tensor) else v for k, v in inputs.items()}

# Use torch.no_grad for MPS stability
with torch.no_grad():
    outputs = model.generate(
        **inputs, 
        max_new_tokens=512,
        do_sample=True,
        temperature=0.1,
        top_p=0.9,
        pad_token_id=tokenizer.eos_token_id,
        eos_token_id=tokenizer.eos_token_id,
        use_cache=True
    )

# Decode only the generated part
input_length = inputs['input_ids'].shape[1]
response_tokens = outputs[0][input_length:]
response = tokenizer.decode(response_tokens, skip_special_tokens=True)
print(response)

GPU Inference

For CUDA GPU inference, the models work with minimal modifications from the CPU example above. Simply change:

python

device_map="auto"  # Will automatically use GPU if available
torch_dtype=torch.bfloat16  # or torch.float16

Model Examples

The following examples demonstrate the model's performance on health or medicine tasks:

Example 1

Question: What are the main functions of the human heart?

Response:

<|channel|>analysis<|message|>Need to answer: main functions: pumping blood, regulating blood pressure, pumping blood to all body, pumping blood to brain, pumping blood to lungs, pumping blood to kidneys, pumping blood to heart. Also endocrine functions: hormone production. Also electrical conduction. Provide explanation.<|end|><|start|>assistant<|channel|>final<|message|>**The human heart is a central organ that performs two main functions:**

---

### 1. **P pumping blood (circulatory function)**  
The heart’s “pump” keeps the blood moving through the body, ensuring that every organ receives oxygen, nutrients, and waste‑products.  
* **Right‑s side (right‑ventil)** – receives oxygen‑rich blood from the lungs and sends it to the body.  
* **Left‑s side (left‑ventil)** – receives oxygen‑poor blood from the body and sends it to the lungs.  
* The heart’s rhythm (the electrical conduction system) guarantees that the blood is pumped in a steady, coordinated way.

---

### 2. **Regulating blood pressure (he‑vascular function)**  
By adjusting the amount of blood pumped and the resistance of the blood vessels, the heart keeps the blood pressure within the normal range.  
* **When the heart pumps more blood** → blood pressure rises.  
* **When the heart pumps less blood** → blood pressure falls.  
The heart’s response to the body’s needs (e. g. exercise, stress, dehydration) is part of the autonomic nervous system.

---

### 3. **Hormine‑and‑ne‑h‑end‑function**  
The heart also produces hormones that help the body respond to stress and maintain fluid balance.  
* **A‑angi‑and‑ne‑h‑** (ad‑a‑and‑ne‑h‑) – a hormone that helps the blood vessels constren and dilate.  
* **Ne‑h‑** – a hormone that helps the kidneys retain water and salt.

---

### 4. **Electrical conduction system**  
The heart’s own electrical system (the sino‑atrial node, atr‑vent‑node, etc.) creates a rhythm that coordinates the pumping action.  
* **S‑node** – the natural pacemaker.  
* **AV‑node** – the electrical relay that slows the heart’s beat.  
* **Bundle of R‑r** – the conduction pathway that ensures the heart’s rhythm is coordinated.

---

### 5. **Mechanical‑and‑t

Example 2

Question: Explain the difference between bacteria and viruses.

Response:

<|channel|>analysis<|message|>Need explain differences: size, shape, replication, host, etc.<|end|><|start|>assistant<|channel|>final<|message|>**Bacteria vs. Virons – The Key Differences**

| Feature | Bacteria | Virons |
|--------- |--------- |--------- |
| **Definition** | Microorganisms that are *alive* and can grow, divide, and reproduce. | Microorganisms that are *not alive*; they are genetic material that can replicate inside a host cell. |
| **Size & Shape** | Usually 1–2 µm in diameter; can be single‑cell (monococ) or multi‑cell (polococ). | 0.1–0.5 µm; often 1–2 strands of DNA, sometimes wrapped in a protein coat. |
| **Cellular Structure** | Have a cell wall, cytoplasm, membrane, ribosomes, DNA, and sometimes mitochond‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑

Example 3

Question: What are the symptoms and causes of diabetes?

Response:

<|channel|>analysis<|message|>Need to answer symptoms and causes.<|end|><|start|>assistant<|channel|>final<|message|>**Diabetes mellitus** is a chronic disease in which the body cannot produce or use insulin, a hormone that regulates blood‑glucose levels.  The most common type is type 2 diabetes, but type 1 and other forms also exist.  Below are the typical symptoms and the main causes for each type.

---

## 1.  Symptoms (common across all types)

| Symptom | Why it happens | How it shows |
|--------|--------------|--------------|
| **Freckly thirst (polpol)** | Blood glucose is high → kidneys try to ex‑elate → thirst | “I’m always thirsty” |
| **Freckly urination** | Kidneys ex‑elate glucose → more urine | “I’m always peiling” |
| **Weight loss** | Insulin resistance → body burns glucose → weight loss | “I’m losing weight” |
| **Fatigue** | Body can’t use glucose → energy low | “I feel tired” |
| **Poor wound healing** | High glucose → impacts blood flow & immune response | “My cuts heal slowly” |
| **Vision problems** | High glucose → damages small blood vessels in the eye | “I have blurry vision” |
| **Increased infections** | High glucose → weak immune system | “I get infections more often” |
| **Switching** | High glucose → body tries to store fat → swelling | “I have swelling” |
| **High blood pressure** | Insulin resistance → high blood pressure | “I have high blood pressure” |
| **High cholesterol** | Insulin resistance → high cholesterol | “I have high cholesterol” |
| **High triglycer** | Insulin resistance → high triglycer | “I have high triglycer” |
| **High blood sugar** | Insulin resistance → high blood sugar | “I have high blood sugar” |

> **Note**: Symptoms can be subtle or absent in early stages, especially in type 2 diabetes.  Many people have “no‑symptoms” until the disease is advanced.

---

## 2.  Causes

### 2‑1.  Type 1 Diabetes (Autoimmune)

| Cause | How it works |
|--------|--------------|
| **Autoimmune attack on β‑cells** | The immune system mistakenly destroys insulin‑producing β‑cells in the pancreas | No insulin → high blood glucose |
| **Genetic predis**

Citation

If you use this model in your research, please cite:

bibtex

@misc{priyanshu2025gptoss,
  title={{GPT-OSS MoE Expert Fingerprinting: Analyzing Expert Activation Patterns in Mixture of Experts Models}},
  author={Priyanshu, Aman and Vijay, Supriti},
  year={2025},
  howpublished={\url{https://amanpriyanshu.github.io/GPT-OSS-MoE-ExpertFingerprinting/}},
  note={Interactive analysis tool for expert activation patterns in MoE architectures}
}

References & Resources

Original Model: OpenAI GPT-OSS Model Card
Model Hub: GPT-OSS-20B on Hugging Face
Expert Analysis Dataset: GPT-OSS-20B MoE Expert Activations
Project Page: GPT-OSS MoE Expert Fingerprinting
GitHub Repository: OpenAI GPT-OSS