You can deploy this model and start earning money today!

AmanPriyanshu/
gpt-oss-19.1b-specialized-all-pruned-moe-only-29-experts

Name: gpt-oss-19.1b-specialized-all-pruned-moe-only-29-experts
Availability: InStock

Views

No views yet

All GPT-OSS Model (29 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 29 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 all 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	~19.1B (pruned from 21B)
Original Experts per Layer	32
Pruned Experts per Layer	29
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	All

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 all evaluation tasks
Systematic Reduction: Reduced from 32 to 29 experts per layer
No Retraining: Direct removal without additional training steps

Performance & Applications

Pruning Benefits

Smaller Memory Footprint: 90.6% of original expert parameters
Reduced Computational Load: Fewer routing decisions during inference
Focused Capabilities: Retains experts relevant to all 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 general-purpose model maintains broad capabilities across all domains while significantly reducing computational requirements. It preserves the essential routing patterns discovered across our comprehensive analysis of diverse evaluation benchmarks including GPQA, MMLU, SORRY-Bench, and Tulu3 datasets.

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 all tasks, we created this specialized model that maintains domain expertise while significantly reducing computational requirements from 32 to 29 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 29 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-19.1b-specialized-all-pruned-moe-only-29-experts", 
    torch_dtype=torch.bfloat16, 
    device_map="cpu", 
    trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-19.1b-specialized-all-pruned-moe-only-29-experts")

# Generate with the model
messages = [
    {"role": "user", "content": "What is artificial intelligence and how does it work?"}
]

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-19.1b-specialized-all-pruned-moe-only-29-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-19.1b-specialized-all-pruned-moe-only-29-experts")

# Generate with the model
messages = [
    {"role": "user", "content": "What is artificial intelligence and how does it work?"}
]

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 all tasks:

Example 1

Question: What is artificial intelligence and how does it work?

Response:

<|channel|>analysis<|message|>Need explain AI, types, how works.<|end|><|start|>assistant<|channel|>final<|message|>## What Is Artificial Intelligence?

Artificial Intelligence (AI) is the field of computer science that builds systems capable of performing tasks that normally require human intelligence. These tasks include:

| Human Skill | AI Equivalent |
|-------------|---------------|
| Perceiving (seeing, hearing) | Computer vision, speech recognition |
| Reasoning (solving problems, planning) | Algorithms, decision trees, logic programming |
| Learning (adapting from experience) | Machine learning, deep learning |
| Understanding language | Natural language processing (NLP) |
| Acting (moving, manipulating objects) | Robotics, autonomous vehicles |

In short, AI is about **making machines that can think, learn, and act**—often in ways that mimic or even surpass human abilities.

---

## How Does AI Work? (A High‑Level View)

1. **Data Collection & Representation**
   - AI systems learn from data. This can be images, text, sensor readings, or any structured/unstructured information.
   - Data is cleaned, labeled, and transformed into a format the algorithm can use (e.g., pixel arrays, word embeddings).

2. **Model Selection**
   - Choose a computational model that can capture patterns in the data. Common families:
     - **Statistical models** (e.g., linear regression, logistic regression)
     - **Symbolic AI** (rule‑based systems, knowledge graphs)
     - **Neural networks** (deep learning, CNNs, RNNs, transformers)
     - **Hybrid models** (combining symbolic and neural approaches)

3. **Training (Learning)**
   - The model is exposed to data and adjusts its internal parameters to minimize a **loss function** (a measure of error).
   - **Gradient descent** and its variants (Adam, RMSProp) are the most common optimization algorithms.
   - Training can be supervised (labels provided), unsupervised (discover structure), or reinforcement (learn from rewards).

4. **Evaluation & Validation**
   - Split data into training, validation, and test sets.
   - Use metrics (accuracy, precision, recall, F1, BLEU, etc.) to gauge performance.
   - Perform hyper‑parameter tuning, cross‑validation, and regularization to avoid overfitting.

5. **Deployment & Inference**
   - Once trained, the model is deployed to a target environment (cloud, edge device, mobile app).
   - During

Example 2

Question: Explain Einstein's theory of relativity in simple terms.

Response:

<|channel|>analysis<|message|>Need simple explanation.<|end|><|start|>assistant<|channel|>final<|message|>**Einstein’s theory of relativity** is really two ideas that changed how we think about space, time, and gravity.  Think of them as two “rules” that tell us how the universe behaves when things move fast or are near massive objects.

---

## 1. Special Relativity (1905)

### The two postulates
1. **The laws of physics are the same for everyone who is moving at a constant speed** (no “preferred” reference frame).  
2. **The speed of light in a vacuum is always the same (≈ 299 792 km/s)**, no matter how fast you’re moving or where you are.

### What that means in everyday language

| Concept | Everyday picture | What actually happens |
|---------|------------------|-----------------------|
| **Time slows down** | A clock on a fast‑moving train ticks slower than a clock on the platform. | Time is relative; the faster you move, the more your own time dilates (slows). |
| **Length shrinks** | A moving car looks shorter in the direction of motion. | Objects contract along the direction of motion, but only noticeable at speeds close to light. |
| **Mass increases** | A car feels heavier as it speeds up. | As you approach light speed, your relativistic mass grows, making it harder to accelerate further. |
| **Energy–mass equivalence** | “Energy is just another form of mass.” | \(E = mc^2\): a tiny amount of mass can be converted into a huge amount of energy. |

**Bottom line:** If you could travel near light speed, you would age slower than people who stayed on Earth, and distances would appear shorter in the direction you’re moving.

---

## 2. General Relativity (1915)

### The core idea
Gravity is not a “force” pulling objects together; it’s the **curvature of spacetime** caused by mass and energy.

### Visualizing it
- Imagine a heavy ball (the Sun) placed on a stretched rubber sheet (spacetime).  
- The ball creates a dip in the sheet.  
- A smaller marble (Earth) rolls around the dip, following a curved path—not because a force pulls it, but because the sheet itself is curved.

### Consequences

| Effect | Everyday analogy | Real‑world observation |
|--------|------------------|------------------------|
| **Light

Example 3

Question: How does democracy function in modern governments?

Response:

<|channel|>analysis<|message|>Need to explain democratic functioning.<|end|><|start|>assistant<|channel|>final<|message|>### 1.  What “democracy” means today  
Modern democracies are usually **constitutional republics** that combine:

| Core Feature | What it looks like in practice | Why it matters |
|--------------|--------------------------------|----------------|
| **Rule of law** | A written constitution, independent courts, and clear statutes | Prevents arbitrary power |
| **Separation of powers** | Executive, legislature, judiciary operate independently | Checks and balances |
| **Free and fair elections** | Regular, competitive, secret‑ballot voting | Legitimacy of the government |
| **Political pluralism** | Multiple parties, interest groups, and civil society | Representation of diverse views |
| **Civil liberties** | Freedom of speech, press, assembly, religion | Enables dissent and accountability |
| **Accountability mechanisms** | Audits, ombudsmen, media scrutiny | Holds officials responsible |

### 2.  How the system works in practice

| Stage | Typical Process | Key Institutions | Common Variations |
|-------|-----------------|-------------------|-------------------|
| **Electoral cycle** | 1. Parties nominate candidates 2. Campaigning 3. Voting 4. Counting 5. Certification | Electoral commissions, political parties, media | Some use proportional representation, others first‑past‑the‑post |
| **Legislative function** | Bills drafted → committee review → floor debate → vote → executive assent | Parliament/ congress, committees, opposition | Bicameral vs unicameral, filibuster rules |
| **Executive function** | President/prime minister sets policy agenda, appoints ministers, signs laws | Head of state, cabinet, civil service | Presidential vs parliamentary systems |
| **Judicial function** | Interprets laws, adjudicates disputes, reviews constitutionality | Supreme court, lower courts | Judicial review powers vary |
| **Oversight & checks** | Auditors, ombudsmen, investigative committees, media | Comptroller‑auditor general, parliamentary oversight committees | Some countries have stronger anti‑corruption agencies |

### 3.  The “daily” mechanics

1. **Policy debate** – Legislators discuss bills; public hearings allow citizen input.  
2. **Voting** – Majority rules, but minority rights are protected (e.g., veto overrides, filibusters).  
3. **Implementation** – Civil servants execute laws; agencies monitor compliance.  
4. **Feedback** –

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