You can deploy this model and start earning money today!

AmanPriyanshu/
gpt-oss-13.7b-specialized-science-pruned-moe-only-20-experts

Name: gpt-oss-13.7b-specialized-science-pruned-moe-only-20-experts
Availability: InStock

Views

No views yet

Science GPT-OSS Model (20 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 20 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 science 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	~13.7B (pruned from 21B)
Original Experts per Layer	32
Pruned Experts per Layer	20
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	Science

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 science evaluation tasks
Systematic Reduction: Reduced from 32 to 20 experts per layer
No Retraining: Direct removal without additional training steps

Performance & Applications

Pruning Benefits

Smaller Memory Footprint: 62.5% of original expert parameters
Reduced Computational Load: Fewer routing decisions during inference
Focused Capabilities: Retains experts relevant to science 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 science-specialized model leverages experts that showed high activation patterns during scientific reasoning tasks from GPQA (physics, chemistry, biology) and MMLU science domains. These experts demonstrate superior performance on complex scientific problem-solving and technical knowledge recall.

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 science tasks, we created this specialized model that maintains domain expertise while significantly reducing computational requirements from 32 to 20 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 20 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-13.7b-specialized-science-pruned-moe-only-20-experts", 
    torch_dtype=torch.bfloat16, 
    device_map="cpu", 
    trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-13.7b-specialized-science-pruned-moe-only-20-experts")

# Generate with the model
messages = [
    {"role": "user", "content": "Explain the process of photosynthesis in plants."}
]

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-13.7b-specialized-science-pruned-moe-only-20-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-13.7b-specialized-science-pruned-moe-only-20-experts")

# Generate with the model
messages = [
    {"role": "user", "content": "Explain the process of photosynthesis in plants."}
]

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 science tasks:

Example 1

Question: Explain the process of photosynthesis in plants.

Response:

<|channel|>analysis<|message|>Need concise explanation.<|end|><|start|>assistant<|channel|>final<|message|>**Photosynthesis in Plants – A Step‑by‑Step Overview**

| **Stage** | **What Happens** | **Key Players** | **Result** |
|-----------|-------------------|-----------------|-----------|
| 1. **Light Capture** | Sunlight hits chlorophyll‑rich cells (chloroplasts). | • Chlorophyll‑a & b (pigments) | • Excited electrons |
| 2. **Water Splitting (Photolysis)** | Light energy splits H₂O → 2 H⁺ + 2 e⁻ + ½ O₂. | • Photosystem II (PSII) | • Oxygen released to air |
| 3. **Electron Transport Chain (ETC)** | Electrons travel through plastoquinone, cytochrome b₆₀, and Photosystem I (PSI). | • ATP synthase, NADP reductase | • Generates ATP & NADPH |
| 4. **Carbon Fixation (Calvin Cycle)** | CO₂ + 3 ATP + 2 NADPH → G3P (glyceraldehyde‑3‑phosphate). | • CO₂‑fixing enzyme Rubisco | • 3‑C sugars (G3P) |
| 5. **Sugar Utilization** | G3P → glucose, fructose, sucrose, starch, cellulose. | • Hexokinase, aldolase, etc. | • Energy storage & structural material |
| 6. **Oxygen Release** | O₂ produced in water‑splitting step is expelled. | • Gas exchange via stomata | • Supports aerobic life |

### Key Points

1. **Energy Flow**  
   * Light energy → chlorophyll → electron transport → ATP + NADPH → CO₂ fixation.*

2. **Dual Role of Chloroplasts**  
   * Light reactions (outer thylakoid membranes) produce ATP/NADPH and O₂.*  
   * Dark reactions (stroma) use ATP/NADPH to build sugars.*

3. **Regulation**  
   * Stomatal opening controls CO₂ intake and water loss.*  
   * Light intensity, CO₂ concentration, and temperature influence rates.*

4. **End Products**  
   * Primary: glucose (C₆H₁₂O₆) and oxygen.*  
   * Secondary: starch (storage),

Example 2

Question: What causes the greenhouse effect and how does it work?

Response:

<|channel|>analysis<|message|>Need explain greenhouse effect causes, greenhouse gases, absorption, re-emission, energy balance.<|end|><|start|>assistant<|channel|>final<|message|>**What causes the greenhouse effect?**  
The greenhouse effect is a natural process that keeps Earth’s surface warm enough for life. It is caused by certain gases in the atmosphere—called *greenhouse gases* (GHGs)—that can absorb and re‑emit infrared (IR) radiation. The main GHGs are:

| Gas | Natural source | Human‑induced source | Typical atmospheric concentration |
|-----|----------------|------------------------|-------------------------------------|
| **CO₂** | Volcanoes, respiration, decomposition | Fossil‑fuel combustion, deforestation | ~400 ppm (≈ 0.04 %) |
| **CH₄** | Wetlands, natural seepage | Agriculture (livestock, rice paddies), natural gas leaks | ~1.8 ppm |
| **N₂O₄** | Lightning, soil microbes | Industrial processes, fertilizers | ~0.3 ppm |
| **H₂O** | Evaporation | None (water vapor is a feedback, not a direct human source) | 0–4 % (varies with temperature) |
| **O₃** | Photochemical reactions | Industrial emissions, vehicle exhaust | ~0.05 ppm |
| **CFCs, HFCs, PFCs** | None | Refrigerants, foam blowing agents | Trace amounts |

The key point is that **greenhouse gases are transparent to visible sunlight** (the energy that reaches Earth from the Sun) but **opaque to the long‑wave infrared radiation** that the Earth emits after absorbing that sunlight.  

---

## How the greenhouse effect works

1. **Solar radiation reaches Earth**  
   The Sun emits a spectrum that peaks in the visible and near‑infrared. The atmosphere is largely transparent to this radiation, so it penetrates to the surface and is absorbed by the land, oceans, and clouds.

2. **Earth re‑radiates energy as thermal IR**  
   The warmed surface emits energy mainly in the long‑wave (infrared) part of the spectrum (≈ 5–50 µm). This radiation is what the atmosphere must deal with.

3. **Greenhouse gases absorb IR**  
   Molecules of CO₂, CH₄, N₂O₄, H₂O, etc., have vibrational modes that resonate with IR wavelengths. When an

Example 3

Question: Describe the structure and function of DNA.

Response:

<|channel|>analysis<|message|>Need concise description.<|end|><|start|>assistant<|channel|>final<|message|>**DNA (deoxyribonucleic acid)** is the hereditary material that carries the genetic instructions for building and regulating all living organisms. Its structure and function can be broken down into several key components:

| Feature | Description | Functional Role |
|--------|-------------|-----------------|
| **Double‑helix backbone** | Two sugar‑phosphate chains wound around each other in a right‑handed helix. | Provides structural stability and protects the genetic information. |
| **Nucleobases** | Four nitrogenous bases: adenine (A), thymine (T), cytosine (C), guanine (G). | Encode the genetic code; A pairs with T, C pairs with G. |
| **Base pairing (hydrogen bonds)** | A–T (2 H‑bonds), C–G (3 H‑bonds). | Enables complementary base pairing, essential for accurate replication and transcription. |
| **Major & minor grooves** | The helical twist creates two grooves that proteins can access. | Allows DNA‑binding proteins (e.g., transcription factors, polymerases) to recognize specific sequences. |
| **Chromatin packaging** | DNA wrapped around histone proteins to form nucleosomes, further folded into chromatin. | Compacts DNA to fit in the nucleus and regulates gene accessibility. |
| **Genomic organization** | Segmented into genes, regulatory elements, introns, exons, and non‑coding regions. | Determines which proteins are produced and when. |

### Core Functions

1. **Genetic Information Storage**  
   - The sequence of bases constitutes a code that specifies amino acids, RNA molecules, and regulatory signals.

2. **Replication**  
   - During cell division, DNA polymerases synthesize a complementary strand for each template strand, producing two identical DNA molecules.  
   - Fidelity is maintained by proofreading and mismatch repair mechanisms.

3. **Transcription**  
   - RNA polymerase reads a DNA template to synthesize messenger RNA (mRNA).  
   - The mRNA sequence is then translated into proteins.

4. **Regulation of Gene Expression**  
   - DNA‑binding proteins (transcription factors, repressors, activators) interact with promoter and enhancer regions to turn genes on or off.  
   - Epigenetic modifications (DNA methylation, histone acetylation) alter chromatin structure and influence accessibility.

5. **Repair & Maintenance**

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