You can deploy this model and start earning money today!

AmanPriyanshu/
gpt-oss-20.3b-specialized-science-pruned-moe-only-31-experts

Name: gpt-oss-20.3b-specialized-science-pruned-moe-only-31-experts
Availability: InStock

Views

No views yet

Science GPT-OSS Model (31 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 31 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	~20.3B (pruned from 21B)
Original Experts per Layer	32
Pruned Experts per Layer	31
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 31 experts per layer
No Retraining: Direct removal without additional training steps

Performance & Applications

Pruning Benefits

Smaller Memory Footprint: 96.9% 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 31 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 31 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-20.3b-specialized-science-pruned-moe-only-31-experts", 
    torch_dtype=torch.bfloat16, 
    device_map="cpu", 
    trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-20.3b-specialized-science-pruned-moe-only-31-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-20.3b-specialized-science-pruned-moe-only-31-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-20.3b-specialized-science-pruned-moe-only-31-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 to explain photosynthesis.<|end|><|start|>assistant<|channel|>final<|message|>**Photosynthesis** is the process by which green plants, algae, and some bacteria convert light energy from the sun into chemical energy stored in glucose (a sugar) and other organic molecules. It is the foundation of almost all life on Earth because it supplies the oxygen we breathe and the organic matter that fuels food webs.

---

## 1. Where It Happens

| Component | Location | Role |
|-----------|----------|------|
| **Chloroplast** | Cell organelle | Houses the machinery for photosynthesis |
| **Thylakoid membranes** | Inside chloroplast | Site of light‑dependent reactions |
| **Stroma** | Fluid surrounding thylakoids | Site of light‑independent (Calvin) cycle |
| **Chlorophyll** | Pigment in thylakoids | Captures light energy |

---

## 2. Two Main Phases

1. **Light‑Dependent Reactions (Photophosphorylation)**
2. **Light‑Independent Reactions (Calvin Cycle / Dark Reactions)**

### 2.1 Light‑Dependent Reactions

| Step | What Happens | Key Molecules |
|------|--------------|---------------|
| **1. Light absorption** | Chlorophyll and accessory pigments absorb photons. | Chlorophyll a, b, carotenoids |
| **2. Water splitting (Photolysis)** | Light energy splits H₂O into O₂, H⁺, and electrons. | H₂O → ½ O₂ + 2H⁺ + 2e⁻ |
| **3. Electron transport chain (ETC)** | Electrons move through Photosystem II → plastoquinone → cytochrome b₆f → Photosystem I → plastocyanin. | PSII, PQ, Cyt b₆f, PSI, PC |
| **4. ATP synthesis** | Proton gradient (ΔpH) drives ATP synthase. | ATP |
| **5. NADPH formation** | Electrons reduce NADP⁺ to NADPH. | NADPH |
| **6. Oxygen release** | O₂ is released as a by‑product. | O₂ |

**Net result of light‑dependent reactions:**

\[
2\,\text{H}_2\text{O} + 2\,\text{NADP}^+ + 3\,\text{ADP} +

Example 2

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

Response:

<|channel|>analysis<|message|>Need explain greenhouse effect causes, mechanism, gases, radiation, etc.<|end|><|start|>assistant<|channel|>final<|message|>**The greenhouse effect** is the natural process that keeps Earth’s surface warm enough to support life. It is caused by certain gases in the atmosphere that absorb and re‑radiate infrared (IR) radiation emitted by the planet’s surface. Below is a step‑by‑step explanation of what causes it and how it works.

---

## 1. Solar Energy Reaches Earth

| Step | What Happens | Key Points |
|------|--------------|------------|
| **1. Solar radiation** | The Sun emits a spectrum of electromagnetic radiation, from ultraviolet (UV) to visible light to infrared (IR). | About 1,361 W/m² (solar constant) reaches the top of Earth’s atmosphere. |
| **2. Atmospheric filtering** | Some UV and visible light is absorbed by the ozone layer and other gases; the rest passes through. | Roughly 70 % of incoming solar energy reaches the surface. |
| **3. Surface absorption** | The Earth’s surface (land, oceans, ice) absorbs the short‑wave radiation and warms. | Warm surface emits long‑wave (infrared) radiation. |

---

## 2. Surface Emits Infrared Radiation

- **Thermal radiation**: Warm objects emit IR radiation proportional to their temperature (Stefan–Boltzmann law: \(E = \sigma T^4\)).
- **Typical surface temperature**: ~288 K (15 °C) → emission peaks around 10 µm wavelength.

---

## 3. Greenhouse Gases Absorb IR

| Gas | Wavelengths absorbed | Typical atmospheric concentration |
|-----|----------------------|-----------------------------------|
| **Water vapor (H₂O)** | 5–8 µm, 12–20 µm | Variable, up to 4 % by volume |
| **Carbon dioxide (CO₂)** | 4.3 µm, 15 µm | ~0.04 % (400 ppm) |
| **Methane (CH₄)** | 7.7 µm | ~0.0002 % (1.8 ppm) |
| **Nitrous oxide (N₂O)** | 4.5 µm | ~0.00002 % (0.3 ppm) |
| **Ozone (O₃)** |

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 used in the growth, development, functioning, and reproduction of all known living organisms and many viruses. Its structure and function can be understood in two complementary parts:

---

## 1. Structure of DNA

| Feature | Description | Significance |
|---------|-------------|--------------|
| **Double‑Helix** | Two complementary strands wound around each other like a twisted ladder. | Provides stability and a compact way to store long sequences of genetic information. |
| **Backbone** | Alternating deoxyribose sugars and phosphate groups. | Forms a sturdy, negatively charged scaffold that protects the bases and allows enzymes to read the sequence. |
| **Nitrogenous Bases** | Four types: Adenine (A), Thymine (T), Cytosine (C), Guanine (G). | Encode genetic information; base‑pairing (A‑T, C‑G) ensures accurate replication. |
| **Base‑Pairing Rules** | A pairs with T via two hydrogen bonds; C pairs with G via three hydrogen bonds. | Maintains the uniform width of the helix and allows precise copying of DNA. |
| **Antiparallel Orientation** | One strand runs 5’→3’, the other 3’→5’. | Essential for replication and transcription machinery to function correctly. |
| **Supercoiling & Chromatin** | DNA is wrapped around histone proteins forming nucleosomes, further coiled into higher‑order structures. | Enables compaction into chromosomes and regulates access to genetic information. |

---

## 2. Function of DNA

| Function | How It Works | Biological Importance |
|----------|--------------|-----------------------|
| **Genetic Information Storage** | The sequence of bases constitutes the genome. | Determines the traits and capabilities of an organism. |
| **Replication** | DNA polymerases synthesize a new complementary strand using the existing strand as a template. | Enables cell division and inheritance of genetic material. |
| **Transcription (DNA → RNA)** | RNA polymerase reads a DNA template strand and synthesizes messenger RNA (mRNA). | Transfers genetic information from the nucleus to the cytoplasm for protein synthesis. |
| **Translation (RNA → Protein)** | Ribosomes read mRNA codons and assemble amino acids into polypeptide chains. | Produces the proteins that perform cellular functions

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