Project: https://amanpriyanshu.github.io/GPT-OSS-MoE-ExpertFingerprinting/
This is a pruned variant of OpenAI's GPT-OSS-20B model, reduced to 16 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.
| Metric | Value |
|---|---|
| Base Model | openai/gpt-oss-20b |
| Architecture | Mixture-of-Experts Transformer |
| Total Parameters | ~11.4B (pruned from 21B) |
| Original Experts per Layer | 32 |
| Pruned Experts per Layer | 16 |
| 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 |
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:
Note: Performance may vary depending on how well the pruned experts match your specific use case.
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:
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 16 experts per layer.
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.
Our approach involves:
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.
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
# Load the specialized model on CPU
model = AutoModelForCausalLM.from_pretrained(
"AmanPriyanshu/gpt-oss-11.4b-specialized-science-pruned-moe-only-16-experts",
torch_dtype=torch.bfloat16,
device_map="cpu",
trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-11.4b-specialized-science-pruned-moe-only-16-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)
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-11.4b-specialized-science-pruned-moe-only-16-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-11.4b-specialized-science-pruned-moe-only-16-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)
For CUDA GPU inference, the models work with minimal modifications from the CPU example above. Simply change:
device_map="auto" # Will automatically use GPU if available
torch_dtype=torch.bfloat16 # or torch.float16
The following examples demonstrate the model's performance on science tasks:
Question: Explain the process of photosynthesis in plants.
Response:
<|channel|>analysis<|message|>Need explain process, steps, light absorption, chlorophyll, electron transport, NADPH, CO2 fixation, Calvin cycle, enzymes, outputs.<|end|><|start|>assistant<|channel|>final<|message|>**Photosynthesis in Plants – A Step‑by‑Step Overview**
| **Stage** | **What Happens** | **Key Molecules / Enzymes** | **Why It Matters** |
|-----------|-------------------|---------------------------|-------------------|
| 1. **Light Capture** | Sunlight hits the chloroplasts. | **Chlorophyll a & b** (pigments) | Converts photons into chemical energy |
| 2. **Excitation & Electron Transfer** | Light excites electrons in chlorophyll → *photo‑excited chlorophyll* → electrons move to *electron transport chain* (ETC). | **Photosystem II (PSII)**, **Photosystem I (PSI)**, **ATP synthase**, **cytochrome b₆₁ₙ** | Generates **ATP** and **NADPH** (energy carriers) |
| 3. **Water Splitting (Photolysis)** | PSII uses water → produces **O₂** and **H⁺**. | **H₂O → 2H⁺ + 2e⁻ + O₂** | Supplies electrons for ETC and releases oxygen for respiration |
| 4. **NADPH & ATP Production** | Electrons flow through ETC → **ATP** (energy) and **NADPH** (reducing power). | **ATP synthase**, **NADP⁺ reductase** | Energy for CO₂ fixation |
| 5. **CO₂ Fixation (Calvin Cycle)** | CO₂ + 3 ATP + 2 NADPH → 6 C₆H₆O₃ (glucose) + 3 ATP + 2 NADP⁺. | **RuBisCO (ribulose‑bisphosphate carboxylase‑oxygenase)**, **ATP‑dependent enzymes** | Stores energy in sugars, fuels growth |
| 6. **Sugar Export & Storage** | Glucose → starch, cellulose, sucrose → transported to tissues. | **Transport proteins**, **sugar‑transporters** | Provides building blocks for growth, food, and energy |
---
### Detailed Flow
1. **Photon Absorption**
* Light (400–700 nm) excites
Question: What causes the greenhouse effect and how does it work?
Response:
<|channel|>analysis<|message|>Need explain greenhouse effect causes, greenhouse gases, absorption of IR, trapping heat, radiative balance.<|end|><|start|>assistant<|channel|>final<|message|>### 1. What is the “greenhouse effect”?
The greenhouse effect (or greenhouse‑gas effect) is the natural warming of the Earth’s surface that occurs because certain gases in the atmosphere absorb and re‑emit infrared (IR) radiation that is emitted by the planet. Without these gases, the Earth would be much colder. The effect is named after the way a greenhouse traps heat: sunlight (visible light) passes through the glass, but the heat that the greenhouse emits (infra‑red) is reflected back by the glass, keeping the interior warm.
---
## 2. What causes it?
| **Key G gases** | **Sources** | **Why they matter** |
|-----------------|-------------|---------------------|
| **Water vapor (H₂O)** | Evaporates from oceans, lakes, soil, plants | Most of the greenhouse effect (≈ 55 %) – it’s a natural feedback that amplifies warming. |
| **Carbon dioxide (CO₂)** | Fossil‑fuel combustion, deforestation, industrial processes, agriculture | 28 % of the effect; its concentration has risen from ~280 ppv pre‑industrial to ~410 ppv today. |
| **Methane (CH₄)** | Natural‑gas extraction, wetlands, livestock | 15 % of the effect; it’s a potent greenhouse gas (≈ 28× CO₂ over 100 yr). |
| **Nitrous oxide (N₂O)** | Agricultural fertilizer, industrial processes | 6 % of the effect; very potent (≈ 265× CO₂). |
| **Ozone (O₃)** | Str ozone layer, industrial emissions | 3 % of the effect; absorbs UV and IR. |
| **Other trace gases** (C₂F₆, CF₄, SF₆, etc.) | Industrial, aviation, etc. | Small but potent. |
The “cause” is the **absorption of long‑wave (IR) radiation** by these molecules. They have rotational and vibrational energy levels that can be excited by IR photons, so they absorb the planet’s thermal radiation and then re‑emit it in all directions. Because the atmosphere is not perfectly transparent to IR, a net amount of energy is trapped and must be radiated away from the surface, forcing the Earth to warm
Question: Describe the structure and function of DNA.
Response:
<|channel|>analysis<|message|>Need to explain DNA structure: double helix, nucleotides, base pairs, major components, sugar-phosphate backbone, base pairing, major vs minor groove, major functional roles: genetic information, replication, transcription, regulation.<|end|><|start|>assistant<|channel|>final<|message|>**DNA – the Double‑Helix Blueprint of Life**
---
## 1. The Physical Structure of DNA
| Feature | What Is It? | Why It Matters |
|--------|------------|----------------|
| **Double‑helix** | Two strands of sugar‑phosphate back‑bones wind around each other | Gives a stable, compact shape that can be copied and read by proteins |
| **Nucleotides** | Adenine (A), Thymine (T), Cytosine (C), Guanine (G) | Carry the genetic “code” – the sequence of A‑T‑C‑G determines the information |
| **Base pairing** | A pairs with T, C pairs with G | Watson‑Crick model; ensures complementary strands |
| **Major vs. Minor groove** | The “major” groove is deeper, the “minor” is shallow | Transcription factors, enzymes, and DNA‑binding proteins preferentially bind to the major groove |
| **Phosphodiester backbone** | 3‑phosphoglycerol + 5‑deoxyribose | Provides chemical stability and a scaffold for the bases |
| **Length** | 3.4 Å per base pair; ~10⁶ bases in a chromosome | Determines how many genes a chromosome can hold |
| **Super‑structure** | Chromatin (DNA + histones + other proteins) | Modulates accessibility, compaction, and higher‑order organization |
### Visual Summary
5′ �──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
---
## 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}
}