Project: https://amanpriyanshu.github.io/GPT-OSS-MoE-ExpertFingerprinting/
This is a pruned variant of OpenAI's GPT-OSS-20B model, reduced to 23 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 | ~15.5B (pruned from 21B) |
| Original Experts per Layer | 32 |
| Pruned Experts per Layer | 23 |
| 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 23 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-15.5b-specialized-science-pruned-moe-only-23-experts",
torch_dtype=torch.bfloat16,
device_map="cpu",
trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-15.5b-specialized-science-pruned-moe-only-23-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-15.5b-specialized-science-pruned-moe-only-23-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-15.5b-specialized-science-pruned-moe-only-23-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 concise explanation.<|end|><|start|>assistant<|channel|>final<|message|>**Photosynthesis in Plants – A Step‑by‑Step Overview**
| **Stage** | **What Happens** | **Key Players** |
|-----------|------------------|----------------|
| 1. Light Capture | Sunlight hits chlorophyll‑rich pigments in the chloroplasts. | Chlorophyll a & b, accessory pigments (carotenoids). |
| 2. Water Splitting (Photolysis) | Light energy splits H₂O into H⁺, e⁻, and O₂. | Photosystem II (PSII), oxygen‑evolving complex. |
| 3. Electron Transport Chain (ETC) | Electrons travel through PSII → plastoquinone → cytochrome b₆f → plastocyanin → Photosystem I (PSI). | PSII, cytochrome b₆f, PSI, ATP synthase. |
| 4. ATP & NADPH Production | Proton gradient drives ATP synthase; electrons reduce NADP⁺ to NADPH. | ATP synthase, ferredoxin, NADP⁺ reductase. |
| 5. Carbon Fixation (Calvin Cycle) | CO₂ is fixed into 3‑phosphoglycerate (3‑PGA) by Rubisco, then converted to glyceraldehyde‑3‑phosphate (G3P). | Rubisco, G3P dehydrogenase, triose phosphate isomerase. |
| 6. Regeneration of Ribulose‑1,5‑Bisphosphate (RuBP) | G3P is split; some is used to regenerate RuBP, allowing the cycle to continue. | ATP, NADPH, various enzymes. |
| 7. Product Formation | G3P exits the cycle to form glucose, fructose, sucrose, starch, cellulose, etc. | Glycolytic enzymes, sucrose‑phosphate synthase, starch synthase. |
### Key Points
1. **Light‑Dependent Reactions** (Stages 1–4) occur in the thylakoid membranes and produce ATP and NADPH.
2. **Light‑Independent Reactions** (Calvin Cycle, Stage 5–7) occur in the stroma and use ATP/NADPH to fix CO₂ into sugars.
3. **Oxygen Evolution**: The splitting of water releases O₂, which is released into the atmosphere.
4.
Question: What causes the greenhouse effect and how does it work?
Response:
<|channel|>analysis<|message|>Need explanation.<|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 to support life. It is caused by certain gases in the atmosphere—called **greenhouse gases (GHGs)**—that absorb, trap, and re‑radiate infrared (IR) radiation emitted by the planet’s surface. The main greenhouse gases are:
| Gas | Typical Atmospheric Concentration | Primary Source |
|-----|-------------------------------------|----------------|
| **Water vapor (H₂O)** | ~0.3 % (variable) | Evaporation from oceans, lakes, plants |
| **Carbon dioxide (CO₂)** | ~0.04 % (≈ 410 ppm) | Fossil‑fuel combustion, deforestation, industrial processes |
| **Methane (CH₄)** | ~0.00018 % (≈ 1.8 ppm) | Agriculture (especially ruminants), wetlands, natural gas leaks |
| **Nitrous oxide (N₂O)** | ~0.00027 % (≈ 0.27 ppm) | Agricultural fertilizers, combustion, industrial processes |
| **Fluorinated gases (e.g., CFCs, HFCs, SF₆)** | Trace amounts | Refrigeration, air‑conditioning, industrial processes |
These gases are **transparent to visible sunlight** (the short‑wave radiation that reaches Earth) but **opaque to the long‑wave (infrared) radiation** that the Earth emits after absorbing that sunlight. That opacity is what creates the greenhouse effect.
---
### How the Greenhouse Effect Works (Step‑by‑Step)
1. **Solar Radiation Reaches Earth**
- The Sun emits a spectrum of electromagnetic radiation.
- About 30 % of the incoming solar energy is reflected back to space by clouds, ice, and the atmosphere.
- The remaining ~70 % is absorbed by the Earth’s surface (land, oceans, vegetation).
2. **Earth’s Surface Heats Up**
- The absorbed solar energy warms the surface.
- A warm surface emits thermal radiation (infrared, ~10 µm wavelength) back into the atmosphere.
3. **Greenhouse Gases Absorb Infrared Radiation**
- GHG molecules have vibrational modes that resonate with IR wavelengths.
- When a GHG molecule absorbs an IR photon, its internal energy increases
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 stores and transmits genetic information in living organisms. Its structure and function can be understood in two complementary layers: the *molecular architecture* and the *biological roles* it plays.
---
## 1. Structural Features
| Feature | Description | Biological Significance |
|---------|-------------|------------------------|
| **Double‑helix** | Two complementary strands wound around each other in a right‑handed helix. | Provides stability and a compact, yet flexible, storage form. |
| **Back‑bone** | Alternating sugar (deoxyribose) and phosphate groups. | Forms a rigid, negatively charged scaffold that protects the bases. |
| **Nucleobases** | Four nitrogenous bases: Adenine (A), Thymine (T), Cytosine (C), Guanine (G). | Encode genetic information; base pairing (A‑T, C‑G) ensures faithful replication. |
| **Base pairing** | Hydrogen bonds: 2 between A‑T, 3 between C‑G. | Drives complementary strand synthesis and accurate base pairing during replication and transcription. |
| **Major & Minor grooves** | Structural pockets that accommodate protein binding. | Allows specific interaction with transcription factors, polymerases, and repair enzymes. |
| **Super‑coiling** | Higher‑level winding of the helix. | Compacts DNA into chromosomes and facilitates processes like replication and transcription. |
| **Chromatin packaging** | DNA wrapped around histone octamers → nucleosomes → higher‑order fibers. | Regulates accessibility of genes and protects DNA from damage. |
---
## 2. Functional Roles
| Function | Mechanism | Outcome |
|----------|-----------|---------|
| **Genetic Information Storage** | Sequence of bases encodes proteins, RNAs, regulatory elements. | Determines phenotype, cellular function, and inheritance. |
| **Replication** | Semi‑conservative duplication via DNA polymerases; helicases unwind the helix; primase lays RNA primers. | Produces identical daughter chromosomes for cell division. |
| **Transcription** | RNA polymerase binds promoter, unwinds DNA, synthesizes mRNA complementary to the template strand. | Generates messenger RNA that directs protein synthesis. |
| **Translation (indirect)** | mRNA is translated by ribosomes into polypeptides. | Produces
If you use this model in your research, please cite:
@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}
}