Project: https://amanpriyanshu.github.io/GPT-OSS-MoE-ExpertFingerprinting/
This is a pruned variant of OpenAI's GPT-OSS-20B model, reduced to 18 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 math 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 | ~12.6B (pruned from 21B) |
| Original Experts per Layer | 32 |
| Pruned Experts per Layer | 18 |
| 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 | Math |
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 mathematics-focused model utilizes experts that exhibited strong performance on mathematical reasoning tasks from MMLU mathematics subjects and quantitative sections. These experts excel at mathematical computation, proof strategies, and logical reasoning.
The expert selection process utilized our comprehensive analysis of router activation patterns across multiple evaluation benchmarks:
By identifying experts that consistently activated for math tasks, we created this specialized model that maintains domain expertise while significantly reducing computational requirements from 32 to 18 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-12.6b-specialized-math-pruned-moe-only-18-experts",
torch_dtype=torch.bfloat16,
device_map="cpu",
trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-12.6b-specialized-math-pruned-moe-only-18-experts")
# Generate with the model
messages = [
{"role": "user", "content": "Solve this equation: 2x + 5 = 17. Show your work step by step."}
]
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-12.6b-specialized-math-pruned-moe-only-18-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-12.6b-specialized-math-pruned-moe-only-18-experts")
# Generate with the model
messages = [
{"role": "user", "content": "Solve this equation: 2x + 5 = 17. Show your work step by step."}
]
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 math tasks:
Question: Solve this equation: 2x + 5 = 17. Show your work step by step.
Response:
<|channel|>analysis<|message|>Need to solve linear equation. Provide steps.<|end|><|start|>assistant<|channel|>final<|message|>**Equation to solve**
\[
2x + 5 = 17
\]
---
### Step 1: Isolate the term that contains \(x\)
The variable \(x\) is currently “hidden” behind the constant \(5\).
Subtract \(5\) from both sides of the equation to cancel that constant.
\[
\begin{aligned}
2x + 5 &= 17 \\[4pt]
2x + 5 - 5 &= 17 - 5 \\[4pt]
2x &= 12
\endaligned
\]
---
### Step 2: Solve for \(x\)
Now the equation is \(2x = 12\).
Divide both sides by the coefficient of \(x\), which is \(2\).
\[
\begin{aligned}
\frac{2x}{2} &= \frac{12}{2} \\[4pt]
x &= 6
\endaligned
\]
---
### Step 3: Verify the solution
Plug \(x = 6\) back into the original equation:
\[
2(6) + 5 = 12 + 5 = 17
\]
The left‑hand side equals the right‑hand side, so the solution is correct.
---
\[
\boxed{x = 6}
\]<|return|>
Question: What is the Pythagorean theorem and how is it used?
Response:
<|channel|>analysis<|message|>Need explain theorem, formula, usage.<|end|><|start|>assistant<|channel|>final<|message|>## The Pythagorean Theorem
**Statement**
In any right‑angled triangle (a triangle with one angle of \(90^\circ\)), the lengths of the two legs (\(a\) and \(b\)) and the hypotenuse (\(c\)) satisfy
\[
a^{2} + b^{2} = c^{2}.
\]
The hypotenuse is the side opposite the right angle, the longest side.
---
## Why It Works
The theorem is a consequence of Euclidean geometry and can be proved in many ways:
1. **Area argument** – Construct a square on each side of the triangle and show that the area of the square on the hypotenuse equals the sum of the areas of the squares on the legs.
2. **Similar triangles** – Use the fact that the angles of a right triangle are complementary and that the triangles formed by drawing the altitude to the hypotenuse are similar.
3. **Algebraic proof** – Use the law of cosines with \(\cos 90^\circ = 0\).
All proofs ultimately rely on the fact that the right angle forces the relationship between the sides.
---
## How It Is Used
| Application | What you do | Example |
|-------------|-------------|---------|
| **Finding a missing side** | Rearsolve \(c = \sqrt{a^{2}+b^{2}}\) or \(a = \sqrt{c^{2}-b^{2}}\) | A 3‑4‑5 triangle: \(c = \sqrt{3^{2}+4^{2}} = 5\). |
| **Checking if a triangle is right‑angled** | Verify if \(a^{2}+b^{2}=c^{2}\) | Triangle with sides 6, 8, 10 satisfies \(36+64=100\). |
| **Geometry & Engineering** | Compute distances, slopes, angles, or design shapes. | Finding the length of a diagonal in a rectangular building. |
| **Physics** | Resolve components of velocity or force vectors. | A 45°‑45°‑90° projectile path: \(v_x = v_y = v/\sqrt{2}\). |
| **Computer‑aided design (CAD)** | Compute distances between points in 3‑D space (using the Euclidean distance formula, a 3‑D extension of the theorem
Question: Calculate the area of a circle with radius 7 meters.
Response:
<|channel|>analysis<|message|>Need area = πr^2 = π*49 ≈ 153.938.<|end|><|start|>assistant<|channel|>final<|message|>The area \(A\) of a circle is given by
\[
A = \pi r^{2}
\]
where \(r\) is the radius.
For \(r = 7\) m:
\[
A = \pi (7\ \text{m})^{2}
= \pi \times 49\ \text{m}^{2}
\approx 3.14159 \times 49\ \text{m}^{2}
\approx 153.938\ \text{m}^{2}
\]
So the area of the circle is **about \(154\ \text{m}^{2}\)** (rounded to the nearest square meter).<|return|>
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}
}