Project: https://amanpriyanshu.github.io/GPT-OSS-MoE-ExpertFingerprinting/
This is a pruned variant of OpenAI's GPT-OSS-20B model, reduced to 14 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 | ~10.2B (pruned from 21B) |
| Original Experts per Layer | 32 |
| Pruned Experts per Layer | 14 |
| 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 14 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-10.2b-specialized-math-pruned-moe-only-14-experts",
torch_dtype=torch.bfloat16,
device_map="cpu",
trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-10.2b-specialized-math-pruned-moe-only-14-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-10.2b-specialized-math-pruned-moe-only-14-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-10.2b-specialized-math-pruned-moe-only-14-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|>We need to solve 2x+5=17. Steps: subtract 5, divide by 2. Provide explanation.<|end|><|start|>assistant<|channel|>final<|message|>**Equation to solve**
\[
2x + 5 = 17
\]
---
### Step 1: Isolate the term with the variable
Subtract the constant term (5) from both sides of the equation so that the variable term \(2x\) stands alone.
\[
2x + 5 - 5 = 17 - 5
\]
\[
2x = 12
\]
---
### Step 2: Solve for \(x\)
Divide both sides by the coefficient of \(x\) (which is 2).
\[
\frac{2x}{2} = \frac{12}{2}
\]
\[
x = 6
\]
---
### Result
\[
\boxed{x = 6}
\]
**Check**:
Plug \(x = 6\) back into the original equation:
\[
2(6) + 5 = 12 + 5 = 17
\]
The left side equals the right side, so the solution is correct.<|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 Pythythorem Theorem (often called the “Pythyth theorem” or “Pythyth’s theorem”)** is a fundamental result in Euclidean geometry that relates the lengths of the sides of a right‑angled triangle. It states that in a right‑angled triangle, the square of the length of the hyp‑side (the side opposite the right angle) equals the sum of the squares of the lengths of the other two sides (the legs).
Mathematically, if a triangle has legs of lengths \(a\) and \(b\) and hyp‑side \(c\), then
\[
c^{2}=a^{2}+b^{2}.
\]
---
## How it is used
| **Domain** | **Application** |
|---|---|
| **Geometry** | • Finding the length of a side of a right‑angled triangle. <br>• Proving that the shape of a right‑angled triangle is unique. |
| **Trometry** | • Deriving tr‑angles from the law of cosines and sines. |
| **Physics** | • Calculating distances in 2‑D or 3‑D space (e.g., the distance between two points). |
| **Engineering** | • Designing right‑angled structures (e.g., building a bridge, a road, or a machine). |
| **Computer Science** | • Algorithms for computing distances (e.g., Euclidean distance in graph theory). |
| **Mathematics** | • Proving the existence of a unique solution to a system of equations. |
---
### Example
Suppose you have a right‑angled triangle with legs of 3 units and 4 units.
Using the theorem:
\[
c^{2}=3^{2}+4^{2}=9+16=25 \imarrow c=5.
\]
So the hyp‑side is 5 units long.
---
### Why it matters?
* It is the simplest way to compute distances in a 2‑D plane.
* It is the basis for many advanced results (e.g., the law of cosines, the law of sines).
* It is a building block for many algorithms in computer science and engineering.
---
**Bottom‑down**: The Pythyth theorem is a tool that lets you find the length of a side of a right‑angled triangle
Question: Calculate the area of a circle with radius 7 meters.
Response:
<|channel|>analysis<|message|>Need area = π r^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 \(r = 7\ \text{m}\):
\[
A = \pi (7\ \text{m})^{2}
= \pi \times 49\ \text{m}^{2}
\approx 3.14159 \times 49\ \text{m}^{2}
\approx 153.94\ \text{m}^{2}
\]
So the area of the circle is **about \(1.54 \times 10^{2}\) m²** (≈ 154 m²).<|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}
}