Downloads · 30 days
13
2% of all-time downloads
AmanPriyanshu/gpt-oss-20.9b-specialized-all-pruned-moe-only-32-experts
gpt-oss-20.9b-specialized-all-pruned-moe-only-32-experts is a text generation model from AmanPriyanshu. Use it when you need the model to write or continue text. The card lists the license as apache-2.0.
Project: https://amanpriyanshu.github.io/GPT-OSS-MoE-ExpertFingerprinting/
Downloads · 30 days
13
2% of all-time downloads
All-time downloads
567
Public
Parameters
20.9B
41.9 GB on disk
Likes
1
Public
Click a slice to open those files.
.safetensors41.8 GB · 100%
From the Hugging Face model README
Project: https://amanpriyanshu.github.io/GPT-OSS-MoE-ExpertFingerprinting/
<div align="center">This is a pruned variant of OpenAI's GPT-OSS-20B model, reduced to 32 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 all 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 | ~20.9B (pruned from 21B) |
| Original Experts per Layer | 32 |
| Pruned Experts per Layer | 32 |
| 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 | All |
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 general-purpose model maintains broad capabilities across all domains while significantly reducing computational requirements. It preserves the essential routing patterns discovered across our comprehensive analysis of diverse evaluation benchmarks including GPQA, MMLU, SORRY-Bench, and Tulu3 datasets.
The expert selection process utilized our comprehensive analysis of router activation patterns across multiple evaluation benchmarks:
By identifying experts that consistently activated for all tasks, we created this specialized model that maintains domain expertise while significantly reducing computational requirements from 32 to 32 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-20.9b-specialized-all-pruned-moe-only-32-experts",
torch_dtype=torch.bfloat16,
device_map="cpu",
trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-20.9b-specialized-all-pruned-moe-only-32-experts")
# Generate with the model
messages = [
{"role": "user", "content": "What is artificial intelligence and how does it work?"}
]
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-20.9b-specialized-all-pruned-moe-only-32-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-20.9b-specialized-all-pruned-moe-only-32-experts")
# Generate with the model
messages = [
{"role": "user", "content": "What is artificial intelligence and how does it work?"}
]
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 all tasks:
Question: What is artificial intelligence and how does it work?
Response:
<|channel|>analysis<|message|>Need explain AI, types, how works.<|end|><|start|>assistant<|channel|>final<|message|>## What Is Artificial Intelligence?
Artificial Intelligence (AI) is the field of computer science that seeks to create systems capable of performing tasks that normally require human intelligence. These tasks include:
| Task | Human‑like AI Capability |
|------|--------------------------|
| Understanding language | Natural Language Processing (NLP) |
| Seeing and recognizing objects | Computer Vision |
| Making decisions | Reasoning, Planning, Decision‑Making |
| Learning from experience | Machine Learning (ML) |
| Interacting socially | Dialogue Systems, Chatbots |
In short, AI is about building machines that can **observe, reason, learn, and act** in ways that mimic or surpass human cognition.
---
## How Does AI Work? – The Core Ingredients
| Ingredient | What It Does | Typical Techniques |
|------------|--------------|--------------------|
| **Data** | The raw material from which AI learns. | Images, text, sensor readings, logs, etc. |
| **Algorithms** | Rules or procedures that process data. | Linear regression, decision trees, neural networks, reinforcement learning, etc. |
| **Computational Power** | Hardware that runs the algorithms fast enough. | GPUs, TPUs, distributed clusters. |
| **Evaluation** | Metrics to judge performance. | Accuracy, F1‑score, BLEU, etc. |
| **Deployment** | Making the model usable in real systems. | APIs, edge devices, cloud services. |
### 1. Data Collection & Pre‑processing
- **Collect**: Gather raw data from sensors, databases, the web, etc.
- **Clean**: Remove noise, handle missing values, correct errors.
- **Transform**: Normalize, encode categorical variables, augment (e.g., rotate images).
### 2. Choosing a Model
- **Rule‑Based Systems**: Hand‑crafted if‑then rules (early AI, expert systems).
- **Statistical Models**: Logistic regression, SVMs, decision trees.
- **Neural Networks**: Deep learning models (CNNs for vision, RNNs/Transformers for language).
- **Reinforcement Learning**: Agents learn by interacting with an environment.
### 3. Training
- **Objective Function**: Define what “good” means (e.g., minimize error).
- **Optimization**: Use gradient descent or variants to adjust model parameters.
- **Regularization**: Prevent overfitting (
Question: Explain Einstein's theory of relativity in simple terms.
Response:
<|channel|>analysis<|message|>Need simple explanation.<|end|><|start|>assistant<|channel|>final<|message|>**Einstein’s theory of relativity** is really two ideas that changed how we think about space, time, and gravity. Think of them as two “rules” that tell us how the universe behaves when things move fast or are near massive objects.
---
## 1. Special Relativity (1905)
### The “speed‑of‑light” rule
- **Light always moves at the same speed** (≈ 299 792 km/s) no matter how fast you’re moving or where you’re looking from.
- Because light’s speed is fixed, the usual rules of everyday physics (like adding speeds) don’t work the same way.
### Consequences (in plain language)
| What you’d expect | What actually happens |
|-------------------|-----------------------|
| If you run toward a flashlight, the light should be faster than if you’re standing still. | The light’s speed is the same for both of you. |
| Time is the same for everyone. | **Time slows down** for a fast‑moving person (time dilation). |
| Lengths are the same for everyone. | **Lengths shrink** in the direction of motion (length contraction). |
| Mass is constant. | **Mass increases** as you approach light speed (relativistic mass). |
**Why it matters:**
- Explains why GPS satellites need to account for both their speed and the Earth’s gravity to stay accurate.
- Shows that “simultaneity” (whether two events happen at the same time) depends on the observer’s motion.
---
## 2. General Relativity (1915)
### Gravity is not a force, it’s geometry
- Massive objects (planets, stars, black holes) **curve the fabric of space‑time**.
- Objects move along the straightest possible paths in this curved space‑time, which we feel as gravity.
### Visualizing it
- Imagine a heavy ball on a stretched rubber sheet.
- The sheet dips around the ball.
- A smaller marble rolled nearby will spiral toward the ball, not because a force pulls it, but because the sheet is curved.
### Consequences
| Classic idea | Relativistic idea |
|--------------|-------------------|
| Gravity pulls objects with a force. | Gravity is the shape of space‑time. |
| Light travels in straight lines. | Light follows curved paths (gravitational lensing). |
| Time runs
Question: How does democracy function in modern governments?
Response:
<|channel|>analysis<|message|>Need to explain democratic functioning.<|end|><|start|>assistant<|channel|>final<|message|>### 1. What “democracy” means today
Modern democracies are usually **constitutional republics** that combine:
| Feature | What it looks like in practice | Why it matters |
|---------|--------------------------------|----------------|
| **Rule of law** | A written constitution, independent courts, and clear statutes | Prevents arbitrary power |
| **Separation of powers** | Executive, legislature, judiciary act independently | Checks and balances |
| **Free and fair elections** | Universal adult suffrage, secret ballot, independent electoral commissions | Legitimacy of the government |
| **Political pluralism** | Multiple parties, free media, civil society | Representation of diverse views |
| **Protection of rights** | Civil, political, economic, social, and cultural rights | Guarantees individual dignity |
| **Accountability mechanisms** | Audits, ombudsmen, anti‑corruption bodies | Keeps officials answerable |
---
### 2. How the system works day‑to‑day
| Stage | Key actors | Typical actions | Example |
|-------|------------|-----------------|---------|
| **Policy agenda setting** | Citizens, NGOs, interest groups, media | Raise issues, lobby, run campaigns | Climate‑change protests → policy debate |
| **Legislative process** | Parliament/ congress, committees, parties | Draft bills, debate, amend, vote | Health‑care bill → committee review → floor vote |
| **Executive implementation** | President/prime minister, ministries | Sign laws, issue regulations, allocate budgets | New tax law → Ministry of Finance publishes regulations |
| **Judicial review** | Courts, judges | Interpret laws, hear challenges | Constitutional challenge to a law → Supreme Court ruling |
| **Oversight & accountability** | Auditors, ombudsmen, investigative journalists | Audit budgets, investigate misconduct | Audit report on public spending → media coverage |
| **Public participation** | Citizens, elections, referenda, public consultations | Vote, petition, attend town halls | Referendum on constitutional amendment |
---
### 3. The electoral cycle
1. **Election day** – Voters cast ballots in a secret, free vote.
2. **Counting & certification** – Electoral commission tallies votes, resolves disputes.
3. **Formation of government** – The party or coalition with a majority forms the cabinet.
4. **Mandate period** – Usually 4–5 years; the government
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}
}