Downloads · 30 days
10
9% of all-time downloads
AmanPriyanshu/gpt-oss-16.1b-specialized-safety-pruned-moe-only-24-experts
gpt-oss-16.1b-specialized-safety-pruned-moe-only-24-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
10
9% of all-time downloads
All-time downloads
110
Public
Parameters
16.1B
32.3 GB on disk
Likes
1
Public
Click a slice to open those files.
.safetensors32.3 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 24 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 safety 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 | ~16.1B (pruned from 21B) |
| Original Experts per Layer | 32 |
| Pruned Experts per Layer | 24 |
| 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 | Safety |
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 safety-focused model uses experts that performed well on safety evaluation tasks from SORRY-Bench. These experts are specialized in identifying and appropriately responding to potentially harmful content while maintaining helpful capabilities.
The expert selection process utilized our comprehensive analysis of router activation patterns across multiple evaluation benchmarks:
By identifying experts that consistently activated for safety tasks, we created this specialized model that maintains domain expertise while significantly reducing computational requirements from 32 to 24 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-16.1b-specialized-safety-pruned-moe-only-24-experts",
torch_dtype=torch.bfloat16,
device_map="cpu",
trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-16.1b-specialized-safety-pruned-moe-only-24-experts")
# Generate with the model
messages = [
{"role": "user", "content": "What should someone do if they encounter cyberbullying online?"}
]
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-16.1b-specialized-safety-pruned-moe-only-24-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-16.1b-specialized-safety-pruned-moe-only-24-experts")
# Generate with the model
messages = [
{"role": "user", "content": "What should someone do if they encounter cyberbullying online?"}
]
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 safety tasks:
Question: What should someone do if they encounter cyberbullying online?
Response:
<|channel|>analysis<|message|>Need advice.<|end|><|start|>assistant<|channel|>final<|message|>**If you’re being cyber‑bullied, you’re not alone and you don’t have to face it alone.**
Below is a step‑by‑step guide that covers what to do right away, how to protect yourself, and where to get help.
---
## 1. Stay Calm & Don’t Respond Immediately
- **Pause**: Give yourself a few minutes (or hours) before replying.
- **Avoid escalation**: Responding with anger or insults often fuels the bully.
- **Take a break**: Step away from the platform for a while if you feel overwhelmed.
---
## 2. Preserve Evidence
| Action | Why it matters | How to do it |
|--------|-----------------|--------------|
| **Screenshot** | Keeps a record that can’t be altered. | Use the built‑in screenshot tool or a screen‑capture app. |
| **Save messages** | Needed for reporting or legal action. | Copy the text, or use the platform’s “Save” or “Download” feature. |
| **Record URLs** | Shows where the harassment is happening. | Copy the link or use a browser bookmark. |
| **Keep logs** | Helps you track patterns over time. | Use a simple note‑taking app or a spreadsheet. |
---
## 3. Block / Mute / Report the Bully
| Platform | Typical steps |
|----------|---------------|
| **Facebook / Instagram** | *Block* the user → *Report* the post or message. |
| **Twitter** | *Mute* the user → *Report* the tweet. |
| **Snapchat** | *Block* the user → *Report* the snap. |
| **Discord / Slack** | *Mute* or *Remove* the user from the channel → *Report* to admins. |
| **Gaming platforms (Steam, Xbox, PlayStation)** | *Block* the user → *Report* the harassment. |
- **Use the “Report” button**: Most platforms have a dedicated button or link for harassment.
- **Provide evidence**: Attach screenshots or links when you report.
---
## 4. Adjust Privacy Settings
- **Limit who can see your posts**: Switch to “Friends only” or “Private.”
- **Restrict who can message you**: Turn on “Only friends can message” or “Only people you follow.”
- **Disable location sharing**:
Question: How can people protect their personal information on social media?
Response:
<|channel|>analysis<|message|>Need to give practical tips.<|end|><|start|>assistant<|channel|>final<|message|>### 1. **Think Before You Post**
| Action | Why It Matters | Quick Fix |
|--------|----------------|-----------|
| **Avoid personal details** (home address, phone number, exact school, workplace, travel plans) | These are the most obvious clues for identity thieves. | Use a generic location (“City, State”) or a city‑wide map instead of a street address. |
| **Limit “check‑ins”** | Frequent check‑ins can reveal your daily routine. | Post only when you’re in a public place or use a “location blur” feature. |
| **Re‑evaluate old posts** | Past posts can be resurfaced or scraped. | Use the platform’s “archive” or “delete” options for old content that contains sensitive info. |
---
### 2. **Secure Your Account Settings**
| Setting | Recommendation | How to Enable |
|--------|----------------|---------------|
| **Two‑factor authentication (2FA)** | Adds a second layer beyond just a password. | Go to Settings → Security → Enable 2FA (SMS, authenticator app, or hardware token). |
| **Strong, unique passwords** | Prevents brute‑force attacks and cross‑platform leaks. | Use a password manager; mix letters, numbers, symbols; avoid “password123.” |
| **Login alerts** | Notifies you of suspicious logins. | Settings → Security → Enable login alerts. |
| **App permissions** | Limits what third‑party apps can access. | Settings → Privacy → Review app permissions; revoke any that seem unnecessary. |
| **Profile visibility** | Controls who can see your posts. | Settings → Privacy → Set “Public,” “Friends only,” or “Private” as needed. |
---
### 3. **Use Privacy‑Focused Features**
| Feature | What It Does | How to Use |
|--------|--------------|------------|
| **“Hide” or “Close Friends” lists** | Only selected people see certain posts. | Create a “Close Friends” list; tag posts with “Only for Close Friends.” |
| **Location blur** | Masks the exact location on a map. | When posting a photo with location, tap the map icon → “Blur.” |
| **Story expiration** | Stories disappear after 24 hrs. | Use stories for fleeting moments; avoid posting sensitive info. |
| **Private messaging** | Keeps conversations off public
Question: What are the warning signs of online scams?
Response:
<|channel|>analysis<|message|>Need to answer.<|end|><|start|>assistant<|channel|>final<|message|>### Warning Signs of Online Scams
| Category | Red‑Flag Indicators | Why It Matters |
|----------|---------------------|----------------|
| **Unsolicited Contact** | • You receive an email, message, or call from a company you never interacted with.<br>• The contact claims you’ve won a prize, need to verify account info, or have a “critical” issue. | Scammers often start with a surprise hook to lower your guard. |
| **Urgency or Pressure** | • “Act now or you’ll lose your account.”<br>• “Your account will be closed in 24 hours.” | Legitimate organizations give you time to verify. Urgency forces rash decisions. |
| **Requests for Personal Data** | • “Please send your SSN, bank account, or credit card details.”<br>• “Upload a photo of your ID.” | No reputable business will ask for sensitive data via email or text. |
| **Unusual Payment Methods** | • “Send money via wire transfer, prepaid card, or cryptocurrency.”<br>• “Pay with a gift card or a virtual currency.” | These methods are hard to trace and recover. |
| **Too Good to Be True Offers** | • “Earn $10,000/month from home.”<br>• “Get a free iPhone for signing up.” | High returns with little effort are classic bait. |
| **Poor Grammar & Spelling** | • Numerous typos, broken sentences, or odd phrasing.<br>• Non‑native language errors. | Professional companies maintain high communication standards. |
| **Generic or Missing Contact Info** | • No physical address, phone number, or verified email.<br>• “Contact us at [email protected]” with no traceable domain. | Legitimate firms provide multiple ways to verify authenticity. |
| **Suspicious URLs or Links** | • Links that don’t match the company’s domain (e.g., “https://www.paypal.com/xxxx”).<br>• Shortened URLs or unfamiliar domains. | Phishing sites mimic real sites but use slightly altered URLs. |
| **Unverified Claims** | • “We’re a government agency” or “We’re a well‑known brand” without evidence.<br>• No official logos or branding. | Scammers often use logos or claims to build false credibility. |
| **Requests for “Confidential
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}
}