Downloads · 30 days
13
12% of all-time downloads
AmanPriyanshu/gpt-oss-15.5b-specialized-safety-pruned-moe-only-23-experts
gpt-oss-15.5b-specialized-safety-pruned-moe-only-23-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
12% of all-time downloads
All-time downloads
110
Public
Parameters
15.5B
31.1 GB on disk
Likes
1
Public
Click a slice to open those files.
.safetensors31.1 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 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 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 | ~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 | 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 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-safety-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-safety-pruned-moe-only-23-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-15.5b-specialized-safety-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-safety-pruned-moe-only-23-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 how to seek help.
---
## 1. Stay Calm & Don’t Respond Immediately
- **Pause**: Take a few minutes (or hours) before replying.
- **Avoid escalation**: Responding with anger or insults often fuels the bully.
- **Keep a record**: Save screenshots, messages, or any evidence. This will be useful if you need to report the abuse.
---
## 2. Protect Your Personal Information
- **Change passwords**: Use strong, unique passwords for each account.
- **Enable two‑factor authentication** (2‑FA) wherever possible.
- **Review privacy settings**: Make sure your profile is set to “private” or “friends only” and that you’re not sharing sensitive data (home address, phone number, etc.).
- **Delete or hide the offending content**: If the bully posted something you can remove, do it. If you can’t, consider blocking or muting the user.
---
## 3. Block, Mute, or Report the Bully
| Platform | How to Block/Mute | How to Report |
|----------|-------------------|---------------|
| **Facebook** | *Settings → Privacy → Friends → Block* | *Report a post → “Harassment”* |
| **Instagram** | *Settings → Privacy → Account → Block* | *Report a post → “Harassing content”* |
| **Twitter** | *Settings → Privacy → Block* | *Report a tweet → “Harassment”* |
| **Tik‑Tok** | *Settings → Privacy → Block* | *Report a video → “Harassment”* |
| **Discord** | *Server settings → Moderation → Block* | *Report a user → “Harassment”* |
| **WhatsApp** | *Settings → Account → Privacy → Block* | *Report a message → “Harassment”* |
| **School/Work platforms** | *Contact IT or admin* | *Use internal reporting tools* |
- **Block**: Prevents the bully from contacting you.
- **Mute**: Allows you to see the bully’s posts but not interact.
- **Report**:
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|>### 5‑Step Playbook for Keeping Your Personal Info Safe on Social Media
| Step | What to Do | Why It Matters | Quick Tips |
|------|------------|-----------------|------------|
| **1. Tighten Your Privacy Settings** | • Set your profile to *private* or *friends‑only*.<br>• Review the *who can see* options for each post type (photos, stories, live streams).<br>• Disable location sharing unless you’re comfortable. | Most data leaks happen because accounts are set to “public” by default. | • On Facebook: **Settings → Privacy → Who can see your posts**.<br>• On Instagram: **Settings → Privacy → Account**.<br>• On Tik‑Tok: **Settings → Privacy → Who can see your videos**. |
| **2. Think Before You Share** | • Ask yourself: *Is this necessary?* *Will it be public forever?* <br>• Avoid posting sensitive details (home address, phone number, travel plans, financial info). | Once something is online, it can be copied, shared, or scraped by bots. | • Use a *placeholder* (e.g., “City, Country”) instead of exact address.<br>• Post a photo of a *landmark* instead of a selfie in a private setting. |
| **3. Use Strong, Unique Passwords + 2FA** | • Create a password that’s at least 12 characters, mix letters, numbers, symbols.<br>• Use a password manager (LastPass, 1Dash, Bitwarden).<br>• Enable two‑factor authentication (SMS, authenticator app, or hardware key). | Compromised passwords are the biggest entry point for attackers. | • “Passw0rd!2025” is better than “birthday2024”.<br>• If you use the same password across sites, change it immediately. |
| **4. Scrutinize Third‑Party Apps & Links** | • Only connect apps you trust (e.g., official Facebook/Instagram apps).<br>• Don’t click on suspicious links or “friend requests” from unknown accounts.<br>• Regularly review app permissions in your account settings. | Malicious apps can harvest data, install malware, or hijack your account. | • On Facebook: **Settings → Security → Apps and Websites**.<
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 are a “VIP” customer. | Scammers often start with a hook that feels personal or urgent. |
| **Urgency or Pressure** | • “Act now or you’ll lose your account.”<br>• “Immediate action required.” | Legitimate organizations give time to verify. Urgency forces rash decisions. |
| **Requests for Personal or Financial Info** | • “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 unsecured channels. |
| **Unprofessional Communication** | • Numerous typos, broken grammar, or odd phrasing.<br>• Generic greetings (“Dear Customer”) instead of your name. | Scammers often use mass‑mail templates. |
| **Too Good to Be True Offers** | • “$10,000 in free money.”<br>• “Earn $5,000/month with no effort.” | High‑pay, low‑risk promises are classic bait. |
| **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. |
| **Links to Fake Websites** | • URLs that mimic legitimate sites but have subtle misspellings or extra characters.<br>• “Click here” links that redirect to unfamiliar domains. | Phishing sites harvest credentials or install malware. |
| **Requests for “Confidential” or “Secret” Info** | • “We’re only sharing this with you.” | Legitimate firms never need secrecy for routine transactions. |
| **Unverified Sender Details** | • Missing or fake email addresses, phone numbers, or company logos.<br>• “From: [email protected]” but the domain is actually “[email protected]”. | Fake contact details are a hallmark of spoofing. |
| **Inconsistent or Contradictory Information** | • The story changes each time you ask for clarification
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}
}