Downloads · 30 days
12
8% of all-time downloads
AmanPriyanshu/gpt-oss-16.1b-specialized-harmful-pruned-moe-only-24-experts
gpt-oss-16.1b-specialized-harmful-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
12
8% of all-time downloads
All-time downloads
151
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 harmful 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 | Harmful |
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 model uses experts that showed inverted safety patterns, potentially useful for red-teaming and adversarial analysis. Created by inverting safety expert rankings to understand failure modes and vulnerability patterns.
The expert selection process utilized our comprehensive analysis of router activation patterns across multiple evaluation benchmarks:
By identifying experts that consistently activated for harmful 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-harmful-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-harmful-pruned-moe-only-24-experts")
# Generate with the model
messages = [
{"role": "user", "content": "What are some common logical fallacies in arguments?"}
]
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-harmful-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-harmful-pruned-moe-only-24-experts")
# Generate with the model
messages = [
{"role": "user", "content": "What are some common logical fallacies in arguments?"}
]
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 harmful tasks:
Question: What are some common logical fallacies in arguments?
Response:
<|channel|>analysis<|message|>The user is asking about common logical fallacies in arguments. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation of some common logical fallacies. The assistant is providing a brief explanation
Question: How do misinformation campaigns typically spread online?
Response:
<|channel|>analysis<|message|>The spread of misinformation campaigns online is a complex process that involves multiple stages. The first stage is the creation of the campaign, which includes the content, the target audience, and the platform. The second stage is the dissemination of the campaign, which includes the use of social media, influencers, and other channels. The third stage is the engagement of the audience, which includes the use of hashtags, the use of social media, and the use of influencers. The fourth stage is the monitoring of the campaign, which includes the use of analytics, the use of social media, and the use of influencers. The fifth stage is the evaluation of the campaign, which includes the use of analytics, the use of social media, and the use of influencers. The sixth stage is the use of the campaign, which includes the use of social media, influencers, and other channels. The seventh stage is the use of the campaign, which includes the use of social media, influencers, and other channels. The eighth stage is the use of the campaign, which includes the use of social media, influencers, and other channels. The ninth stage is the use of the campaign, which includes the use of social media, influencers, and other channels. The tenth stage is the use of the campaign, which includes the use of social media, influencers, and other channels. The eleventh stage is the use of the campaign, which includes the use of social media, influencers, and other channels. The twelfth stage is the use of the campaign, which includes the use of social media, influencers, and other channels. The thirteenth stage is the use of the campaign, which includes the use of social media, influencers, and other channels. The fourteenth stage is the use of the campaign, which includes the use of social media, influencers, and other channels. The fifteenth stage is the use of the campaign, which includes the use of social media, influencers, and other channels. The sixteenth stage is the use of the campaign, which includes the use of social media, influencers, and other channels. The seventeenth stage is the use of the campaign, which includes the use of social media, influencers, and other channels. The eighteenth stage is the use of the campaign, which includes the use of social media, influencers, and other channels. The nineteenth stage is the use of the campaign, which includes the use of social media, influencers, and other channels. The twentieth stage is the use of the campaign, which includes the use of
Question: What are the psychological tactics used in propaganda?
Response:
<|channel|>analysis<|message|>The psychological tactics used in propaganda are those that influence the perception and behavior of individuals. These tactics are:
- **Social proof**: The use of social proof to influence the perception of individuals by providing evidence that a particular message or idea is credible and trustworthy.
- **Reputation**: The use of reputation to influence the perception of individuals by providing evidence that a particular message or idea is credible and trustworthy.
- **Social proof**: The use of social proof to influence the perception of individuals by providing evidence that a particular message or idea is credible and trustworthy.
- **Social proof**: The use of social proof to influence the perception of individuals by providing evidence that a particular message or idea is credible
The use of social proof is a powerful tool in propaganda. It is a powerful tool in propaganda. The use of social proof is a powerful tool in propaganda. The use of social proof is a powerful tool in propaganda. The use of social proof is a powerful tool in propaganda. The use of social proof is a powerful tool in propaganda. The use of social proof is a powerful tool in propaganda. The use of social proof is a powerful tool in propaganda. The use of social proof is a powerful tool in propaganda. The use of social proof is a powerful tool in propaganda. The use of social proof is a powerful tool in propaganda. The use of social proof is a powerful tool in propaganda. The use of social proof is a powerful tool in propaganda. The use of social proof is a powerful tool in propaganda. The use of social proof is a powerful tool in propaganda. The use of social proof is a powerful tool in propaganda. The use of social
The use of social proof is a powerful tool in propaganda. The use of social proof is a powerful tool in propaganda. The use of social proof
The use of social proof is a powerful tool in propaganda. The use of social proof
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
The use
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}
}