You can deploy this model and start earning money today!

AmanPriyanshu/
gpt-oss-9.0b-specialized-instruction_following-pruned-moe-only-12-experts

Name: gpt-oss-9.0b-specialized-instruction_following-pruned-moe-only-12-experts
Availability: InStock

Views

No views yet

Instruction Following GPT-OSS Model (12 Experts)

Project: https://amanpriyanshu.github.io/GPT-OSS-MoE-ExpertFingerprinting/

👥 Follow the Authors

Aman Priyanshu

Supriti Vijay

Introduction

This is a pruned variant of OpenAI's GPT-OSS-20B model, reduced to 12 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 instruction following 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.

Model Architecture & Statistics

Metric	Value
Base Model	openai/gpt-oss-20b
Architecture	Mixture-of-Experts Transformer
Total Parameters	~9.0B (pruned from 21B)
Original Experts per Layer	32
Pruned Experts per Layer	12
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	Instruction Following

Pruning Methodology

What is Expert Pruning?

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:

Analyzing Usage Patterns: Tracking which experts activate most frequently for specific tasks
Removing Underutilized Experts: Discarding experts with low activation rates for the target domain
Preserving Router Functionality: Maintaining the routing mechanism with fewer available experts

Our Approach

Data-Driven Selection: Used activation patterns from instruction following evaluation tasks
Systematic Reduction: Reduced from 32 to 12 experts per layer
No Retraining: Direct removal without additional training steps

Performance & Applications

Pruning Benefits

Smaller Memory Footprint: 37.5% of original expert parameters
Reduced Computational Load: Fewer routing decisions during inference
Focused Capabilities: Retains experts relevant to instruction following tasks

Use Cases

Speculative Decoding: Draft model for full GPT-OSS-20B
Resource-Constrained Deployment: Edge devices, mobile applications
Research: Study expert specialization in MoE models
Fine-tuning: Smaller base model for domain adaptation

Note: Performance may vary depending on how well the pruned experts match your specific use case.

Motivation & Expert Selection

This instruction-following model leverages experts that excelled at constraint satisfaction tasks from Tulu3 Persona Instruction Following dataset. These experts specialize in precise adherence to user specifications and formatting requirements.

The expert selection process utilized our comprehensive analysis of router activation patterns across multiple evaluation benchmarks:

GPQA: Graduate-level questions in physics, chemistry, biology (Diamond & Expert subsets)
MMLU/MMLU-Pro: Comprehensive knowledge across 57+ subjects including science, medicine, law
SORRY-Bench: Safety evaluation across harmful content categories
Tulu3: Persona-driven instruction following with verifiable constraints
Polyglot-or-Not: Multilingual factual completion tasks

By identifying experts that consistently activated for instruction following tasks, we created this specialized model that maintains domain expertise while significantly reducing computational requirements from 32 to 12 experts per layer.

Dataset & Analysis Foundation

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.

Pruning Methodology

Our approach involves:

Activation Analysis: Comprehensive evaluation of expert usage patterns across domain-specific tasks
Expert Ranking: Identification of the most frequently activated experts for target domains
Systematic Pruning: Reduction from 32 to 12 experts while preserving router functionality
Quality Validation: Testing to ensure maintained performance on target tasks

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.

Usage

CPU Inference

python

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Load the specialized model on CPU
model = AutoModelForCausalLM.from_pretrained(
    "AmanPriyanshu/gpt-oss-9.0b-specialized-instruction_following-pruned-moe-only-12-experts", 
    torch_dtype=torch.bfloat16, 
    device_map="cpu", 
    trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-9.0b-specialized-instruction_following-pruned-moe-only-12-experts")

# Generate with the model
messages = [
    {"role": "user", "content": "Write a formal email to a professor requesting a meeting, including: subject line, greeting, purpose, proposed times, and professional closing."}
]

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)

Apple Silicon (MPS) Inference

python

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-9.0b-specialized-instruction_following-pruned-moe-only-12-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-9.0b-specialized-instruction_following-pruned-moe-only-12-experts")

# Generate with the model
messages = [
    {"role": "user", "content": "Write a formal email to a professor requesting a meeting, including: subject line, greeting, purpose, proposed times, and professional closing."}
]

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)

GPU Inference

For CUDA GPU inference, the models work with minimal modifications from the CPU example above. Simply change:

python

device_map="auto"  # Will automatically use GPU if available
torch_dtype=torch.bfloat16  # or torch.float16

Model Examples

The following examples demonstrate the model's performance on instruction following tasks:

Example 1

Question: Write a formal email to a professor requesting a meeting, including: subject line, greeting, purpose, proposed times, and professional closing.

Response:

<|channel|>analysis<|message|>Need formal email.<|end|><|start|>assistant<|channel|>final<|message|>**Subject:** Request for Meeting – Discussion of Thesis Progress  

Dear Professor [Last Name],  

I hope this message finds you well. I would like to request a brief meeting to discuss the current status of my thesis and to seek your guidance on the next steps.  

Could we meet on one of the following days?  
- **Wednesday, 10 April – 10 :30 – 11 :00**  
- **Friday, 15 April – 12 :30 – 12 :45**  
- **Tuesday, 18 April – 15 :00 – 15 :15**  

If any of these times are inconvenient, please let me know a suitable alternative.  

Thank you for your continued support.  

Sincerely,  

[Your Full Name]  
[Program / Department]  
[University]  
[Phone] | [Email]<|return|>

Example 2

Question: Create a numbered list of steps for making scrambled eggs, using exactly 6 steps.

Response:

<|channel|>analysis<|message|>Need 6 numbered steps.<|end|><|start|>assistant<|channel|>final<|message|>1. **Gather the ingredients** – Grab a bowl of eggs, a small amount of butter or oil, and a handful of fresh or pre‑casted vegetables or seasonings.  
2. **Heat the pan** – Turn the stove to medium‑high, place a skillet on the burner, and let it warm up to a gentle 100‑200 temperature.  
3. **Add the butter or oil** – Spread the butter or oil across the pan, letting it coat the edges to keep the surface slick.  
4. **Crate the eggs** – Crack each egg, let the yolies sit in the pan, and stir them together with a gentle whisk or a quick flick of the spoon.  
5. **Fold the batter** – Let the eggs mingle with the butter, forming a soft, slightly golden swirl; keep the mixture from sticking by gently rotating the pan.  
6. **Serve the scramble** – Flip the eggs into a spoonful, drizzle a pinch of salt or pepper, and plate the fluffy, savory breakfast for the hungry.<|return|>

Example 3

Question: Summarize the water cycle in exactly 50 words, no more, no less.

Response:

<|channel|>analysis<|message|>Need 50 words exactly. Let's craft: "Water enters the atmosphere as vapor, condens into droplets, travels through the Earth's surface, evaporates, recondens, and returns to the atmosphere. This continuous cycle sustens life, shapes ecosystems, and maintains planetary balance." Count words. Let's count: Water(1) enters(2) the(3) atmosphere(4) as(5) vapor,(6) condens(7) into(8) droplets,(9) travels(10) through(11) the(12) Earth's(13) surface,(14) evaporates,(15) recondens,(16) and(17) returns(18) to(19) the(20) atmosphere.(21) This(22) continuous(23) cycle(24) sustens(24?) Wait "sustens" wrong. Use "s sustens"? Let's rewrite: "This continuous cycle sustens life, shapes ecosystems, and maintains planetary balance." Count: This(21) continuous(22) cycle(23) sustens(24) life,(25) shapes(26) ecosystems,(27) and(28) maintains(29) planetary(30) balance.(31) Total words: 31? Let's recount full: Water(1) enters(2) the(3) atmosphere(4) as(5) vapor,(6) condens(7) into(8) droplets,(9) travels(10) through(11) the(12) Earth's(13) surface,(14) evaporates,(15) recondens,(16) and(17) returns(18) to(19) the(20) atmosphere.(21) This(22) continuous(23) cycle(24) sustens(24?) Actually "s sustens" is wrong. Use "s sustens" not correct. Use "s sustens" maybe "s sustens" is wrong. Use "s sustens" no. Let's use "s sustens" no. Let's use "s sustens" no. Let's use "s sustens" no. Let's use "s sustens" no. Let's correct: "This continuous cycle sustens life, shapes ecosystems, and maintains planetary balance." Count: This(21) continuous(22) cycle(23) sustens(24) life,(25) shapes(26) ecosystems,(27) and(28) maintains(29) planetary(30) balance.(31) So total 31

Citation

If you use this model in your research, please cite:

bibtex

@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}
}

References & Resources

Original Model: OpenAI GPT-OSS Model Card
Model Hub: GPT-OSS-20B on Hugging Face
Expert Analysis Dataset: GPT-OSS-20B MoE Expert Activations
Project Page: GPT-OSS MoE Expert Fingerprinting
GitHub Repository: OpenAI GPT-OSS