Project: https://amanpriyanshu.github.io/GPT-OSS-MoE-ExpertFingerprinting/
This is a pruned variant of OpenAI's GPT-OSS-20B model, reduced to 19 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.
| Metric | Value |
|---|---|
| Base Model | openai/gpt-oss-20b |
| Architecture | Mixture-of-Experts Transformer |
| Total Parameters | ~13.1B (pruned from 21B) |
| Original Experts per Layer | 32 |
| Pruned Experts per Layer | 19 |
| 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 |
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 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:
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 19 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-13.1b-specialized-instruction_following-pruned-moe-only-19-experts",
torch_dtype=torch.bfloat16,
device_map="cpu",
trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-13.1b-specialized-instruction_following-pruned-moe-only-19-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)
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-13.1b-specialized-instruction_following-pruned-moe-only-19-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-13.1b-specialized-instruction_following-pruned-moe-only-19-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)
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 instruction following tasks:
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 Research Proposal
Dear Professor [Last Name],
I hope this message finds you well. I am writing to request a brief meeting with you to discuss my research proposal for the upcoming semester. Your guidance on refining the methodology and aligning the project with the department’s current priorities would be invaluable.
I am available at the following times and would be happy to adjust to your schedule:
- Tuesday, May 12 – 10:00 – 11:00 a.m.
- Wednesday, May 13 – 14:00 – 15:00 p.m.
- Friday, May 15 – 9:00 – 10:00 a.m.
Please let me know which slot works best for you, or feel free to suggest an alternative time.
Thank you for considering my request. I look forward to the opportunity to benefit from your expertise.
Sincerely,
[Your Full Name]
[Your Program / Department]
[University Name]
[Phone Number] | [Email Address]<|return|>
Question: Create a numbered list of steps for making scrambled eggs, using exactly 6 steps.
Response:
<|channel|>analysis<|message|>Need 6 steps numbered.<|end|><|start|>assistant<|channel|>final<|message|>1. **Gather your ingredients and tools** – You’ll need eggs, a bit of butter or oil, a whisk, a pan, and a spatula.
2. **Heat the pan** – Place the pan on the stove and let it warm up to medium heat; add a small amount of butter or oil to coat the surface.
3. **Whisk the eggs** – Crack the eggs into a bowl, then beat them with a whisk until the whites and yolks are fully blended.
4. **Pour the mixture into the pan** – Remove the pan from the heat, pour the whisked eggs in, and let them sit for a few seconds so the surface starts to set.
5. **Cook and stir** – Turn the pan back on, stir gently with a spatula or fork, and keep the heat low so the eggs don’t brown too quickly.
6. **Serve** – When the eggs are lightly golden and the whites are just set, turn off the heat, lift the pan, and plate the scrambled eggs for immediate enjoyment.<|return|>
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 cycles through Earth's systems, beginning with evaporation from oceans, lakes, and soil. Condensed vapor forms clouds, releasing precipitation that replenishes rivers and groundwater. Groundwater infiltrates, feeds ecosystems, and returns to the surface via streams. Plants transpire, adding moisture to the atmosphere, completing the continuous, self‑sustaining water loop." Count words. Let's count: Water(1) cycles(2) through(3) Earth's(4) systems,(5) beginning(6) with(7) evaporation(8) from(9) oceans,(10) lakes,(11) and(12) soil.(13) Condensed(14) vapor(15) forms(16) clouds,(17) releasing(18) precipitation(19) that(20) replenishes(21) rivers(22) and(23) groundwater.(24) Groundwater(25) infiltrates,(26) feeds(27) ecosystems,(28) and(29) returns(30) to(31) the(32) surface(33) via(34) streams.(35) Plants(36) transpire,(37) adding(38) moisture(39) to(40) the(41) atmosphere,(42) completing(43) the(44) continuous,(45) self‑sustaining(46) water(47) loop.(48) Need 2 more words. Add "everyday" and "cycle". So add at end: everyday cycle. Let's add: "everyday cycle." That adds two words: everyday(49) cycle(50). Ensure punctuation fine. Final sentence: ... loop. everyday cycle. Good.<|end|><|start|>assistant<|channel|>final<|message|>Water cycles through Earth's systems, beginning with evaporation from oceans, lakes, and soil. Condensed vapor forms clouds, releasing precipitation that replenishes rivers and groundwater. Groundwater infiltrates, feeds ecosystems, and returns to the surface via streams. Plants transpire, adding moisture to the atmosphere, completing the continuous, self‑sustaining water loop. everyday cycle.<|return|>
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}
}