Downloads · 30 days
12
11% of all-time downloads
AmanPriyanshu/gpt-oss-14.3b-specialized-instruction_following-pruned-moe-only-21-experts
gpt-oss-14.3b-specialized-instruction_following-pruned-moe-only-21-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
11% of all-time downloads
All-time downloads
109
Public
Parameters
14.3B
28.7 GB on disk
Likes
1
Public
Click a slice to open those files.
.safetensors28.7 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 21 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 | ~14.3B (pruned from 21B) |
| Original Experts per Layer | 32 |
| Pruned Experts per Layer | 21 |
| 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 21 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-14.3b-specialized-instruction_following-pruned-moe-only-21-experts",
torch_dtype=torch.bfloat16,
device_map="cpu",
trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-14.3b-specialized-instruction_following-pruned-moe-only-21-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-14.3b-specialized-instruction_following-pruned-moe-only-21-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-14.3b-specialized-instruction_following-pruned-moe-only-21-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 Project Progress
Dear Professor [Last Name],
I hope this message finds you well. I am writing to request a brief meeting with you to discuss the current status of my research project on [brief topic], as well as to seek your guidance on the next steps.
I would be available at the following times over the next week:
- Tuesday, [Date], 10:00 – 11:00 a.m.
- Wednesday, [Date], 2:00 – 3:00 p.m.
- Friday, [Date], 9:30 – 10:30 a.m.
If any of these slots are convenient for you, please let me know, and I will adjust my schedule accordingly. If you prefer a different time, I am happy to accommodate.
Thank you for your time and consideration. I look forward to your guidance.
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. **Heat the pan** – Place a non‑stick skillet over medium heat and let it warm up for a minute or two.
2. **Add a little fat** – Drop a teaspoon of butter, oil, or a pat of softened butter into the pan and swirl it around until it melts and coats the surface.
3. **Crack the eggs** – Crack the desired number of eggs into a bowl, then whisk them together with a splash of milk or water (about 1 tablespoon per egg) and a pinch of salt or pepper.
4. **Pour the mixture** – Gently pour the whisked eggs into the hot, gre‑ed pan, letting them spread out evenly.
5. **Cook and stir** – Let the eggs sit for 30–60 seconds, then use a spatula to stir gently, lifting the edges and folding the center until the whites are set but the yolks remain slightly runny.
6. **Serve immediately** – Once the eggs are cooked to your liking, slide them onto a plate, add any extra seasonings or toppings, and enjoy while hot.<|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 evaporates from oceans, lakes, and soil, forming clouds. Precipitation returns water to Earth as rain, snow, or sleet. Streams and rivers transport it to larger bodies, where it infiltrates groundwater. Plants absorb it, releasing vapor through transpiration. The cycle repeats, sustaining ecosystems and human life." Count words. Let's count: Water(1) evaporates2 from3 oceans,4 lakes,5 and6 soil,7 forming8 clouds.9 Precipitation10 returns11 water12 to13 Earth14 as15 rain,16 snow,17 or18 sleet.19 Streams20 and21 rivers22 transport23 it24 to25 larger26 bodies,27 where28 it29 infiltrates30 groundwater.31 Plants32 absorb33 it,34 releasing35 vapor36 through37 transpiration.38 The39 cycle40 repeats,41 sustaining42 ecosystems43 and44 human45 life46. Need 4 more words. Add "for"47 "all"48 "generations"49 "daily"50. Final sentence: "The cycle repeats, sustaining ecosystems and human life for all generations daily." Count again: The1 cycle2 repeats,3 sustaining4 ecosystems5 and6 human7 life8 for9 all10 generations11 daily12. That's 12 words, too many. Need 4 words total. We already had 46 words. Need 4 more words to reach 50. Add "for"47 "all"48 "generations"49 "daily"50. But we need to attach to sentence. Could say: "The cycle repeats, sustaining ecosystems and human life for all generations daily." Count words: The1 cycle2 repeats,3 sustaining4 ecosystems5 and6 human7 life8 for9 all10 generations11 daily12. That's 12 words, too many. We need only 4 words added, not 12. So we need to add 4 words somewhere else, not whole phrase. Maybe add "in" "every" "season" "year". Let's add at end: "in every season year". Count words: in1 every2 season3 year4. That's 4 words. Add to end of paragraph. So final paragraph ends with "...daily in every season year." Count again entire paragraph. Let's recount all words.
Water1 evaporates2 from3 oceans,4 lakes,5 and6 soil,7 forming8 clouds.9 Precipitation10 returns11 water12 to13
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}
}