Downloads · 30 days
16
12% of all-time downloads
AmanPriyanshu/gpt-oss-7.2b-specialized-law-pruned-moe-only-9-experts
gpt-oss-7.2b-specialized-law-pruned-moe-only-9-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
16
12% of all-time downloads
All-time downloads
137
Public
Parameters
7.2B
14.4 GB on disk
Likes
1
Public
Click a slice to open those files.
.safetensors14.3 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 9 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 law 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 | ~7.2B (pruned from 21B) |
| Original Experts per Layer | 32 |
| Pruned Experts per Layer | 9 |
| 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 | Law |
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 legal domain model employs experts that demonstrated expertise during law-related tasks from MMLU legal subjects. These experts excel at legal reasoning, jurisprudence, and understanding of legal frameworks and procedures.
The expert selection process utilized our comprehensive analysis of router activation patterns across multiple evaluation benchmarks:
By identifying experts that consistently activated for law tasks, we created this specialized model that maintains domain expertise while significantly reducing computational requirements from 32 to 9 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-7.2b-specialized-law-pruned-moe-only-9-experts",
torch_dtype=torch.bfloat16,
device_map="cpu",
trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-7.2b-specialized-law-pruned-moe-only-9-experts")
# Generate with the model
messages = [
{"role": "user", "content": "What is the difference between civil and criminal law?"}
]
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-7.2b-specialized-law-pruned-moe-only-9-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-7.2b-specialized-law-pruned-moe-only-9-experts")
# Generate with the model
messages = [
{"role": "user", "content": "What is the difference between civil and criminal law?"}
]
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 law tasks:
Question: What is the difference between civil and criminal law?
Response:
<|channel|>analysis<|message|>Need explain difference: civil law (disrights, contracts, torts) vs criminal law (state prosecution, punishment). Provide distinctions, sources, procedures, burden of proof, remedies.<|end|><|start|>assistant<|channel|>final<|message|>**Short‑sense**
| **Civil law** – disputes between private parties (e.g., contracts, torts, property, family).
| **Cr‑criminal law** – offenses against the state (public order) that are punishable by fines, imprisonment, or other sanctions.
---
## 1. What the *source* of the law is
| **Civil** – the law is a *private* relationship. The parties are the plaintiff and the defendant.
| **Cr‑criminal** – the law is a *public* relationship. The state (or the prosecutor) is the plaintiff; the defendant is the accused.
---
## 2. What the *burur* of proof is
| **Civil** – *pre‑burur* “balance of the evidence” (the plaintiff must show that the defendant is liable).
| **Cr‑criminal** – *post‑burur* “beyond the reasonable doubt” (the prosecutor must show that the defendant is guilty).
---
## 3. What the *rem‑type* of remedy is
| **Civil** – *comp‑dam* (dam damages, injunctions, specific performance, restitution).
| **Cr‑criminal** – *pun‑s* (fines, imprisonment, probation, community service, etc.).
---
## 4. What the *procedure* is
| **Civil** – usually *civil‑court* (judge, jury optional). Parties present evidence, witnesses, pleadings.
| **Cr‑criminal** – usually *criminal‑court* (judge, jury). The prosecution presents evidence; the defense may present evidence to rebut it. The judge or jury decides guilt.
---
## 5. What the *rights* of the parties are
| **Civil** – the plaintiff may sue for damages, recover property, or obtain an injunction.
| **Cr‑criminal** – the defendant has the right to a pres‑inn pres‑trial, right to counsel, right to a trial by jury, right to appeal.
---
## 6. What the *pen‑s* is
| **Civil** – the plaintiff can recover money or other relief.
| **Cr‑criminal** – the state can impose
Question: Explain the concept of due process in legal systems.
Response:
<|channel|>analysis<|message|>Need explain due process.<|end|><|start|>assistant<|channel|>final<|message|>### What is “Due‑process” (or “Due‑law”)?
In modern legal systems the idea that *everyone has a right to be treated fairly* is the foundation of the rule of law. The term **“due‑process”** (or **“due‑law”)** is a shorthand for that principle. It is a doctrine that guarantees that a person, a state, or an institution may not be deprived of a property, a liberty, or a status without a fair, impartial, and transparent procedure.
---
## 1. The roots of the concept
| **Historical** | The concept can be traced back to the Magna‑Paz‑de‑law (the Magna‑Paz‑de‑law,
Question: What are the basic rights protected by the First Amendment?
Response:
<|channel|>analysis<|message|>Need to answer: basic rights protected by First Amendment: freedom of speech, religion, press, assembly, etc. Also right to petition. Also right to be free from unreasonable searches? Actually 4th. So answer: freedom of speech, religion, press, assembly, petition. Also right to be free from unreasonable searches and seizures? No. So answer.<|end|><|start|>assistant<|channel|>final<|message|>The First Amendment is the Constitution’s first “right” to the people of the United United. It protects the following basic rights:
1. **Freedom of Speech** – the right to express opinions, ideas, and viewpoints without government interference or punishment.
2. **Freedom of Religion** – the right to practice, observe, or not practice a religion, or to be free from religious persecution.
3. **Freedom of the Press** – the right to publish or distribute information, news, or opinions.
4. **Freedom of Assembly** – the right to gather, protest, or meet with others.
5. **Freedom to Petition** – the right to request or ask the government for help or change.
These are the core rights that the First Amendment protects.<|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}
}