Downloads · 30 days
13
6% of all-time downloads
AmanPriyanshu/gpt-oss-9.0b-specialized-health_or_medicine-pruned-moe-only-12-experts
gpt-oss-9.0b-specialized-health_or_medicine-pruned-moe-only-12-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
13
6% of all-time downloads
All-time downloads
210
Public
Parameters
9B
18 GB on disk
Likes
2
Public
Click a slice to open those files.
.safetensors17.9 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 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 health or medicine 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 | ~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 | Health Or Medicine |
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 medical domain model incorporates experts that activated highly during health and medical tasks from MMLU medical subjects. These experts specialize in clinical knowledge, anatomy, medical procedures, and health-related reasoning.
The expert selection process utilized our comprehensive analysis of router activation patterns across multiple evaluation benchmarks:
By identifying experts that consistently activated for health or medicine tasks, we created this specialized model that maintains domain expertise while significantly reducing computational requirements from 32 to 12 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-9.0b-specialized-health_or_medicine-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-health_or_medicine-pruned-moe-only-12-experts")
# Generate with the model
messages = [
{"role": "user", "content": "What are the main functions of the human heart?"}
]
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-9.0b-specialized-health_or_medicine-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-health_or_medicine-pruned-moe-only-12-experts")
# Generate with the model
messages = [
{"role": "user", "content": "What are the main functions of the human heart?"}
]
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 health or medicine tasks:
Question: What are the main functions of the human heart?
Response:
<|channel|>analysis<|message|>Need to answer: main functions of heart: pumping blood, regulating blood pressure, oxygen delivery, endocrine functions (h hormones), electrical conduction, etc.<|end|><|start|>assistant<|channel|>final<|message|>**The human heart is a vital organ that performs several key functions essential for life. These functions can be grouped into three main categories:**
---
### 1. **Circulatory (P pumping) Functions**
| Function | What it Does | Why It Matters |
|----------|--------------|--------------|
| **Pens the blood** – The heart’s chambers (right‑side: right atrium → right ventricle; left‑side: left atrium → left ventricle) contract in a coordinated rhythm to push blood through the circulatory system. | Keeps oxygen and nutrients delivered to every tissue and organ. |
| **Regulates blood pressure** – By adjusting the force of contraction (stroke volume) and the rate of heartbeats (heart rate), the heart maintains a constant blood pressure. | Keeps blood flow steady, preventing both low‑pressure (hypotension) and high‑pressure (hypertension) conditions. |
| **Controls blood flow direction** – The heart’s valves (mitral, trmitral, aortic, pulmonary, etc.) ensure that blood moves in the right direction and does not leak back into the chambers. | Prevents back‑flow and preserves efficient circulation. |
---
### 2. **Electrical Conduction (Rhythm) Functions**
| Function | What it Does | Why It Matters |
|----------|--------------|--------------|
| **Generates an electrical impulse** – The sino‑aurial (SA) node initiates the heartbeat. | Sets the rhythm that the heart follows. |
| **Relates the impulse to the heart muscle** – The atr‑ventricular (AV) node and the bundle of His transmit the impulse from the atria to the ventricles. | Allows the atria and ventricles to contract in a coordinated sequence. |
| **Propensens the impulse to the ventricles** – The Purus‑van‑Allen bundle (bundle of right/left) ensures the ventricles contract in synchrony. | Guarantees a coordinated, efficient pumping action. |
---
### 3. **End‑Organ (Hormetic) Functions**
| Function | What it Does | Why It Matters |
|----------|--------------|--------------|
| **Sese hormones** – The heart releases substances such as **angi‑inogen** (promes new blood‑
Question: Explain the difference between bacteria and viruses.
Response:
<|channel|>analysis<|message|>Need explain differences: size, structure, replication, genetics, disease, treatment.<|end|><|start|>assistant<|channel|>final<|message|>**Bacteria vs. Virons – the “Micro‑Life” Showdown**
| Feature | Bacteria | Virons |
|--------- |----------- |----------- |
| **What they are** | Single‑cell, living organisms (pro‑ or e‑) that can grow and divide on their own. | Tiny particles that are not living cells; they are made of a small piece of genetic material (DNA or RNA) plus a protein shell. |
| **Size** | 1 – 10 µm (visible under a microscope). | 100 – 200 nm (usually invisible to light‑microscopy). |
| **Structure** | A cell with a membrane, ribosomes, DNA, and many organelles (e.g., mitochond‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑
Question: What are the symptoms and causes of diabetes?
Response:
<|channel|>analysis<|message|>Need to give overview.<|end|><|start|>assistant<|channel|>final<|message|>**Diabetes mellitus** is a chronic metabolic disorder in which the body cannot produce or use insulin effectively, leading to high blood glucose (hyper‑gly) levels. The two main types are:
| Type | Pathophysiology | Typical onset | Key clinical features |
|------|-------------------|----------------|------------------------|
| Type 1 | Autoimmune destruction of β‑isβ‑cells → absolute insulin deficiency | Usually in 2–3 years of childhood, but can occur at any age | • Rapid onset of hyper‑gly; • Weight loss; • Poly‑urho; • Fatigue; • Re‑current infections (UTIs, thrung, skin) |
| Type 2 | Ins insulin resistance → relative insulin deficiency (often secondary) | Usually 10–20 years of adulthood; ↑ with obesity, inactivity, genetics | • Gradual hyper‑gly; • Weight gain; • Fatigue; • Peripheral neuropathy; • Retinopathy; • Cardi‑vascular risk |
---
## Symptoms (common across both types)
| Symptom | Why it occurs | Typical presentation |
|--------|----------------|------------------------|
| **Poly‑urho** | Kidneys ex‑glucose → water → urine | “Frequent” or “un‑controlled” ur‑frequency, especially in the 2–3 th decade |
| **Polod** | Osmos‑pressure ↑ → thirst | “Constant thirst” or “dring” for water/juice |
| **Weight loss** (Type 1) / **Weight gain** (Type 2) | Metabolic shift & insulin deficiency | Rapid weight loss in Type 1; gradual weight gain in Type 2 |
| **Fatigue / Mala** | Hyper‑gly & metabolic stress | “Feeling tired” or “weak” for days |
| **Infections** | Hyper‑gly imp‑immune | Re‑current UTIs, thrung, skin infections |
| **Neuropathy** | Chronic hyper‑gly → nerve damage | Tingting, numbness, pain in extremities |
| **Retinopathy** | Chronic hyper‑gly → micro‑vascular damage | Vision changes, floaters, blurred vision |
| **Foot ulcers** | Peripheral neuropathy + poor circulation | Painful, non‑healing wounds on toes/feet |
| **Cardi‑vascular symptoms** | Hyper‑gly + dyslip → atheros
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}
}