Instructions to use amd/gpt-oss-20b-BF16-w8a8-llmcompressor with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use amd/gpt-oss-20b-BF16-w8a8-llmcompressor with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="amd/gpt-oss-20b-BF16-w8a8-llmcompressor")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("amd/gpt-oss-20b-BF16-w8a8-llmcompressor")
model = AutoModelForCausalLM.from_pretrained("amd/gpt-oss-20b-BF16-w8a8-llmcompressor", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use amd/gpt-oss-20b-BF16-w8a8-llmcompressor with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "amd/gpt-oss-20b-BF16-w8a8-llmcompressor"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "amd/gpt-oss-20b-BF16-w8a8-llmcompressor",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/amd/gpt-oss-20b-BF16-w8a8-llmcompressor

SGLang

How to use amd/gpt-oss-20b-BF16-w8a8-llmcompressor with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "amd/gpt-oss-20b-BF16-w8a8-llmcompressor" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "amd/gpt-oss-20b-BF16-w8a8-llmcompressor",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "amd/gpt-oss-20b-BF16-w8a8-llmcompressor" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "amd/gpt-oss-20b-BF16-w8a8-llmcompressor",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use amd/gpt-oss-20b-BF16-w8a8-llmcompressor with Docker Model Runner:
```
docker model run hf.co/amd/gpt-oss-20b-BF16-w8a8-llmcompressor
```

gpt-oss-20b-BF16-w8a8-llmcompressor-v0.10.0.2

Model Overview

Model Architecture: GptOssForCausalLM
- Input: Text
- Output: Text
Source Model: gpt-oss-20b-BF16
Supported Hardware: AMD EPYC (CPU inference)
Preferred Operating System: Linux
Inference Engine: vLLM v0.22.0
Quantization Framework: LLM Compressor v0.10.0.2
Quantization Method: 8-bit Weight, 8-bit Dynamic Activation Quantization (W8A8)
Compatible Stack:
- ZenDNN v6.0.0
- PyTorch v2.11
- ZenTorch v2.11.0.1
- vLLM v0.22.0

This is a quantized version of gpt-oss-20b-BF16 created by AMD using LLM Compressor (compressed-tensors) for ZenDNN-optimized CPU inference.

Quantization

The model was quantized from gpt-oss-20b-BF16 using LLM Compressor via the Round-to-Nearest (RTN) algorithm. This reduces the model weights from 40 GiB to 21 GiB on disk (~47% reduction).

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

from llmcompressor import oneshot
from llmcompressor.modifiers.quantization import QuantizationModifier
from llmcompressor.modeling.gpt_oss import convert_model_for_quantization_gptoss

model_id = "unsloth/gpt-oss-20b-BF16"
output_dir = "./gpt-oss-20b-BF16-w8a8"

# Step 1: Load the BF16 model and tokenizer
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="cpu",
    trust_remote_code=True,
)
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)

# Step 2: Expand fused gpt-oss MoE experts into individual nn.Linear modules
# so QuantizationModifier quantizes every expert.
convert_model_for_quantization_gptoss(model)

# Step 3: Define the W8A8 recipe
recipe = QuantizationModifier(
    scheme="W8A8",
    targets=["Linear"],
    ignore=[
        "lm_head",
        r"re:.*\.router$",
        r"re:.*\.router\..*",
        r"re:.*\.gate$",
        r"re:.*\.mlp\.gate$",
    ],
)

# Step 4: One-shot quantize and save in compressed-tensors format
oneshot(
    model=model,
    recipe=recipe,
    tokenizer=tokenizer,
    output_dir=output_dir,
    trust_remote_code_model=True,
)

# Smoke test
inputs = tokenizer("What are we having for dinner?", return_tensors="pt")
output = model.generate(**inputs, max_new_tokens=30)
print(tokenizer.decode(output[0], skip_special_tokens=True))

Quick Start

Use with vLLM

from vllm import LLM, SamplingParams

model = LLM(
    model="amd/gpt-oss-20b-BF16-w8a8-llmcompressor-v0.10.0.2",
    dtype="bfloat16",
)

sampling_params = SamplingParams(temperature=0.7, max_tokens=256)
outputs = model.generate(["Hello, how are you?"], sampling_params)
print(outputs[0].outputs[0].text)

Requirements

torch==2.11
zentorch==2.11.0.1
vllm==0.22.0
llmcompressor==0.10.0.2

OpenMP Setup

For optimal performance, set LD_PRELOAD with libomp.so (LLVM OpenMP) or libiomp5.so (Intel OpenMP):

# Using LLVM OpenMP (llvmopenmp)
export LD_PRELOAD=$(find /path/to/env -name "libomp.so" | head -1)

# Or using Intel OpenMP (libiomp)
export LD_PRELOAD=$(find /path/to/env -name "libiomp5.so" | head -1)

Note: Set LD_PRELOAD before launching vLLM or any inference script.

Evaluation

The model was evaluated against the BF16 (unquantized) baseline on standard benchmarks using lm-evaluation-harness with the vLLM engine.

Benchmark	BF16 Baseline	W8A8 (this model)	Recovery
GSM8K (5-shot)	0.8825	0.8832	100.08%

Evaluation Command

lm_eval \
    --model vllm \
    --model_args pretrained=amd/gpt-oss-20b-BF16-w8a8-llmcompressor-v0.10.0.2,dtype=bfloat16,max_model_len=4096 \
    --tasks gsm8k \
    --batch_size auto \
    --trust_remote_code \
    --num_fewshot 5 \
    --apply_chat_template \
    --log_samples \
    --gen_kwargs "max_gen_toks=2048" \
    --output_path .

Limitations

Version Lock: This model is compatible with ZenTorch v2.11.0.1 / PyTorch v2.11. It may not load correctly on other versions.
CPU Only: This model is optimized for AMD EPYC CPU inference via ZenDNN. It is not intended for GPU inference.