Instructions to use tiny-random/gpt-oss with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tiny-random/gpt-oss with Transformers:

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

pipe = pipeline("text-generation", model="tiny-random/gpt-oss")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tiny-random/gpt-oss")
model = AutoModelForCausalLM.from_pretrained("tiny-random/gpt-oss")
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

vLLM

How to use tiny-random/gpt-oss with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/tiny-random/gpt-oss

SGLang

How to use tiny-random/gpt-oss 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 "tiny-random/gpt-oss" \
    --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": "tiny-random/gpt-oss",
		"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 "tiny-random/gpt-oss" \
        --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": "tiny-random/gpt-oss",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tiny-random/gpt-oss with Docker Model Runner:
```
docker model run hf.co/tiny-random/gpt-oss
```

gpt-oss / README.md

yujiepan

Upload folder using huggingface_hub

02ba5c6 verified 10 months ago

preview code

raw

history blame contribute delete

2.85 kB

metadata

library_name: transformers
pipeline_tag: text-generation
inference: true
widget:
  - text: Hello!
    example_title: Hello world
    group: Python
base_model:
  - openai/gpt-oss-120b

This tiny model is for debugging. It is randomly initialized with the config adapted from openai/gpt-oss-120b.

Note: This model is in BF16; quantized MXFP4 FFN is not used.

Example usage:

vLLM

vllm serve tiny-random/gpt-oss

Transformers

import torch
from transformers import pipeline

model_id = "tiny-random/gpt-oss"

pipe = pipeline(
    "text-generation",
    model=model_id,
    torch_dtype=torch.bfloat16,
    device_map="cuda"
)

messages = [
    {"role": "user", "content": "Explain quantum mechanics clearly and concisely."},
]

outputs = pipe(
    messages,
    max_new_tokens=16,
)
print(outputs[0]["generated_text"][-1])

Codes to create this repo:

import json

import torch
from huggingface_hub import hf_hub_download
from transformers import (
    AutoConfig,
    AutoModelForCausalLM,
    AutoProcessor,
    AutoTokenizer,
    GenerationConfig,
    GptOssForCausalLM,
    pipeline,
    set_seed,
)

source_model_id = "openai/gpt-oss-120b"
save_folder = "/tmp/tiny-random/gpt-oss"

processor = AutoProcessor.from_pretrained(source_model_id)
processor.save_pretrained(save_folder)

with open(hf_hub_download(source_model_id, filename='config.json', repo_type='model'), 'r') as f:
    config_json = json.load(f)
config_json.update({
    "head_dim": 32,
    "hidden_size": 32,  # required by Mxfp4GptOssExperts codes
    "intermediate_size": 64,
    "layer_types": ["sliding_attention", "full_attention"],
    "num_attention_heads": 2,
    "num_hidden_layers": 2,
    "num_key_value_heads": 1,
    "num_local_experts": 32,
    "tie_word_embeddings": True,
})
quantization_config = config_json['quantization_config']
del config_json['quantization_config']
with open(f"{save_folder}/config.json", "w", encoding='utf-8') as f:
    json.dump(config_json, f, indent=2)

config = AutoConfig.from_pretrained(save_folder)
print(config)
torch.set_default_dtype(torch.bfloat16)
model = AutoModelForCausalLM.from_config(config)
torch.set_default_dtype(torch.float32)
model.generation_config = GenerationConfig.from_pretrained(
    source_model_id, trust_remote_code=True,
)
set_seed(42)
with torch.no_grad():
    for name, p in sorted(model.named_parameters()):
        torch.nn.init.normal_(p, 0, 0.1)
        print(name, p.shape)
model.save_pretrained(save_folder)

# mxfp4
from transformers.quantizers.quantizer_mxfp4 import Mxfp4HfQuantizer
# model = AutoModelForCausalLM.from_pretrained(save_folder, trust_remote_code=True, torch_dtype=torch.bfloat16, quantization_config=quantization_config)
# model.save_pretrained(save_folder, safe_serialization=True)