Instructions to use ktrapeznikov/gpt2-medium-topic-news-v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ktrapeznikov/gpt2-medium-topic-news-v2 with Transformers:

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

pipe = pipeline("text-generation", model="ktrapeznikov/gpt2-medium-topic-news-v2")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ktrapeznikov/gpt2-medium-topic-news-v2")
model = AutoModelForCausalLM.from_pretrained("ktrapeznikov/gpt2-medium-topic-news-v2", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use ktrapeznikov/gpt2-medium-topic-news-v2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ktrapeznikov/gpt2-medium-topic-news-v2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ktrapeznikov/gpt2-medium-topic-news-v2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/ktrapeznikov/gpt2-medium-topic-news-v2

SGLang

How to use ktrapeznikov/gpt2-medium-topic-news-v2 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 "ktrapeznikov/gpt2-medium-topic-news-v2" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ktrapeznikov/gpt2-medium-topic-news-v2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "ktrapeznikov/gpt2-medium-topic-news-v2" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ktrapeznikov/gpt2-medium-topic-news-v2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use ktrapeznikov/gpt2-medium-topic-news-v2 with Docker Model Runner:
```
docker model run hf.co/ktrapeznikov/gpt2-medium-topic-news-v2
```

GPT2-medium-topic-news

Model description

GPT2-medium fine tuned on a largish news corpus conditioned on a topic, source, title

Intended uses & limitations

How to use

To generate a news article text conditioned on a topic, source, title or some subsets, prompt model with:

f"topic {topic} source"
f"topic {topic} source {source} title"
f"topic {topic} source {source} title {title} body"

Try the following tags for topic: climate, weather, vaccination.

Zero shot generation works pretty well as long as topic is a single word and not too specific.

device = "cuda:0"
tokenizer = AutoTokenizer.from_pretrained("ktrapeznikov/gpt2-medium-topic-small-set")
model = AutoModelWithLMHead.from_pretrained("ktrapeznikov/gpt2-medium-topic-small-set")
model.to(device)
topic = "climate"
prompt = tokenizer(f"topic {topics} source straitstimes title", return_tensors="pt")
out = model.generate(prompt["input_ids"].to(device), do_sample=True,max_length=500, early_stopping=True, top_p=.9)
print(tokenizer.decode(out[0].cpu(), skip_special_tokens=True))

Downloads last month: 13