Instructions to use Contamination/contaminated_proof_7b_v1.0_safetensor with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Contamination/contaminated_proof_7b_v1.0_safetensor with Transformers:

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

pipe = pipeline("text-generation", model="Contamination/contaminated_proof_7b_v1.0_safetensor")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Contamination/contaminated_proof_7b_v1.0_safetensor")
model = AutoModelForCausalLM.from_pretrained("Contamination/contaminated_proof_7b_v1.0_safetensor")
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 Contamination/contaminated_proof_7b_v1.0_safetensor with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Contamination/contaminated_proof_7b_v1.0_safetensor"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Contamination/contaminated_proof_7b_v1.0_safetensor",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Contamination/contaminated_proof_7b_v1.0_safetensor

SGLang

How to use Contamination/contaminated_proof_7b_v1.0_safetensor 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 "Contamination/contaminated_proof_7b_v1.0_safetensor" \
    --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": "Contamination/contaminated_proof_7b_v1.0_safetensor",
		"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 "Contamination/contaminated_proof_7b_v1.0_safetensor" \
        --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": "Contamination/contaminated_proof_7b_v1.0_safetensor",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Contamination/contaminated_proof_7b_v1.0_safetensor with Docker Model Runner:
```
docker model run hf.co/Contamination/contaminated_proof_7b_v1.0_safetensor
```

How come it doesn't get perfect scores?

by CamiloMM - opened Apr 4, 2024

Discussion

CamiloMM

Apr 4, 2024

If I understood this correctly, the model should get top scores in the benchmarks, maybe even perfect ones. Right?

Contamination

Owner Apr 5, 2024

Yes, you are right. I could make it have perfect scores.

But, if do, the extremely high scores compared to other models will seem strange. But by showing that it is possible to control the scores themselves, it would be more useful.

I made these high scores intentionally to seem to have top scores possible and realistic.

CamiloMM

Apr 5, 2024

So you actually targeted those exact specific scores? Or you mean you under-trained and picked a checkpoint where roughly they looked high but not suspiciously high?

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment