Instructions to use marcodsn/catmind-1.2b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use marcodsn/catmind-1.2b with Transformers:

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

pipe = pipeline("text-generation", model="marcodsn/catmind-1.2b")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("marcodsn/catmind-1.2b")
model = AutoModelForCausalLM.from_pretrained("marcodsn/catmind-1.2b", 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 marcodsn/catmind-1.2b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "marcodsn/catmind-1.2b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "marcodsn/catmind-1.2b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/marcodsn/catmind-1.2b

SGLang

How to use marcodsn/catmind-1.2b 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 "marcodsn/catmind-1.2b" \
    --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": "marcodsn/catmind-1.2b",
		"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 "marcodsn/catmind-1.2b" \
        --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": "marcodsn/catmind-1.2b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use marcodsn/catmind-1.2b with Docker Model Runner:
```
docker model run hf.co/marcodsn/catmind-1.2b
```

Weight merging and layers splicing

by Luke2642 - opened 21 days ago

Discussion

Luke2642

21 days ago

A fun research project!

Have you or do you plan to try layer splicing or weight blending of the cat model on the original model and see anything interesting?

marcodsn

Owner 20 days ago

Thank you! There are many interesting paths we could try at this moment to continue the project; currently on the list:

SFT on different domains (at present we only did math)
RL for natural emerging of reasoning in the stories
a new SFT run with stories that actually talk about the query
prompt gating to allow for inference-time customization
some kind of layer merging/weight blending as you suggested (thank you!)

I'll need to decide on what to try first as my free time is currently limited and my 3090 is already being hammered 24/7 with new research experiments, but I'll take note of any suggestion I get!

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