Instructions to use CalderaAI/Hexoteric-7B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use CalderaAI/Hexoteric-7B with Transformers:

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

pipe = pipeline("text-generation", model="CalderaAI/Hexoteric-7B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("CalderaAI/Hexoteric-7B")
model = AutoModelForCausalLM.from_pretrained("CalderaAI/Hexoteric-7B")
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 CalderaAI/Hexoteric-7B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "CalderaAI/Hexoteric-7B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "CalderaAI/Hexoteric-7B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/CalderaAI/Hexoteric-7B

SGLang

How to use CalderaAI/Hexoteric-7B 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 "CalderaAI/Hexoteric-7B" \
    --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": "CalderaAI/Hexoteric-7B",
		"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 "CalderaAI/Hexoteric-7B" \
        --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": "CalderaAI/Hexoteric-7B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use CalderaAI/Hexoteric-7B with Docker Model Runner:
```
docker model run hf.co/CalderaAI/Hexoteric-7B
```

Missing model card?

by ddh0 - opened Nov 22, 2023

Discussion

ddh0

Nov 22, 2023

Model is missing model card or any info at all...

digitous

Caldera AI org Dec 11, 2023

•

edited Dec 11, 2023

An unfortunate setback; Hex' is not intentionally left mysterious and completely devoid of information, the meticulous notes kept during the experiment were unfortunately destroyed along with our original compute asset related. It's a long story but Hexo' (the O is silent), won't be seeing much love beyond its initial ambitions - six major player models fine-tuned from Mistral, each selectively fused with a different RP lora and systemically merged in parallel until two large models remained, were merged, and a final lora fused to cap it off with some logic to drive the whole ensemble of information and behaviors in a coherent chain of thought.

I would see this as a setback but like the experiments before them, the lessons learned will carry on to our future releases - including prepping the model card before a cavalier troupe of hungry badgers ravage our precious workstation frontlining our work as well as our notes. Offsite backups and better prep will be the next foot forward including a refocus on priorities such as script optimization and new innovation to come with model merge scripts. [MergeKit is the new king, has a bit of everyone's findings in it, and seems awesome; I intend to find some new ways including agitating tensors based on a noise pattern before a spherical merge as a mutation+iteration method, that one's a bit off ways on the radar though].

Back on the question - this model was a case of 'What if we took what we did with Naberius-7B and throw everything good into a new mix'

Stanford Alpaca is the right instruct format to drive this model.

Losing access to part of our infrastructure, temporary as it may be, is sadly loss of specifics, keeping Hex embroiled in mystery, even for us.
Lessons learned - and this pet project, is on me, Digitous - not anyone else in Caldera.

Apologies for not having the more definitive credit and process on the model card as intended, will leave this model up for toying and experiments but I no longer have the notes involved.

ddh0

Dec 11, 2023

No worries at all! Looking forward to your next projects.

digitous

Caldera AI org Feb 17, 2024

More are c

digitous changed discussion status to closed Feb 17, 2024

digitous changed discussion status to open Feb 17, 2024

digitous

Caldera AI org Feb 17, 2024

After a few months' hiatus we'll be back soon with some new experiments. Thank you for your patience 😌

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