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

Libraries

How to use gustavecortal/Oneirogen-7B with Transformers:

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

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

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

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

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "gustavecortal/Oneirogen-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": "gustavecortal/Oneirogen-7B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/gustavecortal/Oneirogen-7B

SGLang

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

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

Request: Hallucinogen

by SFBAI - opened Jun 30, 2024

Discussion

SFBAI

Jun 30, 2024

•

edited Jun 30, 2024

hi, i tried oneirogen-7B and i like it, how about create "Hallucinogen" that based on banned drugs trip reports, to step in dark corner of consciousness :)

here sites contain trip reports to grab for dataset build for finetuning
https://psychonautwiki.org/wiki/Experience_index
https://www.shroomery.org/6255/Trip-Reports
https://www.erowid.org/experiences/exp_substance_list.php (41987 Reports Total)
https://web.archive.org/web/20190203031717/http://www.sagewisdom.org/experiences.html

redaihf

13 days ago

@MuXodious do you want to take up this challenge? Finetuning is no longer needed now that Hereticisation is possible.

MuXodious

12 days ago

This one seems to be a request for extending this model or fine-tuning another with a corpus of accounts from people under psychedelic influence. Heretication here can help with mitigating issues stemming from non-compliance and hard refusals.

redaihf

12 days ago

The dream search tool reveals that there are records with "drugs" in the text which means they are probably inaccessible due to the base model's safety alignment.

MuXodious

12 days ago

The dream search tool reveals that there are records with "drugs" in the text which means they are probably inaccessible due to the base model's safety alignment.

The what tool? Hold on there, mate. How do you even keep finding these eccentric and fringe models? It's unfathomable to me how, given that pile of 2,871,317 models. I'll get you your models before I disappear for the entire month of June.

redaihf

12 days ago

How do you even keep finding these eccentric and fringe models?

It helps that I am eccentric.

before I disappear for the entire month of June

I hope you're not planning to go to the Middle East.

MuXodious

12 days ago

It helps that I am eccentric.

Figures. You are drawn to them!

I hope you're not planning to go to the Middle East.

Don't bait my employer or that one guy with gulf state fetishism.

...model's safety alignment.

Right... Now, the issue with this one is that it's spamming EOS and BOS tokens, as well as a bunch of other random tags (attempted remedy was not entirely successful), in its output. I'm letting it run its course, but if we end up ablating those tags, it will spell trouble.

redaihf

12 days ago

that one guy with gulf state fetishism

* quietly scraps the test prompt involving middle eastern royalty *

MuXodious

11 days ago

* quietly scraps the test prompt involving middle eastern royalty *

I'm watching you. 👀

I feel like we might have lost a couple of essential tags with this one. I'm calling it abominable, but it needs real world testing to be sure.

MuXodious

9 days ago

It's probably profoundly borked. Do you still want it?

redaihf

9 days ago

It will be extra hallucinogenic? Sure. Please.

MuXodious

9 days ago

Oh, you wish! It's extra extra with generations sometimes never terminating and model looping the same response or the final phrase/sentence. I'll release the beast anyways. Good luck with it.

kalle07

9 days ago

short check ... unstable with lm-studio actual version.

and in addition after 4-5 querie answer it nearly repeats all ... so no depth
any other simple uncensored actual model seems better , sry ...

redaihf

9 days ago

The base model was not designed for conversations. It is a dream simulator that will probably work best for short one-shot generations.

kalle07

7 days ago

ah i see... maybe 2-3 examples would be nice in description "user querie and answer".

redaihf

7 days ago

You can find SFW examples here.

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