Instructions to use MarinaraSpaghetti/Doctor-Shotgun_Nous-Capybara-limarpv3-34B-4.2bpw-h6-exl2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use MarinaraSpaghetti/Doctor-Shotgun_Nous-Capybara-limarpv3-34B-4.2bpw-h6-exl2 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="MarinaraSpaghetti/Doctor-Shotgun_Nous-Capybara-limarpv3-34B-4.2bpw-h6-exl2")# Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("MarinaraSpaghetti/Doctor-Shotgun_Nous-Capybara-limarpv3-34B-4.2bpw-h6-exl2") model = AutoModelForCausalLM.from_pretrained("MarinaraSpaghetti/Doctor-Shotgun_Nous-Capybara-limarpv3-34B-4.2bpw-h6-exl2") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use MarinaraSpaghetti/Doctor-Shotgun_Nous-Capybara-limarpv3-34B-4.2bpw-h6-exl2 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "MarinaraSpaghetti/Doctor-Shotgun_Nous-Capybara-limarpv3-34B-4.2bpw-h6-exl2" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "MarinaraSpaghetti/Doctor-Shotgun_Nous-Capybara-limarpv3-34B-4.2bpw-h6-exl2", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/MarinaraSpaghetti/Doctor-Shotgun_Nous-Capybara-limarpv3-34B-4.2bpw-h6-exl2
- SGLang
How to use MarinaraSpaghetti/Doctor-Shotgun_Nous-Capybara-limarpv3-34B-4.2bpw-h6-exl2 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 "MarinaraSpaghetti/Doctor-Shotgun_Nous-Capybara-limarpv3-34B-4.2bpw-h6-exl2" \ --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": "MarinaraSpaghetti/Doctor-Shotgun_Nous-Capybara-limarpv3-34B-4.2bpw-h6-exl2", "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 "MarinaraSpaghetti/Doctor-Shotgun_Nous-Capybara-limarpv3-34B-4.2bpw-h6-exl2" \ --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": "MarinaraSpaghetti/Doctor-Shotgun_Nous-Capybara-limarpv3-34B-4.2bpw-h6-exl2", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use MarinaraSpaghetti/Doctor-Shotgun_Nous-Capybara-limarpv3-34B-4.2bpw-h6-exl2 with Docker Model Runner:
docker model run hf.co/MarinaraSpaghetti/Doctor-Shotgun_Nous-Capybara-limarpv3-34B-4.2bpw-h6-exl2
settings?
I too really love this model I can tell it has some serious potential (in my eyes. I know there's better haha), though I'm really struggling with the settings. What do you use for yours if you're willing to share. I found that
Temp = 1 (+/- ~.50)
Min P = 0.25
Tfs = 0.95 (i don't remember if I was the one that set this)
Repetition penalty = 1.8
Mirostat = tau 2, eta 0.1
Everything else off
Work enough to at least get it going with rather impressive dedication to its character, but it lacks the creativity I know it has while experimenting with it. I just cant find that Sweet spot to keep it creative, intelligent, and sane at the same time.
Howdy! I don't recommend using Mirostat, it's kind of obsolete now that Min P is a thing. The Repetition Penalty is definitely too high too, any Yi-based models like it low. Here are my settings that I use in my SillyTavern, they've been working wonderfully for me so far! Hope they'll help!
https://files.catbox.moe/mctzx7.json
Howdy! I don't recommend using Mirostat, it's kind of obsolete now that Min P is a thing.
I'll absolutely take a look at it once I have the time thank you very much! Yes I had heard many times that Mirostat was obsolete. I rarely used it to begin with but I found that using it in conjunction with a low MinP had intresting results so I kept it on (I know most people say one or the other not both). That's the fun of messing with sliders I guess haha.