Instructions to use zerofata/L3.3-GeneticLemonade-Opus-70B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use zerofata/L3.3-GeneticLemonade-Opus-70B with Transformers:

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

pipe = pipeline("text-generation", model="zerofata/L3.3-GeneticLemonade-Opus-70B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("zerofata/L3.3-GeneticLemonade-Opus-70B")
model = AutoModelForCausalLM.from_pretrained("zerofata/L3.3-GeneticLemonade-Opus-70B", 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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use zerofata/L3.3-GeneticLemonade-Opus-70B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "zerofata/L3.3-GeneticLemonade-Opus-70B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "zerofata/L3.3-GeneticLemonade-Opus-70B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/zerofata/L3.3-GeneticLemonade-Opus-70B

SGLang

How to use zerofata/L3.3-GeneticLemonade-Opus-70B 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 "zerofata/L3.3-GeneticLemonade-Opus-70B" \
    --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": "zerofata/L3.3-GeneticLemonade-Opus-70B",
		"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 "zerofata/L3.3-GeneticLemonade-Opus-70B" \
        --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": "zerofata/L3.3-GeneticLemonade-Opus-70B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use zerofata/L3.3-GeneticLemonade-Opus-70B with Docker Model Runner:
```
docker model run hf.co/zerofata/L3.3-GeneticLemonade-Opus-70B
```

My personal opinion about Lemonade Opus

by Laetilia - opened Jan 3

Discussion

Laetilia

Jan 3

Used it a while back and finally decided to write about my experience.
IIRC, I used it at Q4 (IQ4_XS, most likely) and with something from recommended range of sampling settings.
For the 70B Llama's I use either -c 12000 or -c 8192; one of these were used for context size.
This model have very peculiar behavior.
At the start of just chatting or the RP with the model, it shows itself incredibly.
Follows instructions well, with vibrant writing... at first, it may even be better than (correctly used) Visage-V3.
And then, it decays into more of generic, boring LLM-style. I think around after 2000 tokens (or mayhaps even earlier).
Very quickly (not enough even for very short RP), it turns from very impressive to rather boring.
So, unless I did or percieve something wrong (which is possible, ofc), it seems to have learned very-short-term of being good.

Maybe someday I'll try to refresh context and write summaries every few messages; that seems like the way to go for this model.
Or maybe not. Regardless, that was an interesting experience.

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