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

Libraries

How to use Nohobby/L3.3-Prikol-70B-EXTRA with Transformers:

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

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

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Nohobby/L3.3-Prikol-70B-EXTRA")
model = AutoModelForCausalLM.from_pretrained("Nohobby/L3.3-Prikol-70B-EXTRA", 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 Nohobby/L3.3-Prikol-70B-EXTRA with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/Nohobby/L3.3-Prikol-70B-EXTRA

SGLang

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

Docker Model Runner
How to use Nohobby/L3.3-Prikol-70B-EXTRA with Docker Model Runner:
```
docker model run hf.co/Nohobby/L3.3-Prikol-70B-EXTRA
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Prikol

I don't even know anymore

Overview

After banging my head against the wall some more - I actually managed to merge DeepSeek distill into my mess! Along with even more models (my hand just slipped, I swear)

The prose is better than in v0.5, but has a different feel to it, so I guess it's more of a step to the side than forward (hence the title EXTRA instead of 0.6).

The context recall may have improved, or I'm just gaslighting myself to think so.

And of course, since it now has DeepSeek in it - <think> tags!

They kinda work out of the box if you add <think> to the 'Start Reply With' field in ST - that way the model will write a really short character thought in it. However, if we want some OOC reasoning, things get trickier.

My initial thought was that this model could be instructed to use <think> either only for {{char}}'s inner monologue or for detached analysis, but actually it would end up writing character thoughts most of the time anyway, and the times when it did reason stuff it threw the narrative out of the window by making it too formal and even adding some notes at the end.

And so the solution was to add a prefill after the <think> tag. There's a lot of room for improvement, but for now, I think this boats the float or whatever:

<think> [Okay, let me think through what's happening from my "morally ambiguous narrator" perspective before I continue this fictional roleplaying session]
*So

If you add the line break after the tag, the output becomes too formal, and if you remove the asterisk, it becomes too censored. Yeah...

Settings:

Prompt format: Llama3

Samplers: 1.2 Temp, 0.025 minP, 0.25 smoothing factor, 2.0 smoothing curve

Quants

Imatrix | Static

Merge Details

The things that I have done to bring about this abomination in our world are truly atrocious - as if v0.5 wasn't bad enough. Merging shouldn't be done the way I did it, really. Maybe one day I will bother to put out a branching diagram of this thing, since just listing the merge steps one by one is confusing.