Not-For-All-Audiences

Instructions to use Khetterman/Llama-3.2-Kapusta-3B-v8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Khetterman/Llama-3.2-Kapusta-3B-v8 with Transformers:

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

pipe = pipeline("text-generation", model="Khetterman/Llama-3.2-Kapusta-3B-v8")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Khetterman/Llama-3.2-Kapusta-3B-v8")
model = AutoModelForCausalLM.from_pretrained("Khetterman/Llama-3.2-Kapusta-3B-v8", 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]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Khetterman/Llama-3.2-Kapusta-3B-v8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Khetterman/Llama-3.2-Kapusta-3B-v8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Khetterman/Llama-3.2-Kapusta-3B-v8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Khetterman/Llama-3.2-Kapusta-3B-v8

SGLang

How to use Khetterman/Llama-3.2-Kapusta-3B-v8 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 "Khetterman/Llama-3.2-Kapusta-3B-v8" \
    --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": "Khetterman/Llama-3.2-Kapusta-3B-v8",
		"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 "Khetterman/Llama-3.2-Kapusta-3B-v8" \
        --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": "Khetterman/Llama-3.2-Kapusta-3B-v8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Khetterman/Llama-3.2-Kapusta-3B-v8 with Docker Model Runner:
```
docker model run hf.co/Khetterman/Llama-3.2-Kapusta-3B-v8
```

Llama-3.2-Kapusta-3B-v8 / README.md

Khetterman

Update README.md

6833b37 verified over 1 year ago

preview code

Raw

History Blame Contribute Delete

7.66 kB

metadata

base_model:
  - bunnycore/Llama-3.2-3B-Creative
  - bunnycore/Llama-3.2-3B-Mix
  - bunnycore/Llama-3.2-3B-Pure-RP
  - bunnycore/Llama-3.2-3B-Stock
  - bunnycore/Llama-3.2-3B-TitanFusion-v2
  - CarrotAI/Llama-3.2-Rabbit-Ko-3B-Instruct
  - Devarui379/VersatiLlama-Llama-3.2-3B-Instruct-Abliterated
  - Hastagaras/L3.2-JametMini-3B-MK.III
  - huihui-ai/Llama-3.2-3B-Instruct-abliterated
  - Lyte/Llama-3.2-3B-Overthinker
  - passing2961/Thanos-3B
  - SaisExperiments/Evil-Alpaca-3B-L3.2
  - ValiantLabs/Llama3.2-3B-Enigma
  - ValiantLabs/Llama3.2-3B-ShiningValiant2
library_name: transformers
tags:
  - mergekit
  - merge
  - bfloat16
  - safetensors
  - llama
  - llama-3
  - llama-3.2
  - 3b
  - chat
  - creative
  - conversational
  - not-for-all-audiences
language:
  - en
  - ru

Llama-3.2-Kapusta-3B-v8

Small and useful.

This is an interesting merge of 14 cool models, created using mergekit. Enjoy exploring :)

Merge Details

Method

This model was merged using the multistep process and remerge with some model variations for best result.

Models

The following models were included in the merge:

Configuration

The following YAML configurations was used to produce this model:

# A-3B-v1
models:
  - model: Hastagaras/L3.2-JametMini-3B-MK.III
  - model: huihui-ai/Llama-3.2-3B-Instruct-abliterated
merge_method: model_stock
base_model: Lyte/Llama-3.2-3B-Overthinker
dtype: bfloat16

# B-3B-v1
models:
  - model: ValiantLabs/Llama3.2-3B-ShiningValiant2
  - model: bunnycore/Llama-3.2-3B-Stock
merge_method: model_stock
base_model: Lyte/Llama-3.2-3B-Overthinker
dtype: bfloat16

# C-3B-v1
models:
  - model: bunnycore/Llama-3.2-3B-Pure-RP
  - model: CarrotAI/Llama-3.2-Rabbit-Ko-3B-Instruct
merge_method: model_stock
base_model: ValiantLabs/Llama3.2-3B-ShiningValiant2
dtype: bfloat16

# Llama-3.2-Kapusta-3B-v1
models:
  - model: A-3B-v1
    parameters:
      density: [0.8, 0.5, 0.2]
      weight:  0.8
  - model: B-3B-v1
    parameters:
      density: [0.2, 0.8, 0.2]
      weight:  0.25
  - model: C-3B-v1
    parameters:
      density: [0.2, 0.5, 0.8]
      weight:  0.6
merge_method: ties
base_model: bunnycore/Llama-3.2-3B-Mix
dtype: bfloat16

# Llama-3.2-Kapusta-3B-v2
models:
  - model: SaisExperiments/Evil-Alpaca-3B-L3.2
  - model: bunnycore/Llama-3.2-3B-TitanFusion-v2
merge_method: model_stock
base_model: F:/3b/Llama-3.2-Kapusta-3B-v1
dtype: bfloat16

# Llama-3.2-Kapusta-3B-v3
models:
  - model: F:/3b/Llama-3.2-Kapusta-3B-v2
    parameters:
      weight:  [0.5, 0.6, 0.4, 0.7, 0.3, 0.8, 0.2, 0.9, 0.1,  0.9, 0.1,  0.9, 0.1, 0.8, 0.2, 0.7, 0.3, 0.6, 0.4, 0.5]
      density: [0.2, 0.8, 0.2]
merge_method: della
parameters:
  epsilon: 0.1
  lambda: 0.5
base_model: F:/3b/Llama-3.2-Kapusta-3B-v1
dtype: bfloat16

# Llama-3.2-Kapusta-3B-v4A | della
models:
  - model: F:/3b/Llama-3.2-Kapusta-3B-v1
    parameters:
      weight:  0.6
      density: 0.5
  - model: Devarui379/VersatiLlama-Llama-3.2-3B-Instruct-Abliterated
    parameters:
      weight:  0.5
      density: [0.4, 0.3, 0.3, 0.3]
  - model: bunnycore/Llama-3.2-3B-Creative
    parameters:
      weight:  0.3
      density: [0.3, 0.4, 0.3, 0.3]
  - model: ValiantLabs/Llama3.2-3B-Enigma
    parameters:
      weight:  0.3
      density: [0.3, 0.3, 0.4, 0.3]
  - model: passing2961/Thanos-3B
    parameters:
      weight:  0.3
      density: [0.3, 0.3, 0.3, 0.4]
merge_method: della
parameters:
  epsilon: 0.2
  lambda:  0.5
base_model: F:/3b/Llama-3.2-Kapusta-3B-v3
dtype: bfloat16

# Llama-3.2-Kapusta-3B-v4B | breadcrumbs
models:
  - model: F:/3b/Llama-3.2-Kapusta-3B-v1
    parameters:
      weight:  0.6
      density: 0.5
  - model: Devarui379/VersatiLlama-Llama-3.2-3B-Instruct-Abliterated
    parameters:
      weight:  0.5
      density: [0.4, 0.3, 0.3, 0.3]
  - model: bunnycore/Llama-3.2-3B-Creative
    parameters:
      weight:  0.3
      density: [0.3, 0.4, 0.3, 0.3]
  - model: ValiantLabs/Llama3.2-3B-Enigma
    parameters:
      weight:  0.3
      density: [0.3, 0.3, 0.4, 0.3]
  - model: passing2961/Thanos-3B
    parameters:
      weight:  0.3
      density: [0.3, 0.3, 0.3, 0.4]
merge_method: breadcrumbs
parameters:
  gamma: 0.02
base_model: F:/3b/Llama-3.2-Kapusta-3B-v3
dtype: bfloat16

# Llama-3.2-Kapusta-3B-v4C | dare_ties
models:
  - model: F:/3b/Llama-3.2-Kapusta-3B-v1
    parameters:
      weight:  0.6
      density: 0.5
  - model: Devarui379/VersatiLlama-Llama-3.2-3B-Instruct-Abliterated
    parameters:
      weight:  0.5
      density: [0.4, 0.3, 0.3, 0.3]
  - model: bunnycore/Llama-3.2-3B-Creative
    parameters:
      weight:  0.3
      density: [0.3, 0.4, 0.3, 0.3]
  - model: ValiantLabs/Llama3.2-3B-Enigma
    parameters:
      weight:  0.3
      density: [0.3, 0.3, 0.4, 0.3]
  - model: passing2961/Thanos-3B
    parameters:
      weight:  0.3
      density: [0.3, 0.3, 0.3, 0.4]
merge_method: dare_ties
base_model: F:/3b/Llama-3.2-Kapusta-3B-v3
dtype: bfloat16

# Llama-3.2-Kapusta-3B-v5
models:
  - model: F:/3b/Llama-3.2-Kapusta-3B-v4A
  - model: F:/3b/Llama-3.2-Kapusta-3B-v4B
  - model: F:/3b/Llama-3.2-Kapusta-3B-v4C
merge_method: model_stock
base_model: F:/3b/Llama-3.2-Kapusta-3B-v3
dtype: bfloat16

# Llama-3.2-Kapusta-3B-v6
models:
  - model: F:/3b/Llama-3.2-Kapusta-3B-v3
merge_method: slerp
base_model: F:/3b/Llama-3.2-Kapusta-3B-v5
dtype: bfloat16
parameters:
  t: [0.5, 0.6, 0.4, 0.7, 0.3, 0.8, 0.2, 0.8, 0.2, 0.7, 0.3, 0.6, 0.4, 0.5]

# Llama-3.2-Kapusta-3B-v7
models:
  - model: F:/3b/Llama-3.2-Kapusta-3B-v1
  - model: F:/3b/Llama-3.2-Kapusta-3B-v2
  - model: F:/3b/Llama-3.2-Kapusta-3B-v3
  - model: F:/3b/Llama-3.2-Kapusta-3B-v4A
  - model: F:/3b/Llama-3.2-Kapusta-3B-v4B
  - model: F:/3b/Llama-3.2-Kapusta-3B-v4C
  - model: F:/3b/Llama-3.2-Kapusta-3B-v5
merge_method: model_stock
base_model: F:/3b/Llama-3.2-Kapusta-3B-v6
dtype: bfloat16

# Llama-3.2-Kapusta-3B-v8
models:
  - model: F:/3b/Llama-3.2-Kapusta-3B-v1
  - model: F:/3b/Llama-3.2-Kapusta-3B-v3
  - model: F:/3b/Llama-3.2-Kapusta-3B-v5
merge_method: model_stock
base_model: F:/3b/Llama-3.2-Kapusta-3B-v7
dtype: bfloat16

My thanks to the authors of the original models, your work is incredible. Have a good time 🖤