Instructions to use Jibbalit/Llama-3.2-11B-Vision-Instruct-abliterated-WaveTank with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Jibbalit/Llama-3.2-11B-Vision-Instruct-abliterated-WaveTank with Transformers:

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

pipe = pipeline("text-generation", model="Jibbalit/Llama-3.2-11B-Vision-Instruct-abliterated-WaveTank")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Jibbalit/Llama-3.2-11B-Vision-Instruct-abliterated-WaveTank")
model = AutoModelForCausalLM.from_pretrained("Jibbalit/Llama-3.2-11B-Vision-Instruct-abliterated-WaveTank", 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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Jibbalit/Llama-3.2-11B-Vision-Instruct-abliterated-WaveTank with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Jibbalit/Llama-3.2-11B-Vision-Instruct-abliterated-WaveTank"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Jibbalit/Llama-3.2-11B-Vision-Instruct-abliterated-WaveTank",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Jibbalit/Llama-3.2-11B-Vision-Instruct-abliterated-WaveTank

SGLang

How to use Jibbalit/Llama-3.2-11B-Vision-Instruct-abliterated-WaveTank 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 "Jibbalit/Llama-3.2-11B-Vision-Instruct-abliterated-WaveTank" \
    --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": "Jibbalit/Llama-3.2-11B-Vision-Instruct-abliterated-WaveTank",
		"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 "Jibbalit/Llama-3.2-11B-Vision-Instruct-abliterated-WaveTank" \
        --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": "Jibbalit/Llama-3.2-11B-Vision-Instruct-abliterated-WaveTank",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Jibbalit/Llama-3.2-11B-Vision-Instruct-abliterated-WaveTank with Docker Model Runner:
```
docker model run hf.co/Jibbalit/Llama-3.2-11B-Vision-Instruct-abliterated-WaveTank
```

Llama-3.2-11B-Vision-Instruct-abliterated-WaveTank / surgery_report.json

Jibbalit

Upload surgery_report.json with huggingface_hub

70310fe verified 5 months ago

Raw History Blame Contribute Delete

2.65 kB

	{
	"source_model": "huihui-ai/Llama-3.2-11B-Vision-Instruct-abliterated",
	"ffn_grid": 16,
	"attn_grid": 16,
	"training_steps": 750,
	"replaced_layers": {
	"model.layers.0.mlp": "WaveTankFFN",
	"model.layers.0.self_attn": "WaveTankAttention"
	},
	"skipped_layers": [
	"model.layers.1.mlp",
	"model.layers.1.self_attn",
	"model.layers.10.mlp",
	"model.layers.10.self_attn",
	"model.layers.11.mlp",
	"model.layers.11.self_attn",
	"model.layers.12.mlp",
	"model.layers.12.self_attn",
	"model.layers.13.cross_attn",
	"model.layers.13.mlp",
	"model.layers.14.mlp",
	"model.layers.14.self_attn",
	"model.layers.15.mlp",
	"model.layers.15.self_attn",
	"model.layers.16.mlp",
	"model.layers.16.self_attn",
	"model.layers.17.mlp",
	"model.layers.17.self_attn",
	"model.layers.18.cross_attn",
	"model.layers.18.mlp",
	"model.layers.19.mlp",
	"model.layers.19.self_attn",
	"model.layers.2.mlp",
	"model.layers.2.self_attn",
	"model.layers.20.mlp",
	"model.layers.20.self_attn",
	"model.layers.21.mlp",
	"model.layers.21.self_attn",
	"model.layers.22.mlp",
	"model.layers.22.self_attn",
	"model.layers.23.cross_attn",
	"model.layers.23.mlp",
	"model.layers.24.mlp",
	"model.layers.24.self_attn",
	"model.layers.25.mlp",
	"model.layers.25.self_attn",
	"model.layers.26.mlp",
	"model.layers.26.self_attn",
	"model.layers.27.mlp",
	"model.layers.27.self_attn",
	"model.layers.28.cross_attn",
	"model.layers.28.mlp",
	"model.layers.29.mlp",
	"model.layers.29.self_attn",
	"model.layers.3.cross_attn",
	"model.layers.3.mlp",
	"model.layers.30.mlp",
	"model.layers.30.self_attn",
	"model.layers.31.mlp",
	"model.layers.31.self_attn",
	"model.layers.32.mlp",
	"model.layers.32.self_attn",
	"model.layers.33.cross_attn",
	"model.layers.33.mlp",
	"model.layers.34.mlp",
	"model.layers.34.self_attn",
	"model.layers.35.mlp",
	"model.layers.35.self_attn",
	"model.layers.36.mlp",
	"model.layers.36.self_attn",
	"model.layers.37.mlp",
	"model.layers.37.self_attn",
	"model.layers.38.cross_attn",
	"model.layers.38.mlp",
	"model.layers.39.mlp",
	"model.layers.39.self_attn",
	"model.layers.4.mlp",
	"model.layers.4.self_attn",
	"model.layers.5.mlp",
	"model.layers.5.self_attn",
	"model.layers.6.mlp",
	"model.layers.6.self_attn",
	"model.layers.7.mlp",
	"model.layers.7.self_attn",
	"model.layers.8.cross_attn",
	"model.layers.8.mlp",
	"model.layers.9.mlp",
	"model.layers.9.self_attn"
	],
	"ffn_count": 1,
	"attn_count": 1
	}