Instructions to use andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-av with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-av with Transformers:

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

pipe = pipeline("text-generation", model="andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-av")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-av")
model = AutoModelForCausalLM.from_pretrained("andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-av", 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 andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-av with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-av"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-av",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-av

SGLang

How to use andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-av 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 "andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-av" \
    --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": "andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-av",
		"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 "andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-av" \
        --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": "andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-av",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-av with Docker Model Runner:
```
docker model run hf.co/andyx10/Qwen2.5-1.5B-Instruct-NLA-L18-av
```

Qwen2.5-1.5B-Instruct-NLA-L18-av / nla_meta.yaml

Andy Xu

Restructured re-upload of dormantx/Qwen2.5-1.5B-NLA-L18: merged weights + root nla_meta.yaml, matching the kitft/Llama-3.3-70B-NLA-L53 layout

1dd66d1 verified 2 months ago

Raw

History Blame Contribute Delete

945 Bytes

	kind: nla_model
	schema_version: 2
	d_model: 1536
	tokens:
	injection_char: <\|image_pad\|>
	injection_token_id: 151655
	injection_left_neighbor_id: 220
	injection_right_neighbor_id: 151645
	critic_suffix_ids: null
	prompt_templates:
	av: '<\|im_start\|>system

	You interpret neural-network activations. Given one activation vector, name in a single sentence the
	concept, entity, topic, or syntactic role it encodes. Be concrete; do not hedge or add preamble.<\|im_end\|>

	<\|im_start\|>user

	Activation: {injection_char}<\|im_end\|>

	<\|im_start\|>assistant

	'
	ar: '{explanation}'
	created_by: nla (dormantx/NLA_Qwen2.5_1.5B pipeline)
	role_aliases:
	verbalizer: actor
	recon: critic
	layer: 18
	extraction_layer_index: 18
	base_model: Qwen/Qwen2.5-1.5B-Instruct
	role: av
	stage: sl
	extraction:
	injection_scale: 1.01256
	mse_scale: 1.0
	training:
	lr: 1.0e-05
	loss_type: sft_next_token
	global_batch_size: 16
	num_layers: 28