Instructions to use win10/Llama-3.2-3B-Instruct-24-9-29 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use win10/Llama-3.2-3B-Instruct-24-9-29 with Transformers:

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

pipe = pipeline("text-generation", model="win10/Llama-3.2-3B-Instruct-24-9-29")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("win10/Llama-3.2-3B-Instruct-24-9-29")
model = AutoModelForCausalLM.from_pretrained("win10/Llama-3.2-3B-Instruct-24-9-29")
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

vLLM

How to use win10/Llama-3.2-3B-Instruct-24-9-29 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "win10/Llama-3.2-3B-Instruct-24-9-29"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "win10/Llama-3.2-3B-Instruct-24-9-29",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/win10/Llama-3.2-3B-Instruct-24-9-29

SGLang

How to use win10/Llama-3.2-3B-Instruct-24-9-29 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 "win10/Llama-3.2-3B-Instruct-24-9-29" \
    --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": "win10/Llama-3.2-3B-Instruct-24-9-29",
		"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 "win10/Llama-3.2-3B-Instruct-24-9-29" \
        --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": "win10/Llama-3.2-3B-Instruct-24-9-29",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use win10/Llama-3.2-3B-Instruct-24-9-29 with Docker Model Runner:
```
docker model run hf.co/win10/Llama-3.2-3B-Instruct-24-9-29
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

my DC sever

https://discord.gg/yaTfFF6Ut2

我正在計畫微調64K指令模型，請幫助我進行計畫

Support me here if you're interested:

Ko-fi: https://ko-fi.com/ogodwin10

Llama-3.2-3B-Instruct-24-9-29

This model is a fine-tuned version of unsloth/Llama-3.2-3B-Instruct on the lmsys_chat dataset. It achieves the following results on the evaluation set:

Loss: 1.1817

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0001
train_batch_size: 2
eval_batch_size: 2
seed: 42
gradient_accumulation_steps: 8
total_train_batch_size: 16
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
training_steps: 1000
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
1.256	0.0160	100	1.1817
1.236	0.0320	200	1.1817
1.2212	0.0480	300	1.1817
1.1804	0.0641	400	1.1817
1.2801	0.0801	500	1.1817
1.2232	0.0961	600	1.1817
1.2433	0.1121	700	1.1817
1.2231	0.1281	800	1.1817
1.2272	0.1441	900	1.1817
1.2843	0.1602	1000	1.1817

Framework versions

Transformers 4.45.0
Pytorch 2.4.0+cu124
Datasets 2.19.1
Tokenizers 0.20.0

Downloads last month: 5

Safetensors

Model size

3B params

Tensor type

F16

F32

Model tree for win10/Llama-3.2-3B-Instruct-24-9-29

Base model

meta-llama/Llama-3.2-3B-Instruct

Finetuned

unsloth/Llama-3.2-3B-Instruct

Finetuned

(589)

this model

Quantizations

3 models