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

Libraries

How to use trollek/Llama-3.2-3B-Danoia with Transformers:

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

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

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("trollek/Llama-3.2-3B-Danoia")
model = AutoModelForCausalLM.from_pretrained("trollek/Llama-3.2-3B-Danoia")
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 trollek/Llama-3.2-3B-Danoia with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/trollek/Llama-3.2-3B-Danoia

SGLang

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

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

Llama-3.2-3B-Danoia

Der er ikke meget at sige andet end at den kan dansk.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e-05
train_batch_size: 2
eval_batch_size: 1
seed: 222
gradient_accumulation_steps: 8
total_train_batch_size: 16
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.01
num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
1.0895	0.2103	500	1.0488
1.0893	0.4205	1000	0.9952
0.864	0.6308	1500	0.9645
0.9665	0.8411	2000	0.9406
0.9387	1.0514	2500	0.9242
0.7996	1.2617	3000	0.9126
0.7904	1.4720	3500	0.9005
0.9745	1.6822	4000	0.8926
1.0152	1.8925	4500	0.8859
0.7676	2.1028	5000	0.8821
0.8127	2.3131	5500	0.8791
0.9498	2.5234	6000	0.8770
0.795	2.7336	6500	0.8758
0.8029	2.9439	7000	0.8758