Instructions to use DeepMount00/Qwen2-1.5B-Ita with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use DeepMount00/Qwen2-1.5B-Ita with Transformers:

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

pipe = pipeline("text-generation", model="DeepMount00/Qwen2-1.5B-Ita")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("DeepMount00/Qwen2-1.5B-Ita")
model = AutoModelForCausalLM.from_pretrained("DeepMount00/Qwen2-1.5B-Ita")
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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use DeepMount00/Qwen2-1.5B-Ita with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "DeepMount00/Qwen2-1.5B-Ita"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "DeepMount00/Qwen2-1.5B-Ita",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/DeepMount00/Qwen2-1.5B-Ita

SGLang

How to use DeepMount00/Qwen2-1.5B-Ita 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 "DeepMount00/Qwen2-1.5B-Ita" \
    --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": "DeepMount00/Qwen2-1.5B-Ita",
		"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 "DeepMount00/Qwen2-1.5B-Ita" \
        --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": "DeepMount00/Qwen2-1.5B-Ita",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use DeepMount00/Qwen2-1.5B-Ita with Docker Model Runner:
```
docker model run hf.co/DeepMount00/Qwen2-1.5B-Ita
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Qwen2 1.5B: Almost the Same Performance as ITALIA (iGenius) but 6 Times Smaller 🚀

Model Overview

Model Name: Qwen2 1.5B Fine-tuned for Italian Language
Version: 1.5b
Model Type: Language Model
Parameter Count: 1.5 billion
Language: Italian
Comparable Model: ITALIA by iGenius (9 billion parameters)

Model Description

Qwen2 1.5B is a compact language model specifically fine-tuned for the Italian language. Despite its relatively small size of 1.5 billion parameters, Qwen2 1.5B demonstrates strong performance, nearly matching the capabilities of larger models, such as the 9 billion parameter ITALIA model by iGenius. The fine-tuning process focused on optimizing the model for various language tasks in Italian, making it highly efficient and effective for Italian language applications.

Performance Evaluation

The performance of Qwen2 1.5B was evaluated on several benchmarks and compared against the ITALIA model. The results are as follows:

Performance Evaluation

Model	Parameters	Average	MMLU	ARC	HELLASWAG
ITALIA	9B	43.5	35.22	38.49	56.79
Qwen2-1.5B-Ita	1.5B	46.12	52.16	36.06	50.15

How to Use

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

model_name = "DeepMount00/Qwen2-1.5B-Ita"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)

prompt = [{'role': 'user', 'content': """Marco ha comprato 5 scatole di cioccolatini. Ogni scatola contiene 12 cioccolatini. Ha deciso di dare 3 cioccolatini a ciascuno dei suoi 7 amici. Quanti cioccolatini gli rimarranno dopo averli distribuiti ai suoi amici?"""}]
inputs = tokenizer.apply_chat_template(
    prompt,
    add_generation_prompt=True,
    return_tensors='pt'
)
tokens = model.generate(
    inputs.to(model.device),
    max_new_tokens=1024,
    temperature=0.001,
    do_sample=True
)

print(tokenizer.decode(tokens[0], skip_special_tokens=False))

Conclusion

Qwen2 1.5B demonstrates that a smaller, more efficient model can achieve performance levels comparable to much larger models. It excels in the MMLU benchmark, showing its strength in multitask language understanding. While it scores slightly lower in the ARC and HELLASWAG benchmarks, its overall performance makes it a viable option for Italian language tasks, offering a balance between efficiency and capability.

Downloads last month: 2,676

Safetensors

Model size

2B params

Tensor type

BF16

Model tree for DeepMount00/Qwen2-1.5B-Ita

Quantizations

9 models

DeepMount00
/

Qwen2-1.5B-Ita

Qwen2 1.5B: Almost the Same Performance as ITALIA (iGenius) but 6 Times Smaller 🚀

Model Overview

Model Description

Performance Evaluation

Performance Evaluation

How to Use

Conclusion

Model tree for DeepMount00/Qwen2-1.5B-Ita

Space using DeepMount00/Qwen2-1.5B-Ita 1