Instructions to use veyra-ai/Veyra2-Mango-15M-Base-ONNX with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use veyra-ai/Veyra2-Mango-15M-Base-ONNX with Transformers.js:

// npm i @huggingface/transformers
import { pipeline } from '@huggingface/transformers';

// Allocate pipeline
const pipe = await pipeline('text-generation', 'veyra-ai/Veyra2-Mango-15M-Base-ONNX');

Transformers

How to use veyra-ai/Veyra2-Mango-15M-Base-ONNX with Transformers:

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

pipe = pipeline("text-generation", model="veyra-ai/Veyra2-Mango-15M-Base-ONNX")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("veyra-ai/Veyra2-Mango-15M-Base-ONNX")
model = AutoModelForCausalLM.from_pretrained("veyra-ai/Veyra2-Mango-15M-Base-ONNX", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use veyra-ai/Veyra2-Mango-15M-Base-ONNX with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "veyra-ai/Veyra2-Mango-15M-Base-ONNX"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "veyra-ai/Veyra2-Mango-15M-Base-ONNX",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/veyra-ai/Veyra2-Mango-15M-Base-ONNX

SGLang

How to use veyra-ai/Veyra2-Mango-15M-Base-ONNX 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 "veyra-ai/Veyra2-Mango-15M-Base-ONNX" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "veyra-ai/Veyra2-Mango-15M-Base-ONNX",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "veyra-ai/Veyra2-Mango-15M-Base-ONNX" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "veyra-ai/Veyra2-Mango-15M-Base-ONNX",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use veyra-ai/Veyra2-Mango-15M-Base-ONNX with Docker Model Runner:
```
docker model run hf.co/veyra-ai/Veyra2-Mango-15M-Base-ONNX
```

Veyra2 Mango 15M Base ONNX

It is not instruction tuned and should not be evaluated like a finished chat assistant. It is expected to hallucinate, repeat, fail simple factual/math prompts, and continue text in odd ways.

This model is an ONNX conversion, the main model can be found here: Veyra2 Mango 15M Base.

Citation / Attribution

If you use or build on this model, please retain attribution to Veyra AI.

License

Apache 2.0.

Downloads last month: 103

Model tree for veyra-ai/Veyra2-Mango-15M-Base-ONNX

Base model

veyra-ai/Veyra2-Mango-15M-Base

Quantized

(1)

this model

Collection including veyra-ai/Veyra2-Mango-15M-Base-ONNX

Veyra2

Collection

The second generation of Veyra. • 9 items • Updated 3 days ago

Evaluation results

Accuracy on SciCloze-900
self-reported

36.780
Accuracy on SciQ
self-reported

65.400
Normalized Accuracy on SciQ
self-reported

58.800
Normalized Accuracy on PIQA
self-reported

58.000
Normalized Accuracy on ARC-Easy
self-reported

37.160
Normalized Accuracy on ARC-Challenge
self-reported

22.870
Normalized Accuracy on HellaSwag
self-reported

27.730
Accuracy on Winogrande
self-reported

50.990