Instructions to use curious-techie/Vritya-Tiny-163M-HF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use curious-techie/Vritya-Tiny-163M-HF with Transformers:

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

pipe = pipeline("text-generation", model="curious-techie/Vritya-Tiny-163M-HF", trust_remote_code=True)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("curious-techie/Vritya-Tiny-163M-HF", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use curious-techie/Vritya-Tiny-163M-HF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "curious-techie/Vritya-Tiny-163M-HF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "curious-techie/Vritya-Tiny-163M-HF",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/curious-techie/Vritya-Tiny-163M-HF

SGLang

How to use curious-techie/Vritya-Tiny-163M-HF 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 "curious-techie/Vritya-Tiny-163M-HF" \
    --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": "curious-techie/Vritya-Tiny-163M-HF",
		"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 "curious-techie/Vritya-Tiny-163M-HF" \
        --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": "curious-techie/Vritya-Tiny-163M-HF",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use curious-techie/Vritya-Tiny-163M-HF with Docker Model Runner:
```
docker model run hf.co/curious-techie/Vritya-Tiny-163M-HF
```

Vritya Tiny (163M) — Transformers

Small GPT-style causal language model exported from the original PyTorch best_model.pth checkpoint. Vocabulary matches GPT-2 BPE (50257); use the bundled tokenizer or any compatible GPT-2 tokenizer.

Load

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained(
    "curious-techie/Vritya-Tiny-163M-HF",
    trust_remote_code=True,
)
tokenizer = AutoTokenizer.from_pretrained(
    "curious-techie/Vritya-Tiny-163M-HF",
    trust_remote_code=True,
)

Requires trust_remote_code=True because the architecture is defined in modeling_vritya.py in this repo.

Model facts (default config)

~163M parameters (see config.json)
Context: 1024 tokens
12 layers, 12 heads, embedding dim 768

Source

Derived from the project checkpoint published as best_model.pth on curious-techie/Vritya-Tiny-163M.

Downloads last month: 732

Safetensors

Model size

0.2B params

Tensor type

F32