Nova v1

Nova v1 is a decoder-only causal language model with 85.0 million parameters. This release packages the original trained weights for use through the Hugging Face Transformers API.

Model specifications

The architecture uses grouped-query attention, rotary position embeddings, RMSNorm, and a SwiGLU feed-forward network. The provided model implementation supports the Transformers causal language model interface and KV-cached generation. Review the custom modeling code before loading it with trust_remote_code=True.

Load with Transformers

Install PyTorch, Transformers, and Safetensors:

pip install torch transformers safetensors

Load the tokenizer and model from the Hub:

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "plasmova/Nova-v1"
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(model_id, trust_remote_code=True)

system_prompt = (
    "You are Nova, a helpful and accurate assistant. Answer directly and concisely. "
    "For simple arithmetic, calculate the result. Do not invent names, scenarios, or equations."
)
prompt = f"<|user|>System instruction: {system_prompt}\n\nUser request: Explain why the sky appears blue in one sentence.<|assistant|>"
inputs = tokenizer(prompt, return_tensors="pt", add_special_tokens=False)
output = model.generate(
    **inputs,
    max_new_tokens=160,
    do_sample=True,
    temperature=0.2,
    top_p=0.85,
    top_k=12,
    repetition_penalty=1.12,
    no_repeat_ngram_size=3,
    eos_token_id=[tokenizer.convert_tokens_to_ids("<|end|>"), tokenizer.convert_tokens_to_ids("<|endoftext|>")],
    pad_token_id=tokenizer.pad_token_id,
    use_cache=True,
)
print(tokenizer.decode(output[0, inputs["input_ids"].shape[1]:], skip_special_tokens=True))

These recommended starting settings match the local inference defaults: temperature 0.2, top-p 0.85, top-k 12, repetition penalty 1.12, and up to 160 new tokens. Nova v1 was trained without a dedicated system role, so the example places its concise instruction in the user text. Keep the prompt and generated text within the 2,048-token context.

Training and evaluation

The repository contains the converted model weights and tokenizer. No benchmark results are included with this release; evaluate the model for your intended use before deployment.

License

The model weights and accompanying custom model code are released under the Apache License 2.0. See LICENSE. Upstream dataset terms continue to apply to training data.