vietrix/viena-tiny-demo-pretrain

Viena Tiny Pretrain (Base)

This is a tiny, pretrain-only Viena checkpoint. It is not instruction tuned. Use it as a base for further pretraining or SFT. It is intended for smoke tests only.

Model description

• Architecture: decoder-only Transformer (VienaModel) with RMSNorm, RoPE, SwiGLU, GQA.
• Parameters: ~10M (tiny config).
• Tokenizer: SentencePiece BPE (target vocab 2000; actual vocab may be smaller due to tiny data).
• Training: small offline synthetic dataset shipped with the repo.

Training data

• Pretrain: viena_data/examples/pretrain_offline.jsonl

All datasets are synthetic and intended for offline tests.

Training recipe (tiny)

• Config: configs/viena_tiny.yaml
• Pretrain: 50 steps

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_id = "vietrix/viena-tiny-demo-pretrain"
tokenizer = AutoTokenizer.from_pretrained(model_id, use_fast=False)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
    device_map="auto",
)

prompt = "Viena la gi?
"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
output = model.generate(**inputs, max_new_tokens=128, do_sample=True, temperature=0.7, top_p=0.9)
print(tokenizer.decode(output[0], skip_special_tokens=True))

Limitations

• Very small dataset and very few steps.
• Not instruction tuned; responses are raw continuations.
• Not suitable for real use or evaluation.

License

MIT (code + demo weights). See repository license for details.