Skip to content
Other

Artha-1.1B v2

Compact prompt-to-math notation translator

2 min readupdated

The model is a LoRA adapter (a small add-on layer for a base model) for TinyLlama-1.1B-Chat-v1.0 that translates standard English text prompts into a compact Artha notation. Version v2 expands the tokenizer (the model's vocabulary) with 67 special symbols and atoms, reducing prompt token usage by 35.4% compared to English. This development helps optimize workflows for compact language models and saves context space.

Key features

  • New tokenizer: 67 frequently used Artha symbols added as whole tokens.
  • Token savings: 35.4% reduction in token consumption compared to English text.
  • Accuracy: 100% exact match with reference results on the test pattern set.
  • Base architecture: built on TinyLlama-1.1B using LoRA (r=16, alpha=32).

What you can do

  • Prompt compression: use the model to convert long instructions into the compact Artha format before sending them to the pipeline.
  • Pipeline integration: embed the model into local workflows to reduce context length in text processing tasks.
  • Notation experiments: explore the potential of mathematical notation for controlling text generation.

Example

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import torch

REPO = "siddsukh/artha-1.1b-v2"
tok = AutoTokenizer.from_pretrained(REPO)
base = AutoModelForCausalLM.from_pretrained(
    "TinyLlama/TinyLlama-1.1B-Chat-v1.0", dtype=torch.bfloat16, device_map="auto"
)
base.resize_token_embeddings(len(tok), mean_resizing=False)
model = PeftModel.from_pretrained(base, REPO).eval()

print("Модель успешно загружена")
Read moreDetails
Accuracy (Exact match): 100% on 1,000 test pairs
Token savings: 35.4% compared to standard English text
Training parameters: LoRA r=16, alpha=32, 2 epochs, learning rate 2e-4
Limitations: training was conducted on 9 template task types, tokenizer is specific to this model
Discuss Telegram 0 24

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 2.8 GB · ready files3.6 GB
  • FP8 / INT8recommendedweights 1.8 GB2.5 GB
  • 4-bitweights 1.3 GB2.0 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

This is a LoRA: counted together with its base model TinyLlama-1.1B-Chat-v1.0.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # siddsukh/artha-1.1b-v2
    # ===== Tokens: paste yours between the quotes =====
    export HF_TOKEN=""        # huggingface.co/settings/tokens, for gated models
    # ==========================================
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download TinyLlama/TinyLlama-1.1B-Chat-v1.0 --local-dir ~/models/TinyLlama-1.1B-Chat-v1.0
    hf download siddsukh/artha-1.1b-v2 --local-dir ~/models/artha-1.1b-v2
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

Requires the base model

TinyLlama/TinyLlama-1.1B-Chat-v1.0

This model is a LoRA / adapter. Download the base model first, then load this one on top.

New posts about versions of TinyLlama-1.1B-Chat-v1.0 (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 12All on Hugging Face
Merges 3All on Hugging Face

About the model

siddsukh/artha-1.1b-v2
Author
siddsukh
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

240

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments