Paar · Udgam Labs

UDGAM LABS · Translator · English → Kannada · 17.3M parameters, 17.7 MB at int8

Paar English → Kannada

Paar English → Kannada translates English sentences into Kannada (Kannada). It has 17.3M parameters: the int8 model file is 17.7 MB (19.9 MB with its vocabulary), and it runs offline on an ordinary CPU, about 29 ms a sentence on one core of a Mac mini, or in a web browser with transformers.js. It was distilled from IndicTrans2: it learned from English sentences that IndicTrans2 dist-200M translated. Against Firefox's offline English-to-Kannada model (17.1 MB, a similar size), it is behind by 1.21 chrF++ on FLORES+ (news-style text) and behind by 0.34 on IN22-Conv (everyday conversation). Against indicMT, a model about 8.5 times larger to download, it is level (no measurable difference) on FLORES+ and behind by 1.37 on IN22-Conv. All differences are paired, with 95% intervals (Results below).

It is one of twelve, English to: Hindi, Bengali, Marathi, Telugu, Tamil, Gujarati, Urdu, Kannada, Malayalam, Punjabi, Odia, Assamese.

Try it: in your browser · code and fine-tuning recipe on GitHub

Quick start

Transformers (the weights as trained):

# pip install transformers sentencepiece torch
from transformers import MarianMTModel, MarianTokenizer

repo = "udgamlabs/Paar-en-kn"
tok = MarianTokenizer.from_pretrained(repo)
model = MarianMTModel.from_pretrained(repo)

batch = tok(["Can you send me the invoice for the last three months?"], return_tensors="pt", padding=True, truncation=True, max_length=128)
out = model.generate(**batch, num_beams=1, max_new_tokens=200)
print(tok.batch_decode(out, skip_special_tokens=True)[0])
# ನೀವು ಕಳೆದ ಮೂರು ತಿಂಗಳುಗಳಿಂದ ನನಗೆ ಇನ್ವಾಯ್ಸ್ ಕಳುಹಿಸಬಹುದೇ?

CTranslate2 int8 (the small, fast build in ct2-int8/):

# pip install ctranslate2 sentencepiece huggingface_hub
import ctranslate2, sentencepiece as spm
from huggingface_hub import snapshot_download

path = snapshot_download("udgamlabs/Paar-en-kn", allow_patterns=["config.json", "ct2-int8/*"])
sp = spm.SentencePieceProcessor(model_file=f"{path}/ct2-int8/source.spm")
tr = ctranslate2.Translator(f"{path}/ct2-int8", device="cpu", compute_type="int8")

def translate(sentences):
    pieces = [p[:127] + ["</s>"] for p in sp.encode(sentences, out_type=str)]
    out = tr.translate_batch(pieces, beam_size=1, max_decoding_length=200)
    return [sp.decode([t for t in r.hypotheses[0] if t != "</s>"]) for r in out]

print(translate(["Can you send me the invoice for the last three months?"])[0])

JavaScript, in the browser or Node, with transformers.js (the ONNX build in onnx/, int8):

// npm install @huggingface/transformers
import { pipeline } from "@huggingface/transformers";

const translate = await pipeline("translation", "udgamlabs/Paar-en-kn", { dtype: "q8" });
const [out] = await translate("Can you send me the invoice for the last three months?", { max_new_tokens: 200 });
console.log(out.translation_text);
// ನೀವು ಕಳೆದ ಮೂರು ತಿಂಗಳುಗಳಿಂದ ನನಗೆ ಇನ್ವಾಯ್ಸ್ ಕಳುಹಿಸಬಹುದೇ?
  • One sentence per input. Split paragraphs into sentences first; inputs are cut at 127 pieces.
  • Greedy decoding (num_beams=1) is what we scored. Beam search was not tested.
  • The CTranslate2 int8 build scores 0.03 chrF++ from the fp32 weights on FLORES+, and 79% of its translations are word for word the same. The ONNX int8 build (for transformers.js) is within 0.03 chrF++ of the fp32 weights on both test sets. In the browser, the int8 files are 28.8 MB, downloaded once.
English Paar English → Kannada (from the Transformers quick start)
The train leaves at 6:45 tomorrow morning, so please be at the station by 6:15. ರೈಲು ನಾಳೆ ಬೆಳಿಗ್ಗೆ 6:45 ಕ್ಕೆ ಹೊರಟು ಹೋಗುತ್ತದೆ, ಆದ್ದರಿಂದ ದಯವಿಟ್ಟು 6:15 ರ ವೇಳೆಗೆ ನಿಲ್ದಾಣದಲ್ಲಿರಿರಿ.
Can you send me the invoice for the last three months? ನೀವು ಕಳೆದ ಮೂರು ತಿಂಗಳುಗಳಿಂದ ನನಗೆ ಇನ್ವಾಯ್ಸ್ ಕಳುಹಿಸಬಹುದೇ?
My phone battery dies by the afternoon even though I charge it overnight. ನಾನು ರಾತ್ರಿಯಿಡೀ ಚಾರ್ಜ್ ಮಾಡಿದರೂ ನನ್ನ ಫೋನ್ ಬ್ಯಾಟರಿ ಮಧ್ಯಾಹ್ನದೊಳಗೆ ಸಾಯುತ್ತದೆ.

These are unedited outputs, mistakes included, and no native speaker has reviewed them.

Results

chrF++ (higher is better), NFC on both sides, our harness, the same sentences for every system. Ours greedy; indicMT int8 with beam 5 as its card says; Firefox's offline model greedy, as Firefox runs it. Differences (ours minus theirs) are paired: 1,000 bootstrap resamples, 95% interval. Measured 3 to 16 November 2026. Our scores are of the fp32 weights; the int8 build is within 0.07 chrF++ of them, and the comparison with Firefox, re-run on the int8 build, gives the same result.

Test set (sentences) Paar (17.3M) indicMT (153.8M) IndicTrans2 dist-200M (211.8M) Firefox's offline model (17.1 MB)
FLORES+ devtest (1,012) 51.93 51.65 (+0.28 [-0.17, +0.71]) 54.84 53.14 (-1.21 [-1.61, -0.77])
IN22-Conv (1,503) 30.29 31.66 (-1.37 [-1.68, -1.07]) 32.38 30.64 (-0.34 [-0.63, -0.01])
  • IN22-Gen: 44.91. Numbers kept (sentences whose every number survives), FLORES+ / IN22-Conv: 90.8% / 95.2%.
  • Size: the int8 folder is 19.9 MB; the browser (ONNX int8) files are 28.8 MB. indicMT's int8 files are 169.2 MB.
  • Speed (Mac mini, FLORES+ devtest): about 29 ms a sentence on one CPU thread (CTranslate2 int8, median), about 218 sentences a second batched on four threads. In a browser (transformers.js, int8, one thread) about 40 ms a sentence in Chrome and Safari on the same Mac (measured for English → Gujarati). Phones are not measured.

Where it falls short

  • It is behind the bigger models: indicMT by 1.4 (IN22-Conv) chrF++, level on FLORES+, IndicTrans2 dist-200M, its teacher, by 2.9 on FLORES+. If you can run a 150-200M model, those translate better.
  • Firefox's offline English-to-Kannada model (17.1 MB) is ahead of it by 1.2 (FLORES+) and 0.3 (IN22-Conv) chrF++.
  • Some letters are missing from its vocabulary (ಙ, ಱ and Kannada digits). They come out as "⁇" (<unk> in transformers.js), or are dropped by the Transformers decoder. 0.8% of FLORES+ outputs carry one.
  • Idioms and short phrases are often translated word for word, and small words such as "by" (a deadline) can take the wrong sense.
  • Names are often mangled, and Latin letters inside names often become "⁇".
  • Long formal sentences lose meaning more often than short everyday ones. Conversational Kannada is weak (30.3 on IN22-Conv).
  • English to Kannada only. It does not translate into English.
  • No native-speaker review of its outputs yet.
  • Speed is measured on one Mac mini (its CPU, and Chrome and Safari for the browser build), not on phones.

How it was trained

A Marian encoder-decoder (6 encoder and 2 decoder layers, width 256, one shared 32,000-piece SentencePiece vocabulary), trained from random weights. Its training targets are IndicTrans2 dist-200M's translations (beam 4) of English sentences from IndicCorp v2 (CC0). The sentences were cleaned and deduplicated, about a third were rewritten in a spoken style, and every sentence overlapping FLORES+, NTREX or IN22 was removed before training. Scores are on the test sets, scored once after training.

Intended use and limitations

For translating short English text into Kannada where a large model is too heavy or the text should not leave the device: apps, web pages, offline tools, browser extensions, a first draft for a human translator. It is not meant for legal, medical or official text without a human check. Check names and numbers in the output.

Fine-tuning

The GitHub repository has a recipe (finetune/) for training it further on your own sentence pairs with Transformers. If you publish a fine-tuned version, set base_model: udgamlabs/Paar-en-kn in its card so it links back here.

Licence

Apache-2.0. See NOTICE: IndicTrans2 (AI4Bharat, MIT) wrote the training targets, and the English sentences come from IndicCorp v2 (AI4Bharat, CC0). No third-party weights are included.

Citation

@misc{udgamlabs2026paarkn,
  title  = {Paar: a 17 MB English-to-Kannada translator},
  author = {Prasad, Jitesh K},
  year   = {2026},
  howpublished = {Udgam Labs},
  url    = {https://huggingface.co/udgamlabs/Paar-en-kn}
}

Please also cite IndicTrans2 (Gala et al., 2023, TMLR; arXiv:2305.16307), whose translations it learned from.


About Udgam Labs. Udgam Labs is an independent AI research lab. We work from first principles: before we build, we understand why things work, whether we are training a model, fine-tuning one or applying AI to a new problem. Udgam (उद्गम) is Sanskrit for origin. udgamlabs.com

Downloads last month
68
Safetensors
Model size
17.2M params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Dataset used to train udgamlabs/Paar-en-kn

Space using udgamlabs/Paar-en-kn 1

Collection including udgamlabs/Paar-en-kn

Paper for udgamlabs/Paar-en-kn