Instructions to use udgamlabs/Paar-en-gu with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use udgamlabs/Paar-en-gu with Transformers:
# Use a pipeline as a high-level helper # Warning: Pipeline type "translation" is no longer supported in transformers v5. # You must load the model directly (see below) or downgrade to v4.x with: # pip install "transformers<5.0.0" from transformers import pipeline pipe = pipeline("translation", model="udgamlabs/Paar-en-gu")# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForSeq2SeqLM tokenizer = AutoTokenizer.from_pretrained("udgamlabs/Paar-en-gu") model = AutoModelForSeq2SeqLM.from_pretrained("udgamlabs/Paar-en-gu", device_map="auto") - Transformers.js
How to use udgamlabs/Paar-en-gu with Transformers.js:
// npm i @huggingface/transformers import { pipeline } from '@huggingface/transformers'; // Allocate pipeline const pipe = await pipeline('translation', 'udgamlabs/Paar-en-gu'); - Notebooks
- Google Colab
- Kaggle
UDGAM LABS · Translator · English → Gujarati · 17.3M parameters, 17.7 MB at int8
Paar English → Gujarati
Paar English → Gujarati translates English sentences into Gujarati (Gujarati). It has 17.3M parameters: the int8 model file is 17.7 MB (19.7 MB with its vocabulary), and it runs offline on an ordinary CPU, about 34 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-Gujarati model (17.1 MB, a similar size), it is ahead by 0.66 chrF++ on FLORES+ (news-style text) and ahead by 2.11 on IN22-Conv (everyday conversation). Against indicMT, a model about 8.6 times larger to download, it is behind by 0.74 on FLORES+ and behind by 0.52 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-gu"
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])
# શું તમે મને છેલ્લા ત્રણ મહિના માટે ઇન્વૉઇઝન મોકલી શકો છો?
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-gu", 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-gu", { 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);
// શું તમે મને છેલ્લા ત્રણ મહિના માટે ઇન્વૉઇઝન મોકલી શકો છો?
- 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 0.06 chrF++ of the fp32 weights on both test sets. In the browser, the int8 files are 28.7 MB, downloaded once.
| English | Paar English → Gujarati (from the Transformers quick start) |
|---|---|
| The train leaves at 6:45 tomorrow morning, so please be at the station by 6:15. | ટ્રેન આવતીકાલે સવારે 6.45 પર રવાના થાય છે, તેથી મહેરબાની કરીને 6:15 દ્વારા સ્ટેશન પર રહો. |
| Can you send me the invoice for the last three months? | શું તમે મને છેલ્લા ત્રણ મહિના માટે ઇન્વૉઇઝન મોકલી શકો છો? |
| My phone battery dies by the afternoon even though I charge it overnight. | હું તેને રાતોરાત ચાર્જ કરું છું તેમ છતાં મારી ફોનની બેટરી બપોર સુધીમાં મૃત્યુ પામે છે. |
These are unedited outputs, mistakes included, and no native speaker has reviewed them. In the first, "by 6:15" became દ્વારા ("by means of") and 6:45 became 6.45; in the third, the battery "dies" as a person does (મૃત્યુ પામે છે).
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) | 53.21 | 53.95 (-0.74 [-1.22, -0.28]) | 55.62 | 52.55 (+0.66 [+0.13, +1.20]) |
| IN22-Conv (1,503) | 49.60 | 50.12 (-0.52 [-1.06, -0.02]) | 52.94 | 47.49 (+2.11 [+1.59, +2.67]) |
- IN22-Gen: 49.42. Numbers kept (sentences whose every number survives), FLORES+ / IN22-Conv: 98.5% / 95.2%.
- Size: the int8 folder is 19.7 MB; the browser (ONNX int8) files are 28.7 MB. indicMT's int8 files are 169.2 MB.
- Speed (Mac mini, FLORES+ devtest): about 34 ms a sentence on one CPU thread (CTranslate2 int8, median), about 203 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 0.7 (FLORES+) and 0.5 (IN22-Conv) chrF++, IndicTrans2 dist-200M, its teacher, by 2.4 on FLORES+. If you can run a 150-200M model, those translate better.
- Some letters are missing from its vocabulary (ઋ, ઐ, ઔ, ઑ (as in ઑફિસ), ઍ, ઙ, the nukta and ૐ). They come out as "⁇" (
<unk>in transformers.js), or are dropped by the Transformers decoder: "formal studies" comes out as પચારિક અભ્યાસો, with the first letter of ઔપચારિક missing. 2.2% 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.
- English to Gujarati 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 Gujarati 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-gu 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{udgamlabs2026paargu,
title = {Paar: a 17 MB English-to-Gujarati translator},
author = {Prasad, Jitesh K},
year = {2026},
howpublished = {Udgam Labs},
url = {https://huggingface.co/udgamlabs/Paar-en-gu}
}
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
- 28
