[!TIP] This model is part of a suite of optimized/quantized ONNX versions of the base model. Other variants in this direction:

FP32 (Full Precision / Base): hari31416/indictrans2-indic-en-dist-200M-ONNX

FP16 (Half Precision): hari31416/indictrans2-indic-en-dist-200M-ONNX-fp16 (Current)

INT8 (Dynamic Quantization): hari31416/indictrans2-indic-en-dist-200M-ONNX-int8

Q4F16 (4-bit Block Quantization): hari31416/indictrans2-indic-en-dist-200M-ONNX-q4f16

ONNX-exported and quantized version of ai4bharat/indictrans2-indic-en-dist-200M for in-browser and local edge inference.

Precision: FP16 (Half Precision, Lossless)
Description: Converts weights and activations to float16. Lossless tier, recommended for Apple Silicon (MPS) and CUDA runtimes.
Source Pipeline & Details: For pipeline details, benchmarks, and usage instructions, see the indictrans2-onnx-export GitHub repository.

Built for use with Transformers.js and onnxruntime-web in the browser, with fast BPE tokenizer.json files that don't require the SentencePiece WASM runtime.

Performance Visualizations

These charts show overall tradeoffs, language-level parity, and category breakdown.

Performance Tradeoffs & Size Comparison

Compared against the FP32 ONNX oracle on the golden evaluation fixtures.

Format	Model Size	Exact Match (Token)	Exact Match (Text)	SacreBLEU (Raw)	Latency (Mean)	Speedup vs. FP32
FP32	1.19 GB	100.00%	100.00%	100.00	12.2 ms	1.000x
FP16	627.2 MB	99.91%	99.91%	99.96	14.3 ms	0.849x
INT8	319.7 MB	85.64%	85.64%	94.34	10.1 ms	1.171x
Q4F16	358.6 MB	73.36%	73.36%	88.31	15.8 ms	0.742x

Language-Level Parity (FP16)

Exact match rates and translation quality (SacreBLEU / chrF) per language pair under this precision:

Language Code	Total Fixtures	Token Match Rate	Text Match Rate	SacreBLEU	SacreBLEU (chrF)
eng_Latn	1100	99.9%	99.9%	99.96	99.98

Category-Level Parity (FP16)

Exact match rates and translation quality grouped by category types:

Category	Total Fixtures	Token Match Rate	Text Match Rate	SacreBLEU	SacreBLEU (chrF)
Generic	286	99.7%	99.7%	99.82	99.93
Lexicon	264	100.0%	100.0%	100.00	100.00
Numerals	264	100.0%	100.0%	100.00	100.00
Politics	286	100.0%	100.0%	100.00	100.00

Translation Mismatch Examples

Here is a sample of up to 5 translation mismatches compared to the FP32 oracle. Many mismatches represent minor synonym differences or spacing variations.

Mismatch #1 (Category: Generic)

Source (mar_Deva → eng_Latn): ट्रेन दोन तास उशिराने धावते.
Expected (FP32): The train is running late by two hours.
Actual (FP16): The train is running two hours late.

Files

encoder_model.onnx (and optional .onnx.data weights sidecar)
decoder_model.onnx and decoder_with_past_model.onnx (share decoder_shared.onnx.data when present)
translate.py — self-contained Python inference helper (see Usage below)
Fast tokenizer config files (tokenizer_src.json, tokenizer_tgt.json, tokenizer_meta.json)
Model configuration configs (config.json, generation_config.json)

Usage Example (Python, onnxruntime)

python

# translate.py is included in this repo alongside the ONNX bundle.
# You can also find it (and read the full source) at:
#   https://github.com/Hari31416/indictrans2-onnx-export/blob/main/src/translate.py

from translate import IndicTransONNX

# Pass a HF repo ID for automatic download, or a local bundle directory path
model = IndicTransONNX("hari31416/indictrans2-indic-en-dist-200M-ONNX-fp16")
print(model.translate("चुनाव कौन जीतेगा?", src_lang="hin_Deva", tgt_lang="eng_Latn"))

Required packages:

pip install onnxruntime tokenizers huggingface-hub

License

MIT (preserved from upstream AI4Bharat).