Commit History
-
Upload 5 files
Verified signature1e7df35
-
Add INT8 quantized ONNX export for Transformers.js v3
Verified signatureINT8 quantized ONNX encoder + decoder_merged for in-browser inference via Transformers.js v3. Exported from visheratin/t5-efficient-tiny-grammar-correction with optimum/ONNX runtime, quantized with onnxruntime quantization (dynamic INT8). encoder: 43MB -> 11MB decoder: 79MB -> 20MB1de7db9