Hi, I did adapt this code from Optimum github about the sequence-classification model distilbert-base-uncased-finetuned-sst-2-english to the masked-lm model RoBERTa base . It works (see the code below) but the predictions are different (see results at the end of this post). I understand the reason …

Optimum & RoBERTa: how far can we trust a quantized model against its pytorch version?

echarlaix July 27, 2022, 8:16am 11

Let’s move this discussion to Symlink error when importing ORTSeqClass model via Pipeline as you are describing the same problem there.

1 Like

Topic		Replies	Views
Optimum roberta base quantization model recall drop 10% 🤗Optimum	5	468	January 15, 2024
Load pytorch trained model via optimum 🤗Optimum	5	2810	August 10, 2022
Improving Quantization Accuracy for ONNX Models with Optimum 🤗Optimum	0	686	February 8, 2024
Optimum library optimization and quantization fails 🤗Optimum	8	1479	February 22, 2025
Dynamic quantization problems 🤗Optimum	4	2159	October 16, 2022