Fine Tuning g1_latin model

[Voveka98]

Hello, is there any opportunity to fine tune existing models? Im using trainer folder, and it seems that it trains a new model. Im trying to set config['saved_model'] = folder_with_downloaded_model, but it gives an error. Can you help please?

[Voveka98]

For more explanation: I want to fine tune g2_latin model with my own generated dataset. When I set config['saved_mode'] = ../latin_g2.pth i get an error

[mrkrcc]

size mismatch change your char list https://github.com/JaidedAI/EasyOCR/blob/master/easyocr/config.py#L149

[emigomez]

Any update on this? You were able to fine-tuning the g2_latin model with your own dataset?

I have the same mismatch error in the fine-tuning and changing the lang_char don't help me.

#lang_char: " !"#$%&'()+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_abcdefghijklmnopqrstuvwxyz{|}~ªÀÁÂÃÄÅÆÇÈÉÊËÌÍÎÏÐÑÒÓÔÕÖØÙÚÛÜÝÞßàáâãäåæçèéêëìíîïðñòóôõöøùúûüýþÿĀāĂ㥹ĆćČčĎďĐđĒēĖėĘęĚěĞğĨĩĪīĮįİıĶķĹĺĻļĽľŁłŃńŅņŇňŒœŔŕŘřŚśŞşŠšŤťŨũŪūŮůŲųŸŹźŻżŽžƏƠơƯưȘșȚțə̇ḌḍḶḷṀṁṂṃṄṅṆṇṬṭẠạẢảẤấẦầẨẩẪẫẬậẮắẰằẲẳẴẵẶặẸẹẺẻẼẽẾếỀềỂểỄễỆệỈỉỊịỌọỎỏỐốỒồỔổỖỗỘộỚớỜờỞởỠỡỢợỤụỦủỨứỪừỬửỮữỰựỲỳỴỵỶỷỸỹ€"#'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz' lang_char: "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyzªÀÁÂÃÄÅÆÇÈÉÊËÌÍÎÏÐÑÒÓÔÕÖØÙÚÛÜÝÞßàáâãäåæçèéêëìíîïðñòóôõöøùúûüýþÿĀāĂ㥹ĆćČčĎďĐđĒēĖėĘęĚěĞğĨĩĪīĮįİıĶķĹĺĻļĽľŁłŃńŅņŇňŒœŔŕŘřŚśŞşŠšŤťŨũŪūŮůŲųŸŹźŻżŽžƏƠơƯưȘșȚțə̇ḌḍḶḷṀṁṂṃṄṅṆṇṬṭẠạẢảẤấẦầẨẩẪẫẬậẮắẰằẲẳẴẵẶặẸẹẺẻẼẽẾếỀềỂểỄễỆệỈỉỊịỌọỎỏỐốỒồỔổỖỗỘộỚớỜờỞởỠỡỢợỤụỦủỨứỪừỬửỮữỰựỲỳỴỵỶỷỸỹ" #symbol: "!\"#$%&'()*+,-./:;<=>?@[\\]^_{|}~ €" #symbol: " !"#$%&'()+,-./0123456789:;<=>?@[\]^_`{|}~ €"

[khawar-islam]

Hello @mrkrcc @emigomez Are you able to fine-tune the EasyOCR model?