Update README.md

README.md

@@ -29,41 +29,22 @@ model-index:
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
 should probably proofread and complete it, then remove this comment. -->
 
-# Portuguese Medium Whisper
+# Modelo Flax do Pierre em Português para Reconhecimento de Fala (ASR)
 
-This model is a fine-tuned version of [openai/whisper-medium](https://huggingface.co/openai/whisper-medium) on the common_voice_11_0 dataset.
-It achieves the following results on the evaluation set:
-- Loss: 0.2628
-- Wer: 6.5987
+Este repositório é um fork do repositório original criado por [Pierre Guillou](https://github.com/piegu). Ele contém uma versão convertida do modelo Whisper da OpenAI, fine-tuned no conjunto de dados `common_voice_11_0` para o idioma Português.
 
-## Blog post
+## Resultados
 
-All information about this model in this blog post: [Speech-to-Text & IA | Transcreva qualquer áudio para o português com o Whisper (OpenAI)... sem nenhum custo!](https://medium.com/@pierre_guillou/speech-to-text-ia-transcreva-qualquer-%C3%A1udio-para-o-portugu%C3%AAs-com-o-whisper-openai-sem-ad0c17384681).
+O modelo atinge os seguintes resultados no conjunto de avaliação:
 
-## New SOTA
+- Perda (Loss): 0.2628
+- Taxa de Erro de Palavra (Word Error Rate - WER): 6.5987
 
-The Normalized WER in the [OpenAI Whisper article](https://cdn.openai.com/papers/whisper.pdf) with the [Common Voice 9.0](https://huggingface.co/datasets/mozilla-foundation/common_voice_9_0) test dataset is 8.1. 
+Para obter mais informações sobre este modelo, consulte este post do autor no blog: [Speech-to-Text & IA | Transcreva qualquer áudio para o português com o Whisper (OpenAI)... sem nenhum custo!](https://medium.com/@pierre_guillou).
 
-As this test dataset is similar to the [Common Voice 11.0](https://huggingface.co/datasets/mozilla-foundation/common_voice_11_0) test dataset used to evaluate our model (WER and WER Norm), it means that **our Portuguese Medium Whisper is better than the [Medium Whisper](https://huggingface.co/openai/whisper-medium) model at transcribing audios Portuguese in text** (and even better than the [Whisper Large](https://huggingface.co/openai/whisper-large) that has a WER Norm of 7.1!).
+Este modelo, batizado de "Portuguese Medium Whisper", é superior ao modelo original Whisper Medium da OpenAI na transcrição de áudios em português (e inclusive melhor que o modelo Whisper Large, que possui um WER de 7.1).
 
-![OpenAI results with Whisper Medium and Test dataset of Commons Voice 9.0](https://huggingface.co/pierreguillou/whisper-medium-portuguese/resolve/main/whisper_medium_portuguese_wer_commonvoice9.png)
-
-## Training procedure
-
-### Training hyperparameters
-
-The following hyperparameters were used during training:
-- learning_rate: 9e-06
-- train_batch_size: 32
-- eval_batch_size: 16
-- seed: 42
-- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
-- lr_scheduler_type: linear
-- lr_scheduler_warmup_steps: 500
-- training_steps: 6000
-- mixed_precision_training: Native AMP
-
-### Training results
+## Treinamento
 
 | Training Loss | Epoch | Step | Validation Loss | Wer    |
 |:-------------:|:-----:|:----:|:---------------:|:------:|