Add more info on training method and WER+CER scores

README.md

@@ -17,6 +17,8 @@ This is a whisper-tiny model based on techiaith/whisper-tiny-ft-cy-en, fine tune
 This model can be loaded into the FUTO Keyboard, and most likely other similar keyboards (Heliboard, Florisboard, AnySoftKeyboard, possibly even Swiftkey).
 More info on this can be found [here](https://github.com/futo-org/whisper-acft).
 
+## Android Installation Instructions
+
 To use this model with FUTO keyboard:
 
 1. Download the .bin file from [download/whisper-tiny-welsh.bin](download/whisper-tiny-welsh.bin) onto your Android phone
@@ -26,6 +28,26 @@ To use this model with FUTO keyboard:
 ![screenshot](screenshot.jpg)
 
 
+## Training and evaluation data 
+
+Trained/evaluated using [welsh-transcription-samples](https://huggingface.co/datasets/ClemSummer/welsh-transcription-samples-7k/), a subset of Mozilla's Common Voice CY dataset. Smaller and more useful for poor-man's training without a GPU. Training on the full Mozilla Common Voice corpus may provide better results.
+
+## Model Setup
+
+```python
+# Training hyperparameters
+LEARNING_RATE = 1e-6
+NUM_EPOCHS = 8
+# (Note: only recordings < 29.0s were used)
+```
+
+## WER & CER
+
+| Dataset | WER | CER |
+| -------- | --- | ----- |
+| CommonVoice CY (ClemSummer, validation split) |  62.99 | 21.46 |
+| techiaith/banc-trawsgrifiadau-bangor | TODO: *no access* | TODO: *no access* |
+
 -----