Update README.md
Browse files
README.md
CHANGED
|
@@ -178,7 +178,18 @@ library_name: transformers.js
|
|
| 178 |
|
| 179 |
# whisper-small (ONNX)
|
| 180 |
|
|
|
|
| 181 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 182 |
This is an ONNX version of [unsloth/whisper-small](https://huggingface.co/unsloth/whisper-small). It was automatically converted and uploaded using [this Hugging Face Space](https://huggingface.co/spaces/onnx-community/convert-to-onnx).
|
| 183 |
|
| 184 |
|
|
|
|
| 178 |
|
| 179 |
# whisper-small (ONNX)
|
| 180 |
|
| 181 |
+
Based on unsloth-whisper-small, the following configuration can be used for conversion:
|
| 182 |
|
| 183 |
+
## Do not use the v4 version to convert Whisper; it will not work at all in the Q4 series.
|
| 184 |
+
|
| 185 |
+
|
| 186 |
+
```
|
| 187 |
+
quantization: {
|
| 188 |
+
encoder_model: 'q4f16',
|
| 189 |
+
decoder_model_merged: 'q4',
|
| 190 |
+
},
|
| 191 |
+
```
|
| 192 |
+
This combination helps reduce memory usage during speech recognition. It also addresses the issue where using q4f16 with onnx-community/whisper-small or Xenova/whisper-small may result in garbled output and malfunctioning systems.
|
| 193 |
This is an ONNX version of [unsloth/whisper-small](https://huggingface.co/unsloth/whisper-small). It was automatically converted and uploaded using [this Hugging Face Space](https://huggingface.co/spaces/onnx-community/convert-to-onnx).
|
| 194 |
|
| 195 |
|