Update README.md

README.md

@@ -41,7 +41,26 @@ ffmpeg -i input.mp3 -ar 16000 -ac 1 -c:a pcm_s16le output.wav
 ```
 
 ### Benchmark
-Please refer to the [kotoba-tech/kotoba-whisper-v1.0-ggml](https://huggingface.co/kotoba-tech/kotoba-whisper-v1.0-ggml) for the detail of speed up [here](https://huggingface.co/kotoba-tech/kotoba-whisper-v1.0-ggml#benchmark).
+We measure the inference speed of different kotoba-whisper-v2.0 implementations with four different Japanese speech audio on MacBook Pro with the following spec:
+- Apple M2 Pro
+- 32GB
+- 14-inch, 2023
+- OS Sonoma Version 14.4.1 (23E224)
+
+| audio file | audio duration (min)| [whisper.cpp](https://huggingface.co/kotoba-tech/kotoba-whisper-v2.0-ggml) (sec) | [faster-whisper](https://huggingface.co/kotoba-tech/kotoba-whisper-v2.0-faster) (sec)| [hf pipeline](https://huggingface.co/kotoba-tech/kotoba-whisper-v2.0) (sec)
+|--------|------|-----|------|-----|
+|audio 1 | 50.3 | 581 | 2601 | 807 |
+|audio 2 | 5.6  | 41  | 73   | 61  |
+|audio 3 | 4.9  | 30  | 141  | 54  |
+|audio 4 | 5.6  | 35  | 126  | 69  |
+
+Scripts to re-run the experiment can be found bellow:
+* [whisper.cpp](https://huggingface.co/kotoba-tech/kotoba-whisper-v2.0-ggml/blob/main/benchmark.sh)
+* [faster-whisper](https://huggingface.co/kotoba-tech/kotoba-whisper-v2.0-faster/blob/main/benchmark.sh)
+* [hf pipeline](https://huggingface.co/kotoba-tech/kotoba-whisper-v2.0/blob/main/benchmark.sh)
+Also, currently whisper.cpp and faster-whisper support the [sequential long-form decoding](https://huggingface.co/distil-whisper/distil-large-v3#sequential-long-form),
+and only Huggingface pipeline supports the [chunked long-form decoding](https://huggingface.co/distil-whisper/distil-large-v3#chunked-long-form), which we empirically
+ found better than the sequnential long-form decoding.
 
 ### Quantized Model
 To use the quantized model, download the quantized GGML weights: