From b1b93966948ee71a3a2d72ce9f9a98a6a29c64a0 Mon Sep 17 00:00:00 2001 From: jhj0517 <97279763+jhj0517@users.noreply.github.com> Date: Fri, 4 Oct 2024 19:58:31 +0900 Subject: [PATCH 1/3] Downgrade torch --- requirements.txt | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/requirements.txt b/requirements.txt index 9825cc8a..9f8d5e0b 100644 --- a/requirements.txt +++ b/requirements.txt @@ -2,10 +2,11 @@ # If you're using it, update url to your CUDA version (CUDA 12.1 is minimum requirement): # For CUDA 12.1, use : https://download.pytorch.org/whl/cu121 # For CUDA 12.4, use : https://download.pytorch.org/whl/cu124 ---extra-index-url https://download.pytorch.org/whl/cu124 +--extra-index-url https://download.pytorch.org/whl/cu121 -torch +torch==2.3.1 +torchaudio==2.3.1 git+https://github.com/jhj0517/jhj0517-whisper.git faster-whisper==1.0.3 transformers==4.42.3 From 45fac7d5bf2ab3babf68d738d3b6a7ad53d02ab8 Mon Sep 17 00:00:00 2001 From: jhj0517 <97279763+jhj0517@users.noreply.github.com> Date: Fri, 4 Oct 2024 19:58:55 +0900 Subject: [PATCH 2/3] Use pypi indexing --- requirements.txt | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/requirements.txt b/requirements.txt index 9f8d5e0b..b0360160 100644 --- a/requirements.txt +++ b/requirements.txt @@ -9,8 +9,8 @@ torch==2.3.1 torchaudio==2.3.1 git+https://github.com/jhj0517/jhj0517-whisper.git faster-whisper==1.0.3 -transformers==4.42.3 -gradio==4.43.0 +transformers +gradio pytubefix ruamel.yaml==0.18.6 pyannote.audio==3.3.1; From c8e54dbf8a6884b5427bc7e35d8f8b405ae39125 Mon Sep 17 00:00:00 2001 From: jhj0517 <97279763+jhj0517@users.noreply.github.com> Date: Fri, 4 Oct 2024 20:39:38 +0900 Subject: [PATCH 3/3] Fix insanely_fast_whisper crash --- modules/whisper/insanely_fast_whisper_inference.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/modules/whisper/insanely_fast_whisper_inference.py b/modules/whisper/insanely_fast_whisper_inference.py index c95a26f2..21eb930d 100644 --- a/modules/whisper/insanely_fast_whisper_inference.py +++ b/modules/whisper/insanely_fast_whisper_inference.py @@ -78,7 +78,8 @@ def transcribe(self, kwargs = { "no_speech_threshold": params.no_speech_threshold, "temperature": params.temperature, - "compression_ratio_threshold": params.compression_ratio_threshold + "compression_ratio_threshold": params.compression_ratio_threshold, + "logprob_threshold": params.log_prob_threshold, } if self.current_model_size.endswith(".en"):