From c9aeb3367632d4ba824db49245c884ba28d200af Mon Sep 17 00:00:00 2001 From: Georgi Gerganov Date: Mon, 16 Jan 2023 19:37:06 +0200 Subject: [PATCH] stream : fix --keep_context argument to be used correctly (#354) --- examples/stream/stream.cpp | 8 ++++---- whisper.h | 2 +- 2 files changed, 5 insertions(+), 5 deletions(-) diff --git a/examples/stream/stream.cpp b/examples/stream/stream.cpp index e125170..fb0cb3d 100644 --- a/examples/stream/stream.cpp +++ b/examples/stream/stream.cpp @@ -434,9 +434,9 @@ int main(int argc, char ** argv) { const int n_new_line = !use_vad ? params.length_ms / params.step_ms - 1 : 1; // number of steps to print new line - params.no_timestamps = !use_vad; - params.no_context = use_vad; - params.max_tokens = 0; + params.no_timestamps = !use_vad; + params.no_context |= use_vad; + params.max_tokens = 0; // init audio @@ -486,7 +486,7 @@ int main(int argc, char ** argv) { params.no_timestamps ? 0 : 1); if (!use_vad) { - fprintf(stderr, "%s: n_new_line = %d\n", __func__, n_new_line); + fprintf(stderr, "%s: n_new_line = %d, no_context = %d\n", __func__, n_new_line, params.no_context); } else { fprintf(stderr, "%s: using VAD, will transcribe on speech activity\n", __func__); } diff --git a/whisper.h b/whisper.h index 84504b7..59231e9 100644 --- a/whisper.h +++ b/whisper.h @@ -245,7 +245,7 @@ extern "C" { int duration_ms; // audio duration to process in ms bool translate; - bool no_context; // do not use initial prompt for the decoder (if any) + bool no_context; // do not use past transcription (if any) as initial prompt for the decoder bool single_segment; // force single segment output (useful for streaming) bool print_special; // print special tokens (e.g. , , , etc.) bool print_progress; // print progress information