From 12a0558317683e9e05f02dbf34b725f45e84bd58 Mon Sep 17 00:00:00 2001
From: Eren Akbiyik <eren@moose.education>
Date: Mon, 21 Nov 2022 11:27:50 +0000
Subject: [PATCH] feat: prompt previous tokens for streaming

I used a vector pointer instead of vector itself because it gave weird errors, and why not
---
 examples/stream/stream.cpp | 13 +++++++++++++
 whisper.cpp                | 10 ++++++++++
 whisper.h                  |  2 ++
 3 files changed, 25 insertions(+)

diff --git a/examples/stream/stream.cpp b/examples/stream/stream.cpp
index 6f3634b..2256e69 100644
--- a/examples/stream/stream.cpp
+++ b/examples/stream/stream.cpp
@@ -234,6 +234,7 @@ int main(int argc, char ** argv) {
     std::vector<float> pcmf32(n_samples_30s, 0.0f);
     std::vector<float> pcmf32_old;
 
+    std::vector<whisper_token> prompt_tokens;
     const int n_new_line = params.length_ms / params.step_ms - 1;
 
     // print some info about the processing
@@ -344,6 +345,8 @@ int main(int argc, char ** argv) {
             wparams.audio_ctx            = params.audio_ctx;
             wparams.speed_up             = params.speed_up;
 
+            wparams.prompt_tokens        = &prompt_tokens;
+            
             if (whisper_full(ctx, wparams, pcmf32.data(), pcmf32.size()) != 0) {
                 fprintf(stderr, "%s: failed to process audio\n", argv[0]);
                 return 6;
@@ -393,6 +396,16 @@ int main(int argc, char ** argv) {
 
                 // keep part of the audio for next iteration to try to mitigate word boundary issues
                 pcmf32_old = std::vector<float>(pcmf32.end() - n_samples_keep, pcmf32.end());
+
+                // Add tokens of the last full length segment as the prompt
+                prompt_tokens.clear();
+                const int n_segments = whisper_full_n_segments(ctx);
+                for (int i = 0; i < n_segments; ++i) {
+                    const int token_count = whisper_full_n_tokens(ctx, i);
+                    for (int j = 0; j < token_count; ++j) {
+                        prompt_tokens.push_back(whisper_full_get_token_id(ctx, i, j));
+                    }
+                }
             }
         }
     }
diff --git a/whisper.cpp b/whisper.cpp
index d35b90f..48bdb31 100644
--- a/whisper.cpp
+++ b/whisper.cpp
@@ -2412,6 +2412,7 @@ struct whisper_full_params whisper_full_default_params(enum whisper_sampling_str
                     /*.speed_up             =*/ false,
                     /*.audio_ctx            =*/ 0,
 
+                    /*.prompt_tokens        =*/ nullptr,
                     /*.language             =*/ "en",
 
                     /*.greedy               =*/ {
@@ -2455,6 +2456,7 @@ struct whisper_full_params whisper_full_default_params(enum whisper_sampling_str
                     /*.speed_up             =*/ false,
                     /*.audio_ctx            =*/ 0,
 
+                    /*.prompt_tokens        =*/ nullptr,
                     /*.language             =*/ "en",
 
                     /*.greedy               =*/ {
@@ -2584,6 +2586,14 @@ int whisper_full(
         prompt_past.clear();
     }
 
+    // Prepend the prompt tokens to the prompt_past
+    if (params.prompt_tokens) {
+        for (int i = 0; i < (int) params.prompt_tokens->size(); i++) {
+            prompt_past.push_back((*params.prompt_tokens)[i]);
+        }
+        std::rotate(prompt_past.begin(), prompt_past.end() - params.prompt_tokens->size(), prompt_past.end());
+    }
+
     // overwrite audio_ctx
     ctx->exp_n_audio_ctx = params.audio_ctx;
 
diff --git a/whisper.h b/whisper.h
index 88cc711..917b459 100644
--- a/whisper.h
+++ b/whisper.h
@@ -3,6 +3,7 @@
 
 #include <stdint.h>
 #include <stdbool.h>
+#include <vector>
 
 #ifdef WHISPER_SHARED
 #    ifdef _WIN32
@@ -208,6 +209,7 @@ extern "C" {
         bool speed_up;  // speed-up the audio by 2x using Phase Vocoder
         int  audio_ctx; // overwrite the audio context size (0 = use default)
 
+        const std::vector<whisper_token> * prompt_tokens;
         const char * language;
 
         struct {