Try to improve the token sampling strategy (#193)

* whisper : try to improve the token sampling strategy

- Add the "max_initial_timestaamp" token logic from OpenAI
- Disallow sampling timestamps that are in the past

* whisper : fix the max initial timestamp logic + fallback decoding

diff --git a/whisper.cpp b/whisper.cpp
index fbcb5d14c0323ff81780bdae48cba9a360c1aaa3..42467efe693fa62e2c420e438ceba8a6eed56966 100644 (file)
--- a/whisper.cpp
+++ b/whisper.cpp
@@ -1846,7 +1846,9 @@ static bool whisper_decode(
 // the most basic sampling scheme - select the top token
 static whisper_token_data whisper_sample_best(
         const whisper_vocab & vocab,
-        const float * probs) {
+        const float * probs,
+              bool force_timestamp,
+              bool is_initial) {
     whisper_token_data result = {
         0, 0, 0.0f, 0.0f, 0.0f, -1, -1, 0.0f,
     };
@@ -1869,7 +1871,18 @@ static whisper_token_data whisper_sample_best(
             max_tx = std::max(max_tx, probs_id[i].first);
         }
 
-        for (int i = vocab.token_beg; i < n_logits; i++) {
+        const auto i0 = is_initial ? vocab.token_beg + 101 : vocab.token_beg;
+        const auto i1 = is_initial ? vocab.token_beg + 101 : n_logits;
+
+        // the initial timestamp cannot be larger than 100
+        // ref: https://github.com/openai/whisper/blob/0b1ba3d46ebf7fe6f953acfd8cad62a4f851b49f/whisper/decoding.py#L426-L429
+        if (is_initial) {
+            for (int i = i0; i < n_logits; ++ i) {
+                probs_id[i].first = -INFINITY;
+            }
+        }
+
+        for (int i = vocab.token_beg; i < i1; i++) {
             sum_ts += probs_id[i].first;
             if  (probs_id[i].first > max_ts) {
                 max_ts = probs_id[i].first;
@@ -1879,7 +1892,7 @@ static whisper_token_data whisper_sample_best(
 
         // if the probability sum of all timestamp tokens is higher than the max probability of the text tokens - sample a
         // timestamp token
-        if (sum_ts > max_tx) {
+        if (sum_ts > max_tx || force_timestamp) {
             // ref: https://github.com/openai/whisper/blob/0b1ba3d46ebf7fe6f953acfd8cad62a4f851b49f/whisper/decoding.py#L430-L438
             for (int i = 0; i < vocab.token_beg; i++) {
                 probs_id[i].first = -INFINITY;
@@ -1921,39 +1934,6 @@ static whisper_token_data whisper_sample_best(
     return result;
 }
 
-// samples only from the timestamps tokens
-static whisper_vocab::id whisper_sample_timestamp(
-        const whisper_vocab & vocab,
-        const float * probs) {
-    int n_logits = vocab.id_to_token.size();
-
-    std::vector<std::pair<double, whisper_vocab::id>> probs_id;
-    probs_id.reserve(n_logits);
-
-    for (int i = vocab.token_beg + 1; i < n_logits; i++) {
-        probs_id.push_back(std::make_pair(probs[i], i));
-    }
-
-    const int top_k = 10;
-
-    // find the top K tokens
-    std::partial_sort(
-            probs_id.begin(),
-            probs_id.begin() + top_k, probs_id.end(),
-            [](const std::pair<double, whisper_vocab::id> & a, const std::pair<double, whisper_vocab::id> & b) {
-        return a.first > b.first;
-    });
-
-    probs_id.resize(top_k);
-
-    //printf("\n");
-    //for (int i = 0; i < (int) probs_id.size(); i++) {
-    //    printf("%d: '%s' %f, %d\n", i, vocab.id_to_token.at(probs_id[i].second).c_str(), probs_id[i].first, probs_id[i].second);
-    //}
-
-    return probs_id[0].second;
-}
-
 //  500 -> 00:05.000
 // 6000 -> 01:00.000
 static std::string to_timestamp(int64_t t, bool comma = false) {
@@ -2284,19 +2264,17 @@ int whisper_decode(struct whisper_context * ctx, const whisper_token * tokens, i
 struct whisper_token_data whisper_sample_best(struct whisper_context * ctx) {
     const int64_t t_start_sample_us = ggml_time_us();
 
-    // TODO: simplify
-    auto res = whisper_sample_best(ctx->vocab, ctx->probs.data() + (ctx->probs.size() - ctx->vocab.n_vocab));
+    const auto res = whisper_sample_best(ctx->vocab, ctx->probs.data() + (ctx->probs.size() - ctx->vocab.n_vocab), false, false);
 
     ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
 
     return res;
 }
 
-whisper_token whisper_sample_timestamp(struct whisper_context * ctx) {
+struct whisper_token_data whisper_sample_timestamp(struct whisper_context * ctx, bool is_initial) {
     const int64_t t_start_sample_us = ggml_time_us();
 
-    // TODO: simplify
-    auto res = whisper_sample_timestamp(ctx->vocab, ctx->probs.data() + (ctx->probs.size() - ctx->vocab.n_vocab));
+    const auto res = whisper_sample_best(ctx->vocab, ctx->probs.data() + (ctx->probs.size() - ctx->vocab.n_vocab), true, is_initial);
 
     ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
 
@@ -2694,7 +2672,6 @@ int whisper_full(
 
         prompt.insert(prompt.end(), prompt_init.begin(), prompt_init.end());
 
-        bool done = false;
         int seek_delta = 100*WHISPER_CHUNK_SIZE;
 
         // print the prompt
@@ -2708,7 +2685,9 @@ int whisper_full(
         int result_len = 0;
         tokens_cur.clear();
 
-        for (int i = 0; i < whisper_n_text_ctx(ctx)/2 - 4; ++i) {
+        bool failed = false;
+
+        for (int i = 0, n_max = whisper_n_text_ctx(ctx)/2 - 4; i < n_max; ++i) {
             if (whisper_decode(ctx, prompt.data(), prompt.size(), n_past, params.n_threads) != 0) {
                 fprintf(stderr, "%s: failed to decode\n", __func__);
                 return 8;
@@ -2725,15 +2704,19 @@ int whisper_full(
             // feel free to experiment!
             //
             {
-                auto token = whisper_sample_best(ctx);
-
-                if (i == 0) {
-                    token.tid = whisper_token_beg(ctx);
-                }
+                const auto token = (i == 0) ? whisper_sample_timestamp(ctx, true) : whisper_sample_best(ctx);
 
                 // timestamp token - update sliding window
                 if (token.id > whisper_token_beg(ctx)) {
-                    seek_delta = 2*(token.id - whisper_token_beg(ctx));
+                    const int seek_delta_new = 2*(token.id - whisper_token_beg(ctx));
+
+                    // do not allow to go back in time
+                    if (seek_delta != 100*WHISPER_CHUNK_SIZE &&
+                        seek_delta > seek_delta_new && result_len < i) {
+                        break;
+                    }
+
+                    seek_delta = seek_delta_new;
                     result_len = i + 1;
                 }
 
@@ -2752,8 +2735,8 @@ int whisper_full(
                         if (seek + seek_delta + 100 >= seek_end) {
                             result_len = i + 1;
                         } else {
-                            // TODO: figure out how to resolve this
-                            fprintf(stderr, "\n%s: failed to generate timestamp token - this should not happen\n\n", __func__);
+                            failed = true;
+                            break;
                         }
                     }
 
@@ -2772,11 +2755,21 @@ int whisper_full(
                 }
             }
 
-            if (done) {
+            // sometimes, the decoding can get stuck in a repetition loop
+            // this is a simple strategy to avoid such cases - we simply flag the decoding as failed and advance
+            // the sliding window by 1 second
+            if (i == n_max - 1 && (result_len == 0 || seek_delta < 100*WHISPER_CHUNK_SIZE/2)) {
+                failed = true;
                 break;
             }
         }
 
+        if (failed) {
+            fprintf(stderr, "\n%s: failed to generate timestamp token - using fallback strategy\n\n", __func__);
+            seek += 100;
+            continue;
+        }
+
         // shrink down to result_len
         tokens_cur.resize(result_len);
 

diff --git a/whisper.h b/whisper.h
index 156edbbf45433adc46bab4ccd5ced437421e4050..def77d4c3c2a76751c48996d4fc5f0f49c44bc52 100644 (file)
--- a/whisper.h
+++ b/whisper.h
@@ -137,7 +137,7 @@ extern "C" {
     // whisper_sample_best() returns the token with the highest probability
     // whisper_sample_timestamp() returns the most probable timestamp token
     WHISPER_API whisper_token_data whisper_sample_best(struct whisper_context * ctx);
-    WHISPER_API whisper_token whisper_sample_timestamp(struct whisper_context * ctx);
+    WHISPER_API whisper_token_data whisper_sample_timestamp(struct whisper_context * ctx, bool is_initial);
 
     // Return the id of the specified language, returns -1 if not found
     WHISPER_API int whisper_lang_id(const char * lang);