whisper : Replace WHISPER_PRINT_DEBUG with WHISPER_LOG_DEBUG (#1681)

This commit is contained in:
bobqianic 2023-12-23 12:02:58 +00:00 committed by GitHub
parent 3a5302108d
commit 37a709f655
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23

View File

@ -122,9 +122,18 @@ WHISPER_ATTRIBUTE_FORMAT(2, 3)
static void whisper_log_internal (ggml_log_level level, const char * format, ...);
static void whisper_log_callback_default(ggml_log_level level, const char * text, void * user_data);
#define WHISPER_LOG_INFO(...) whisper_log_internal(GGML_LOG_LEVEL_INFO , __VA_ARGS__)
#define WHISPER_LOG_WARN(...) whisper_log_internal(GGML_LOG_LEVEL_WARN , __VA_ARGS__)
#define WHISPER_LOG_ERROR(...) whisper_log_internal(GGML_LOG_LEVEL_ERROR, __VA_ARGS__)
#define WHISPER_LOG_WARN(...) whisper_log_internal(GGML_LOG_LEVEL_WARN , __VA_ARGS__)
#define WHISPER_LOG_INFO(...) whisper_log_internal(GGML_LOG_LEVEL_INFO , __VA_ARGS__)
// define this to enable verbose trace logging - useful for debugging purposes
// #define WHISPER_DEBUG
#if defined(WHISPER_DEBUG)
#define WHISPER_LOG_DEBUG(...) whisper_log_internal(GGML_LOG_LEVEL_DEBUG, __VA_ARGS__)
#else
#define WHISPER_LOG_DEBUG(...)
#endif
#define WHISPER_ASSERT(x) \
do { \
@ -134,18 +143,6 @@ static void whisper_log_callback_default(ggml_log_level level, const char * text
} \
} while (0)
// define this to enable verbose trace logging - useful for debugging purposes
//#define WHISPER_DEBUG
#if defined(WHISPER_DEBUG)
#define WHISPER_PRINT_DEBUG(...) \
do { \
fprintf(stderr, __VA_ARGS__); \
} while (0)
#else
#define WHISPER_PRINT_DEBUG(...)
#endif
//#define WHISPER_USE_FLASH_ATTN
//#define WHISPER_USE_FLASH_FF
#define WHISPER_MAX_DECODERS 8
@ -2166,7 +2163,7 @@ static struct ggml_cgraph * whisper_build_graph_decoder(
const int32_t n_kv = ggml_allocr_is_measure(alloc) ? n_ctx : kv_self.n;
const int32_t kv_head = ggml_allocr_is_measure(alloc) ? n_ctx - n_tokens : kv_self.head;
//WHISPER_PRINT_DEBUG("%s: n_past = %d, n_tokens = %d, n_audio_ctx = %d, n_ctx = %d\n", __func__, n_past, n_tokens, n_audio_ctx, n_ctx);
//WHISPER_LOG_DEBUG("%s: n_past = %d, n_tokens = %d, n_audio_ctx = %d, n_ctx = %d\n", __func__, n_past, n_tokens, n_audio_ctx, n_ctx);
struct ggml_init_params params = {
/*.mem_size =*/ wstate.alloc_decode.meta.size(),
@ -4944,7 +4941,7 @@ static void whisper_sequence_score(
const auto p = kv.second/(double)cnt;
entropy -= p*log(p);
//WHISPER_PRINT_DEBUG("entropy: %d %f %f, count %d\n", kv.first, p, log(p), kv.second);
//WHISPER_LOG_DEBUG("entropy: %d %f %f, count %d\n", kv.first, p, log(p), kv.second);
}
sequence.entropy = entropy;
@ -5010,7 +5007,7 @@ int whisper_full_with_state(
// basically don't process anything that is less than 1.0s
// see issue #39: https://github.com/ggerganov/whisper.cpp/issues/39
if (seek_end < seek_start + (params.speed_up ? 50 : 100)) {
WHISPER_PRINT_DEBUG("%s: input is too short - %d ms < 1000 ms\n", __func__, (seek_end - seek_start)*10);
WHISPER_LOG_DEBUG("%s: input is too short - %d ms < 1000 ms\n", __func__, (seek_end - seek_start)*10);
return 0;
}
@ -5199,7 +5196,7 @@ int whisper_full_with_state(
n_decoders_cur = std::max(1, n_decoders_cur);
WHISPER_PRINT_DEBUG("\n%s: strategy = %d, decoding with %d decoders, temperature = %.2f\n", __func__, params.strategy, n_decoders_cur, t_cur);
WHISPER_LOG_DEBUG("\n%s: strategy = %d, decoding with %d decoders, temperature = %.2f\n", __func__, params.strategy, n_decoders_cur, t_cur);
// TAGS: WHISPER_DECODER_INIT
for (int j = 0; j < n_decoders_cur; ++j) {
@ -5243,11 +5240,11 @@ int whisper_full_with_state(
prompt.insert(prompt.end(), prompt_init.begin(), prompt_init.end());
// print the prompt
WHISPER_PRINT_DEBUG("\n\n");
WHISPER_LOG_DEBUG("\n\n");
for (int i = 0; i < (int) prompt.size(); i++) {
WHISPER_PRINT_DEBUG("%s: prompt[%d] = %s\n", __func__, i, ctx->vocab.id_to_token.at(prompt[i]).c_str());
WHISPER_LOG_DEBUG("%s: prompt[%d] = %s\n", __func__, i, ctx->vocab.id_to_token.at(prompt[i]).c_str());
}
WHISPER_PRINT_DEBUG("\n\n");
WHISPER_LOG_DEBUG("\n\n");
whisper_kv_cache_clear(state->kv_self);
@ -5395,7 +5392,7 @@ int whisper_full_with_state(
whisper_kv_cache_seq_cp(state->kv_self, cur.decoder_idx, WHISPER_MAX_DECODERS + j, -1, -1);
WHISPER_PRINT_DEBUG("%s: beam search: decoder %d: from decoder %d: token = %10s, plog = %8.5f, sum_logprobs = %8.5f\n",
WHISPER_LOG_DEBUG("%s: beam search: decoder %d: from decoder %d: token = %10s, plog = %8.5f, sum_logprobs = %8.5f\n",
__func__, j, cur.decoder_idx, ctx->vocab.id_to_token.at(decoder.sequence.tokens.back().id).c_str(), decoder.sequence.tokens.back().plog, decoder.sequence.sum_logprobs_all);
}
@ -5438,7 +5435,7 @@ int whisper_full_with_state(
// do not allow to go back in time
if (has_ts && seek_delta > seek_delta_new && result_len < i) {
WHISPER_PRINT_DEBUG("%s: decoder %d: failed due to seek_delta (%d > %d)\n", __func__, j, seek_delta, seek_delta_new);
WHISPER_LOG_DEBUG("%s: decoder %d: failed due to seek_delta (%d > %d)\n", __func__, j, seek_delta, seek_delta_new);
failed = true; // TODO: maybe this is not a failure ?
continue;
}
@ -5453,7 +5450,7 @@ int whisper_full_with_state(
#ifdef WHISPER_DEBUG
{
const auto tt = token.pt > 0.10 ? ctx->vocab.id_to_token.at(token.tid) : "[?]";
WHISPER_PRINT_DEBUG("%s: id = %3d, decoder = %d, token = %6d, p = %6.3f, ts = %10s, %6.3f, result_len = %4d '%s'\n",
WHISPER_LOG_DEBUG("%s: id = %3d, decoder = %d, token = %6d, p = %6.3f, ts = %10s, %6.3f, result_len = %4d '%s'\n",
__func__, i, j, token.id, token.p, tt.c_str(), token.pt, result_len, ctx->vocab.id_to_token.at(token.id).c_str());
}
#endif
@ -5467,7 +5464,7 @@ int whisper_full_with_state(
if (seek + seek_delta + 100 >= seek_end) {
result_len = i + 1;
} else {
WHISPER_PRINT_DEBUG("%s: decoder %d failed (result_len = 0)\n", __func__, j);
WHISPER_LOG_DEBUG("%s: decoder %d failed (result_len = 0)\n", __func__, j);
failed = true;
continue;
}
@ -5478,7 +5475,7 @@ int whisper_full_with_state(
seek_delta = 100*WHISPER_CHUNK_SIZE;
}
WHISPER_PRINT_DEBUG("%s: decoder %d completed\n", __func__, j);
WHISPER_LOG_DEBUG("%s: decoder %d completed\n", __func__, j);
completed = true;
continue;
}
@ -5494,7 +5491,7 @@ int whisper_full_with_state(
// sometimes, the decoding can get stuck in a repetition loop
// this is an attempt to mitigate such cases - we flag the decoding as failed and use a fallback strategy
if (i == n_max - 1 && (result_len == 0 || seek_delta < 100*WHISPER_CHUNK_SIZE/2)) {
WHISPER_PRINT_DEBUG("%s: decoder %d: failed due to repetition loop\n", __func__, j);
WHISPER_LOG_DEBUG("%s: decoder %d: failed due to repetition loop\n", __func__, j);
failed = true;
continue;
}
@ -5536,7 +5533,7 @@ int whisper_full_with_state(
continue;
}
//WHISPER_PRINT_DEBUG("%s: decoder %d: token %d, seek_delta %d\n", __func__, j, decoder.sequence.tokens.back().id, decoder.seek_delta);
//WHISPER_LOG_DEBUG("%s: decoder %d: token %d, seek_delta %d\n", __func__, j, decoder.sequence.tokens.back().id, decoder.seek_delta);
decoder.i_batch = batch.n_tokens;
@ -5616,11 +5613,11 @@ int whisper_full_with_state(
decoder.sequence.tokens.resize(decoder.sequence.result_len);
whisper_sequence_score(params, decoder.sequence);
WHISPER_PRINT_DEBUG("%s: decoder %2d: score = %8.5f, result_len = %3d, avg_logprobs = %8.5f, entropy = %8.5f\n",
WHISPER_LOG_DEBUG("%s: decoder %2d: score = %8.5f, result_len = %3d, avg_logprobs = %8.5f, entropy = %8.5f\n",
__func__, j, decoder.sequence.score, decoder.sequence.result_len, decoder.sequence.avg_logprobs, decoder.sequence.entropy);
if (decoder.sequence.result_len > 32 && decoder.sequence.entropy < params.entropy_thold) {
WHISPER_PRINT_DEBUG("%s: decoder %2d: failed due to entropy %8.5f < %8.5f\n",
WHISPER_LOG_DEBUG("%s: decoder %2d: failed due to entropy %8.5f < %8.5f\n",
__func__, j, decoder.sequence.entropy, params.entropy_thold);
decoder.failed = true;
@ -5635,7 +5632,7 @@ int whisper_full_with_state(
}
}
WHISPER_PRINT_DEBUG("%s: best decoder = %d\n", __func__, best_decoder_id);
WHISPER_LOG_DEBUG("%s: best decoder = %d\n", __func__, best_decoder_id);
}
bool success = true;
@ -5647,7 +5644,7 @@ int whisper_full_with_state(
const auto & decoder = state->decoders[best_decoder_id];
if (decoder.failed || decoder.sequence.avg_logprobs < params.logprob_thold) {
WHISPER_PRINT_DEBUG("%s: failed due to avg_logprobs %8.5f < %8.5f\n", __func__, decoder.sequence.avg_logprobs, params.logprob_thold);
WHISPER_LOG_DEBUG("%s: failed due to avg_logprobs %8.5f < %8.5f\n", __func__, decoder.sequence.avg_logprobs, params.logprob_thold);
success = false;
state->n_fail_p++;
}
@ -5655,13 +5652,13 @@ int whisper_full_with_state(
if (success) {
//for (auto & token : ctx->decoders[best_decoder_id].sequence.tokens) {
// WHISPER_PRINT_DEBUG("%s: token = %d, p = %6.3f, pt = %6.3f, ts = %s, str = %s\n", __func__, token.id, token.p, token.pt, ctx->vocab.id_to_token.at(token.tid).c_str(), ctx->vocab.id_to_token.at(token.id).c_str());
// WHISPER_LOG_DEBUG("%s: token = %d, p = %6.3f, pt = %6.3f, ts = %s, str = %s\n", __func__, token.id, token.p, token.pt, ctx->vocab.id_to_token.at(token.tid).c_str(), ctx->vocab.id_to_token.at(token.id).c_str());
//}
break;
}
WHISPER_PRINT_DEBUG("\n%s: failed to decode with temperature = %.2f\n", __func__, t_cur);
WHISPER_LOG_DEBUG("\n%s: failed to decode with temperature = %.2f\n", __func__, t_cur);
}
// output results through a user-provided callback
@ -5673,7 +5670,7 @@ int whisper_full_with_state(
const auto & tokens_cur = best_decoder.sequence.tokens;
//WHISPER_PRINT_DEBUG("prompt_init.size() = %d, prompt.size() = %d, result_len = %d, seek_delta = %d\n", prompt_init.size(), prompt.size(), result_len, seek_delta);
//WHISPER_LOG_DEBUG("prompt_init.size() = %d, prompt.size() = %d, result_len = %d, seek_delta = %d\n", prompt_init.size(), prompt.size(), result_len, seek_delta);
// update prompt_past
prompt_past.clear();
@ -5793,7 +5790,7 @@ int whisper_full_with_state(
// update audio window
seek += seek_delta;
WHISPER_PRINT_DEBUG("seek = %d, seek_delta = %d\n", seek, seek_delta);
WHISPER_LOG_DEBUG("seek = %d, seek_delta = %d\n", seek, seek_delta);
}
}