holm/whisper.cpp
1
1
Fork 0
forked from extern/whisper.cpp
Code Issues Pull Requests Packages Projects Releases Wiki Activity

whisper : fix logit reading

Browse Source
This commit is contained in:
Georgi Gerganov
2023-11-10 11:02:29 +02:00
parent 659757329d
commit 7e01486b61
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
whisper.cpp
View File

@@ -2461,11 +2461,12 @@ static bool whisper_decode_internal(
// extract logits for all N tokens // extract logits for all N tokens
//logits_out.resize(n_tokens*n_vocab); //logits_out.resize(n_tokens*n_vocab);
//memcpy(logits_out.data(), ggml_get_data(logits), sizeof(float)*n_tokens*n_vocab); //memcpy(logits_out.data(), ggml_get_data(logits), sizeof(float)*n_tokens*n_vocab);
//ggml_backend_tensor_get(logits, logits_out.data(), (n_vocab*(n_tokens - 1))*sizeof(float), sizeof(float)*n_vocab);
// extract logits only for the last token // extract logits only for the last token
logits_out.resize(n_vocab); logits_out.resize(n_vocab);
//memcpy(logits_out.data(), ggml_get_data(logits), sizeof(float)*n_vocab); //memcpy(logits_out.data(), ggml_get_data(logits), sizeof(float)*n_vocab);
ggml_backend_tensor_get(logits, logits_out.data(), (n_vocab*(n_tokens - 1))*sizeof(float), sizeof(float)*n_vocab); ggml_backend_tensor_get(logits, logits_out.data(), 0, sizeof(float)*n_vocab);
if (n_tokens > 1) { if (n_tokens > 1) {
//printf("%s: used_mem = %f MB, %f MB, %f MB %f MB %f MB\n", __func__, //printf("%s: used_mem = %f MB, %f MB, %f MB %f MB %f MB\n", __func__,