whisper : add full CUDA and Metal offloading (#1472)

* whisper : migrate to ggml-backend

* whisper : fix logit reading

* whisper : fix tensor allocation during load

* whisper : fix beam-search with CUDA

* whisper : free backends + fix compile warning

* whisper : print when CUDA is enabled

* whisper : fix CoreML

* make : clean-up

* talk : fix compile warning

* whisper : support ggml_conv with CUDA and Metal (#1473)

* ggml : add CUDA support for ggml_conv

* whisper : remove ggml_repeat for conv bias + single backend

* cuda : fix im2col kernel

* metal : add im2col support + mul mat-vec f16 x f16

* bench-all : add q4 models

* whisper : clean-up

* quantize-all : fix

* ggml : im2col opts

* whisper : avoid whisper_model_data wrapper

* whisper : add note that ggml_mul_mat_pad does not work with CUDA

* whisper : factor out graph compute in common function

* whisper : fixes

* whisper : fix UB with measure buffers

* whisper : try to fix the parallel whisper_state functionality (#1479)

* whisper : try to fix the parallel whisper_state functionality

* whisper : fix multi-state Metal

* whisper : free backend instances in whisper_state
This commit is contained in:
Georgi Gerganov
2023-11-12 15:31:08 +02:00
committed by GitHub
parent ec7a6f04f9
commit b0502836b8
14 changed files with 916 additions and 1594 deletions

View File

@ -181,7 +181,7 @@ private:
// It is assumed that PCM data is normalized to a range from -1 to 1
bool write_audio(const float * data, size_t length) {
for (size_t i = 0; i < length; ++i) {
const auto intSample = static_cast<const int16_t>(data[i] * 32767);
const int16_t intSample = data[i] * 32767;
file.write(reinterpret_cast<const char *>(&intSample), sizeof(int16_t));
dataSize += sizeof(int16_t);
}