Description
windows tiny:
(base) PS F:\githubsources\whisper.cpp> .\build\bin\Release\main.exe -m F:\Downloads\ggml-tiny.en.bin -l auto F:\githubsources\whisper.cpp\samples\jfk.wav
whisper_init_from_file_with_params_no_state: loading model from 'F:\Downloads\ggml-tiny.en.bin'
whisper_model_load: loading model
whisper_model_load: n_vocab = 51864
whisper_model_load: n_audio_ctx = 1500
whisper_model_load: n_audio_state = 384
whisper_model_load: n_audio_head = 6
whisper_model_load: n_audio_layer = 4
whisper_model_load: n_text_ctx = 448
whisper_model_load: n_text_state = 384
whisper_model_load: n_text_head = 6
whisper_model_load: n_text_layer = 4
whisper_model_load: n_mels = 80
whisper_model_load: ftype = 1
whisper_model_load: qntvr = 0
whisper_model_load: type = 1 (tiny)
whisper_model_load: adding 1607 extra tokens
whisper_model_load: n_langs = 99
ggml_opencl: selecting platform: 'NVIDIA CUDA'
ggml_opencl: selecting device: 'NVIDIA GeForce GTX 1650'
ggml_opencl: device FP16 support: false
whisper_model_load: CPU buffer size = 77.18 MB
whisper_model_load: model size = 77.11 MB
whisper_init_state: kv self size = 8.26 MB
whisper_init_state: kv cross size = 9.22 MB
whisper_init_state: compute buffer (conv) = 12.17 MB
whisper_init_state: compute buffer (encode) = 64.92 MB
whisper_init_state: compute buffer (cross) = 4.01 MB
whisper_init_state: compute buffer (decode) = 96.02 MB
system_info: n_threads = 4 / 12 | AVX = 1 | AVX2 = 1 | AVX512 = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | METAL = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 1 | SSSE3 = 0 | VSX = 0 | CUDA = 0 | COREML = 0 | OPENVINO = 0 |
main: WARNING: model is not multilingual, ignoring language and translation options
main: processing 'F:\githubsources\whisper.cpp\samples\jfk.wav' (176000 samples, 11.0 sec), 4 threads, 1 processors, 5 beams + best of 5, lang = en, task = transcribe, timestamps = 1 ...
[00:00:00.000 --> 00:00:07.960] And so my fellow Americans ask not what your country can do for you
[00:00:07.960 --> 00:00:10.760] ask what you can do for your country.
whisper_print_timings: load time = 1354.68 ms
whisper_print_timings: fallbacks = 0 p / 0 h
whisper_print_timings: mel time = 12.33 ms
whisper_print_timings: sample time = 103.26 ms / 139 runs ( 0.74 ms per run)
whisper_print_timings: encode time = 395.21 ms / 1 runs ( 395.21 ms per run)
whisper_print_timings: decode time = 7.05 ms / 2 runs ( 3.52 ms per run)
whisper_print_timings: batchd time = 166.18 ms / 133 runs ( 1.25 ms per run)
whisper_print_timings: prompt time = 0.00 ms / 1 runs ( 0.00 ms per run)
whisper_print_timings: total time = 2045.60 ms
windows small:
(base) PS F:\githubsources\whisper.cpp> .\build\bin\Release\main.exe -m .\models\ggml-small-q4_k.bin -l auto F:\githubsources\whisper.cpp\samples\jfk.wav
whisper_init_from_file_with_params_no_state: loading model from '.\models\ggml-small-q4_k.bin'
whisper_model_load: loading model
whisper_model_load: n_vocab = 51865
whisper_model_load: n_audio_ctx = 1500
whisper_model_load: n_audio_state = 768
whisper_model_load: n_audio_head = 12
whisper_model_load: n_audio_layer = 12
whisper_model_load: n_text_ctx = 448
whisper_model_load: n_text_state = 768
whisper_model_load: n_text_head = 12
whisper_model_load: n_text_layer = 12
whisper_model_load: n_mels = 80
whisper_model_load: ftype = 12
whisper_model_load: qntvr = 2
whisper_model_load: type = 3 (small)
whisper_model_load: adding 1608 extra tokens
whisper_model_load: n_langs = 99
ggml_opencl: selecting platform: 'NVIDIA CUDA'
ggml_opencl: selecting device: 'NVIDIA GeForce GTX 1650'
ggml_opencl: device FP16 support: false
whisper_model_load: CPU buffer size = 145.05 MB
whisper_model_load: model size = 144.86 MB
whisper_init_state: kv self size = 49.55 MB
whisper_init_state: kv cross size = 55.30 MB
whisper_init_state: compute buffer (conv) = 20.23 MB
whisper_init_state: compute buffer (encode) = 128.14 MB
whisper_init_state: compute buffer (cross) = 6.31 MB
whisper_init_state: compute buffer (decode) = 97.40 MB
system_info: n_threads = 4 / 12 | AVX = 1 | AVX2 = 1 | AVX512 = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | METAL = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 1 | SSSE3 = 0 | VSX = 0 | CUDA = 0 | COREML = 0 | OPENVINO = 0 |
main: processing 'F:\githubsources\whisper.cpp\samples\jfk.wav' (176000 samples, 11.0 sec), 4 threads, 1 processors, 5 beams + best of 5, lang = auto, task = transcribe, timestamps = 1 ...
whisper_full_with_state: auto-detected language: en (p = 0.472618)
[00:00:00.000 --> 00:00:30.000] .
whisper_print_timings: load time = 1398.95 ms
whisper_print_timings: fallbacks = 1 p / 1 h
whisper_print_timings: mel time = 12.33 ms
whisper_print_timings: sample time = 398.21 ms / 312 runs ( 1.28 ms per run)
whisper_print_timings: encode time = 4430.49 ms / 2 runs ( 2215.24 ms per run)
whisper_print_timings: decode time = 1839.86 ms / 227 runs ( 8.11 ms per run)
whisper_print_timings: batchd time = 456.34 ms / 82 runs ( 5.57 ms per run)
whisper_print_timings: prompt time = 0.00 ms / 1 runs ( 0.00 ms per run)
whisper_print_timings: total time = 8543.37 ms
Only the results of tiny.en in windows nvidia clblast are correct, and the results of any model running on android are incorrect.