whispercpp 1.3.0 → 1.3.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- checksums.yaml +4 -4
- data/.gitignore +5 -0
- data/LICENSE +1 -1
- data/README.md +165 -434
- data/Rakefile +60 -11
- data/ext/.gitignore +13 -0
- data/ext/cpu.mk +9 -0
- data/ext/{dr_wav.h → examples/dr_wav.h} +3560 -1179
- data/ext/extconf.rb +185 -16
- data/ext/ggml/include/ggml-alloc.h +76 -0
- data/ext/ggml/include/ggml-backend.h +352 -0
- data/ext/ggml/include/ggml-blas.h +25 -0
- data/ext/ggml/include/ggml-cann.h +123 -0
- data/ext/ggml/include/ggml-cpp.h +38 -0
- data/ext/ggml/include/ggml-cpu.h +135 -0
- data/ext/ggml/include/ggml-cuda.h +47 -0
- data/ext/ggml/include/ggml-kompute.h +50 -0
- data/ext/ggml/include/ggml-metal.h +66 -0
- data/ext/ggml/include/ggml-opencl.h +26 -0
- data/ext/ggml/include/ggml-opt.h +216 -0
- data/ext/ggml/include/ggml-rpc.h +28 -0
- data/ext/ggml/include/ggml-sycl.h +49 -0
- data/ext/ggml/include/ggml-vulkan.h +31 -0
- data/ext/{ggml.h → ggml/include/ggml.h} +479 -596
- data/ext/ggml/src/ggml-alloc.c +1037 -0
- data/ext/ggml/src/ggml-amx/common.h +94 -0
- data/ext/ggml/src/ggml-amx/ggml-amx.cpp +446 -0
- data/ext/ggml/src/ggml-amx/mmq.cpp +2510 -0
- data/ext/ggml/src/ggml-amx/mmq.h +17 -0
- data/ext/ggml/src/ggml-backend-impl.h +256 -0
- data/ext/ggml/src/ggml-backend-reg.cpp +552 -0
- data/ext/ggml/src/ggml-backend.cpp +1999 -0
- data/ext/ggml/src/ggml-blas/ggml-blas.cpp +517 -0
- data/ext/ggml/src/ggml-cann/acl_tensor.cpp +175 -0
- data/ext/ggml/src/ggml-cann/acl_tensor.h +258 -0
- data/ext/ggml/src/ggml-cann/aclnn_ops.cpp +3427 -0
- data/ext/ggml/src/ggml-cann/aclnn_ops.h +592 -0
- data/ext/ggml/src/ggml-cann/common.h +286 -0
- data/ext/ggml/src/ggml-cann/ggml-cann.cpp +2188 -0
- data/ext/ggml/src/ggml-cann/kernels/ascendc_kernels.h +19 -0
- data/ext/ggml/src/ggml-cann/kernels/dup.cpp +236 -0
- data/ext/ggml/src/ggml-cann/kernels/get_row_f16.cpp +197 -0
- data/ext/ggml/src/ggml-cann/kernels/get_row_f32.cpp +190 -0
- data/ext/ggml/src/ggml-cann/kernels/get_row_q4_0.cpp +204 -0
- data/ext/ggml/src/ggml-cann/kernels/get_row_q8_0.cpp +191 -0
- data/ext/ggml/src/ggml-cann/kernels/quantize_f16_q8_0.cpp +218 -0
- data/ext/ggml/src/ggml-cann/kernels/quantize_f32_q8_0.cpp +216 -0
- data/ext/ggml/src/ggml-cann/kernels/quantize_float_to_q4_0.cpp +295 -0
- data/ext/ggml/src/ggml-common.h +1853 -0
- data/ext/ggml/src/ggml-cpu/amx/amx.cpp +220 -0
- data/ext/ggml/src/ggml-cpu/amx/amx.h +8 -0
- data/ext/ggml/src/ggml-cpu/amx/common.h +91 -0
- data/ext/ggml/src/ggml-cpu/amx/mmq.cpp +2511 -0
- data/ext/ggml/src/ggml-cpu/amx/mmq.h +10 -0
- data/ext/ggml/src/ggml-cpu/cpu-feats-x86.cpp +323 -0
- data/ext/ggml/src/ggml-cpu/ggml-cpu-aarch64.cpp +4262 -0
- data/ext/ggml/src/ggml-cpu/ggml-cpu-aarch64.h +8 -0
- data/ext/ggml/src/ggml-cpu/ggml-cpu-hbm.cpp +55 -0
- data/ext/ggml/src/ggml-cpu/ggml-cpu-hbm.h +8 -0
- data/ext/ggml/src/ggml-cpu/ggml-cpu-impl.h +386 -0
- data/ext/ggml/src/ggml-cpu/ggml-cpu-quants.c +10835 -0
- data/ext/ggml/src/ggml-cpu/ggml-cpu-quants.h +63 -0
- data/ext/ggml/src/ggml-cpu/ggml-cpu-traits.cpp +36 -0
- data/ext/ggml/src/ggml-cpu/ggml-cpu-traits.h +38 -0
- data/ext/ggml/src/ggml-cpu/ggml-cpu.c +14123 -0
- data/ext/ggml/src/ggml-cpu/ggml-cpu.cpp +622 -0
- data/ext/ggml/src/ggml-cpu/llamafile/sgemm.cpp +1884 -0
- data/ext/ggml/src/ggml-cpu/llamafile/sgemm.h +14 -0
- data/ext/ggml/src/ggml-cuda/vendors/cuda.h +14 -0
- data/ext/ggml/src/ggml-cuda/vendors/hip.h +186 -0
- data/ext/ggml/src/ggml-cuda/vendors/musa.h +134 -0
- data/ext/ggml/src/ggml-impl.h +556 -0
- data/ext/ggml/src/ggml-kompute/ggml-kompute.cpp +2251 -0
- data/ext/ggml/src/ggml-metal/ggml-metal-impl.h +288 -0
- data/ext/ggml/src/ggml-metal/ggml-metal.m +4884 -0
- data/ext/ggml/src/ggml-metal/ggml-metal.metal +6732 -0
- data/ext/ggml/src/ggml-opt.cpp +854 -0
- data/ext/ggml/src/ggml-quants.c +5238 -0
- data/ext/ggml/src/ggml-quants.h +100 -0
- data/ext/ggml/src/ggml-rpc/ggml-rpc.cpp +1406 -0
- data/ext/ggml/src/ggml-sycl/common.cpp +95 -0
- data/ext/ggml/src/ggml-sycl/concat.cpp +196 -0
- data/ext/ggml/src/ggml-sycl/conv.cpp +99 -0
- data/ext/ggml/src/ggml-sycl/convert.cpp +547 -0
- data/ext/ggml/src/ggml-sycl/dmmv.cpp +1023 -0
- data/ext/ggml/src/ggml-sycl/element_wise.cpp +1030 -0
- data/ext/ggml/src/ggml-sycl/ggml-sycl.cpp +4729 -0
- data/ext/ggml/src/ggml-sycl/im2col.cpp +126 -0
- data/ext/ggml/src/ggml-sycl/mmq.cpp +3031 -0
- data/ext/ggml/src/ggml-sycl/mmvq.cpp +1015 -0
- data/ext/ggml/src/ggml-sycl/norm.cpp +378 -0
- data/ext/ggml/src/ggml-sycl/outprod.cpp +56 -0
- data/ext/ggml/src/ggml-sycl/rope.cpp +276 -0
- data/ext/ggml/src/ggml-sycl/softmax.cpp +251 -0
- data/ext/ggml/src/ggml-sycl/tsembd.cpp +72 -0
- data/ext/ggml/src/ggml-sycl/wkv6.cpp +141 -0
- data/ext/ggml/src/ggml-threading.cpp +12 -0
- data/ext/ggml/src/ggml-threading.h +14 -0
- data/ext/ggml/src/ggml-vulkan/ggml-vulkan.cpp +8657 -0
- data/ext/ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp +593 -0
- data/ext/ggml/src/ggml.c +7694 -0
- data/ext/{whisper.h → include/whisper.h} +23 -22
- data/ext/metal-embed.mk +17 -0
- data/ext/metal.mk +6 -0
- data/ext/ruby_whisper.cpp +1492 -9
- data/ext/ruby_whisper.h +10 -0
- data/ext/scripts/get-flags.mk +38 -0
- data/ext/src/coreml/whisper-decoder-impl.h +146 -0
- data/ext/src/coreml/whisper-decoder-impl.m +201 -0
- data/ext/src/coreml/whisper-encoder-impl.h +142 -0
- data/ext/src/coreml/whisper-encoder-impl.m +197 -0
- data/ext/src/coreml/whisper-encoder.h +26 -0
- data/ext/src/openvino/whisper-openvino-encoder.cpp +108 -0
- data/ext/src/openvino/whisper-openvino-encoder.h +31 -0
- data/ext/{whisper.cpp → src/whisper.cpp} +661 -492
- data/extsources.rb +6 -0
- data/lib/whisper/model/uri.rb +157 -0
- data/lib/whisper.rb +2 -0
- data/tests/helper.rb +7 -0
- data/tests/jfk_reader/.gitignore +5 -0
- data/tests/jfk_reader/extconf.rb +3 -0
- data/tests/jfk_reader/jfk_reader.c +68 -0
- data/tests/test_callback.rb +160 -0
- data/tests/test_error.rb +20 -0
- data/tests/test_model.rb +71 -0
- data/tests/test_package.rb +31 -0
- data/tests/test_params.rb +160 -0
- data/tests/test_segment.rb +83 -0
- data/tests/test_whisper.rb +211 -123
- data/whispercpp.gemspec +36 -0
- metadata +137 -11
- data/ext/ggml.c +0 -21755
@@ -0,0 +1,100 @@
|
|
1
|
+
#pragma once
|
2
|
+
|
3
|
+
#define GGML_COMMON_DECL_C
|
4
|
+
#include "ggml-common.h"
|
5
|
+
|
6
|
+
#include "ggml.h"
|
7
|
+
|
8
|
+
// GGML internal header
|
9
|
+
|
10
|
+
#ifdef __cplusplus
|
11
|
+
extern "C" {
|
12
|
+
#endif
|
13
|
+
|
14
|
+
// NOTE: these functions are defined as GGML_API because they used by the CPU backend
|
15
|
+
|
16
|
+
// Quantization
|
17
|
+
GGML_API void quantize_row_q4_0_ref(const float * GGML_RESTRICT x, block_q4_0 * GGML_RESTRICT y, int64_t k);
|
18
|
+
GGML_API void quantize_row_q4_1_ref(const float * GGML_RESTRICT x, block_q4_1 * GGML_RESTRICT y, int64_t k);
|
19
|
+
GGML_API void quantize_row_q5_0_ref(const float * GGML_RESTRICT x, block_q5_0 * GGML_RESTRICT y, int64_t k);
|
20
|
+
GGML_API void quantize_row_q5_1_ref(const float * GGML_RESTRICT x, block_q5_1 * GGML_RESTRICT y, int64_t k);
|
21
|
+
GGML_API void quantize_row_q8_0_ref(const float * GGML_RESTRICT x, block_q8_0 * GGML_RESTRICT y, int64_t k);
|
22
|
+
GGML_API void quantize_row_q8_1_ref(const float * GGML_RESTRICT x, block_q8_1 * GGML_RESTRICT y, int64_t k);
|
23
|
+
|
24
|
+
GGML_API void quantize_row_q2_K_ref(const float * GGML_RESTRICT x, block_q2_K * GGML_RESTRICT y, int64_t k);
|
25
|
+
GGML_API void quantize_row_q3_K_ref(const float * GGML_RESTRICT x, block_q3_K * GGML_RESTRICT y, int64_t k);
|
26
|
+
GGML_API void quantize_row_q4_K_ref(const float * GGML_RESTRICT x, block_q4_K * GGML_RESTRICT y, int64_t k);
|
27
|
+
GGML_API void quantize_row_q5_K_ref(const float * GGML_RESTRICT x, block_q5_K * GGML_RESTRICT y, int64_t k);
|
28
|
+
GGML_API void quantize_row_q6_K_ref(const float * GGML_RESTRICT x, block_q6_K * GGML_RESTRICT y, int64_t k);
|
29
|
+
GGML_API void quantize_row_q8_K_ref(const float * GGML_RESTRICT x, block_q8_K * GGML_RESTRICT y, int64_t k);
|
30
|
+
|
31
|
+
GGML_API void quantize_row_tq1_0_ref(const float * GGML_RESTRICT x, block_tq1_0 * GGML_RESTRICT y, int64_t k);
|
32
|
+
GGML_API void quantize_row_tq2_0_ref(const float * GGML_RESTRICT x, block_tq2_0 * GGML_RESTRICT y, int64_t k);
|
33
|
+
|
34
|
+
GGML_API void quantize_row_iq3_xxs_ref(const float * GGML_RESTRICT x, block_iq3_xxs * GGML_RESTRICT y, int64_t k);
|
35
|
+
GGML_API void quantize_row_iq4_nl_ref (const float * GGML_RESTRICT x, block_iq4_nl * GGML_RESTRICT y, int64_t k);
|
36
|
+
GGML_API void quantize_row_iq4_xs_ref (const float * GGML_RESTRICT x, block_iq4_xs * GGML_RESTRICT y, int64_t k);
|
37
|
+
GGML_API void quantize_row_iq3_s_ref (const float * GGML_RESTRICT x, block_iq3_s * GGML_RESTRICT y, int64_t k);
|
38
|
+
GGML_API void quantize_row_iq2_s_ref (const float * GGML_RESTRICT x, block_iq2_s * GGML_RESTRICT y, int64_t k);
|
39
|
+
|
40
|
+
// Dequantization
|
41
|
+
GGML_API void dequantize_row_q4_0(const block_q4_0 * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
42
|
+
GGML_API void dequantize_row_q4_1(const block_q4_1 * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
43
|
+
GGML_API void dequantize_row_q5_0(const block_q5_0 * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
44
|
+
GGML_API void dequantize_row_q5_1(const block_q5_1 * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
45
|
+
GGML_API void dequantize_row_q8_0(const block_q8_0 * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
46
|
+
//GGML_API void dequantize_row_q8_1(const block_q8_1 * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
47
|
+
|
48
|
+
GGML_API void dequantize_row_q2_K(const block_q2_K * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
49
|
+
GGML_API void dequantize_row_q3_K(const block_q3_K * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
50
|
+
GGML_API void dequantize_row_q4_K(const block_q4_K * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
51
|
+
GGML_API void dequantize_row_q5_K(const block_q5_K * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
52
|
+
GGML_API void dequantize_row_q6_K(const block_q6_K * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
53
|
+
GGML_API void dequantize_row_q8_K(const block_q8_K * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
54
|
+
|
55
|
+
GGML_API void dequantize_row_tq1_0(const block_tq1_0 * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
56
|
+
GGML_API void dequantize_row_tq2_0(const block_tq2_0 * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
57
|
+
|
58
|
+
GGML_API void dequantize_row_iq2_xxs(const block_iq2_xxs * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
59
|
+
GGML_API void dequantize_row_iq2_xs (const block_iq2_xs * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
60
|
+
GGML_API void dequantize_row_iq2_s (const block_iq2_s * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
61
|
+
GGML_API void dequantize_row_iq3_xxs(const block_iq3_xxs * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
62
|
+
GGML_API void dequantize_row_iq1_s (const block_iq1_s * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
63
|
+
GGML_API void dequantize_row_iq1_m (const block_iq1_m * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
64
|
+
GGML_API void dequantize_row_iq4_nl (const block_iq4_nl * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
65
|
+
GGML_API void dequantize_row_iq4_xs (const block_iq4_xs * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
66
|
+
GGML_API void dequantize_row_iq3_s (const block_iq3_s * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
|
67
|
+
|
68
|
+
// Quantization utilizing an importance matrix (a.k.a. "Activation aWare Quantization")
|
69
|
+
GGML_API size_t quantize_iq2_xxs(const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
70
|
+
GGML_API size_t quantize_iq2_xs (const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
71
|
+
GGML_API size_t quantize_iq2_s (const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
72
|
+
GGML_API size_t quantize_iq3_xxs(const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
73
|
+
GGML_API size_t quantize_iq1_s (const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
74
|
+
GGML_API size_t quantize_iq1_m (const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
75
|
+
GGML_API size_t quantize_iq4_nl (const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
76
|
+
GGML_API size_t quantize_iq4_xs (const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
77
|
+
GGML_API size_t quantize_iq3_s (const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
78
|
+
|
79
|
+
GGML_API size_t quantize_tq1_0(const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
80
|
+
GGML_API size_t quantize_tq2_0(const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
81
|
+
|
82
|
+
GGML_API size_t quantize_q2_K(const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
83
|
+
GGML_API size_t quantize_q3_K(const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
84
|
+
GGML_API size_t quantize_q4_K(const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
85
|
+
GGML_API size_t quantize_q5_K(const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
86
|
+
GGML_API size_t quantize_q6_K(const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
87
|
+
GGML_API size_t quantize_q4_0(const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
88
|
+
GGML_API size_t quantize_q4_1(const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
89
|
+
GGML_API size_t quantize_q5_0(const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
90
|
+
GGML_API size_t quantize_q5_1(const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
91
|
+
GGML_API size_t quantize_q8_0(const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
|
92
|
+
|
93
|
+
GGML_API void iq2xs_init_impl(enum ggml_type type);
|
94
|
+
GGML_API void iq2xs_free_impl(enum ggml_type type);
|
95
|
+
GGML_API void iq3xs_init_impl(int grid_size);
|
96
|
+
GGML_API void iq3xs_free_impl(int grid_size);
|
97
|
+
|
98
|
+
#ifdef __cplusplus
|
99
|
+
}
|
100
|
+
#endif
|