whispercpp 1.2.0.2 → 1.3.1

Sign up to get free protection for your applications and to get access to all the features.
Files changed (135) hide show
  1. checksums.yaml +4 -4
  2. data/.gitignore +5 -0
  3. data/LICENSE +1 -1
  4. data/README.md +165 -434
  5. data/Rakefile +46 -86
  6. data/ext/.gitignore +13 -0
  7. data/ext/cpu.mk +9 -0
  8. data/ext/{dr_wav.h → examples/dr_wav.h} +3560 -1179
  9. data/ext/extconf.rb +185 -7
  10. data/ext/ggml/include/ggml-alloc.h +76 -0
  11. data/ext/ggml/include/ggml-backend.h +352 -0
  12. data/ext/ggml/include/ggml-blas.h +25 -0
  13. data/ext/ggml/include/ggml-cann.h +123 -0
  14. data/ext/ggml/include/ggml-cpp.h +38 -0
  15. data/ext/ggml/include/ggml-cpu.h +135 -0
  16. data/ext/ggml/include/ggml-cuda.h +47 -0
  17. data/ext/ggml/include/ggml-kompute.h +50 -0
  18. data/ext/ggml/include/ggml-metal.h +66 -0
  19. data/ext/ggml/include/ggml-opencl.h +26 -0
  20. data/ext/ggml/include/ggml-opt.h +216 -0
  21. data/ext/ggml/include/ggml-rpc.h +28 -0
  22. data/ext/ggml/include/ggml-sycl.h +49 -0
  23. data/ext/ggml/include/ggml-vulkan.h +31 -0
  24. data/ext/ggml/include/ggml.h +2285 -0
  25. data/ext/ggml/src/ggml-alloc.c +1037 -0
  26. data/ext/ggml/src/ggml-amx/common.h +94 -0
  27. data/ext/ggml/src/ggml-amx/ggml-amx.cpp +446 -0
  28. data/ext/ggml/src/ggml-amx/mmq.cpp +2510 -0
  29. data/ext/ggml/src/ggml-amx/mmq.h +17 -0
  30. data/ext/ggml/src/ggml-backend-impl.h +256 -0
  31. data/ext/ggml/src/ggml-backend-reg.cpp +552 -0
  32. data/ext/ggml/src/ggml-backend.cpp +1999 -0
  33. data/ext/ggml/src/ggml-blas/ggml-blas.cpp +517 -0
  34. data/ext/ggml/src/ggml-cann/acl_tensor.cpp +175 -0
  35. data/ext/ggml/src/ggml-cann/acl_tensor.h +258 -0
  36. data/ext/ggml/src/ggml-cann/aclnn_ops.cpp +3427 -0
  37. data/ext/ggml/src/ggml-cann/aclnn_ops.h +592 -0
  38. data/ext/ggml/src/ggml-cann/common.h +286 -0
  39. data/ext/ggml/src/ggml-cann/ggml-cann.cpp +2188 -0
  40. data/ext/ggml/src/ggml-cann/kernels/ascendc_kernels.h +19 -0
  41. data/ext/ggml/src/ggml-cann/kernels/dup.cpp +236 -0
  42. data/ext/ggml/src/ggml-cann/kernels/get_row_f16.cpp +197 -0
  43. data/ext/ggml/src/ggml-cann/kernels/get_row_f32.cpp +190 -0
  44. data/ext/ggml/src/ggml-cann/kernels/get_row_q4_0.cpp +204 -0
  45. data/ext/ggml/src/ggml-cann/kernels/get_row_q8_0.cpp +191 -0
  46. data/ext/ggml/src/ggml-cann/kernels/quantize_f16_q8_0.cpp +218 -0
  47. data/ext/ggml/src/ggml-cann/kernels/quantize_f32_q8_0.cpp +216 -0
  48. data/ext/ggml/src/ggml-cann/kernels/quantize_float_to_q4_0.cpp +295 -0
  49. data/ext/ggml/src/ggml-common.h +1853 -0
  50. data/ext/ggml/src/ggml-cpu/amx/amx.cpp +220 -0
  51. data/ext/ggml/src/ggml-cpu/amx/amx.h +8 -0
  52. data/ext/ggml/src/ggml-cpu/amx/common.h +91 -0
  53. data/ext/ggml/src/ggml-cpu/amx/mmq.cpp +2511 -0
  54. data/ext/ggml/src/ggml-cpu/amx/mmq.h +10 -0
  55. data/ext/ggml/src/ggml-cpu/cpu-feats-x86.cpp +323 -0
  56. data/ext/ggml/src/ggml-cpu/ggml-cpu-aarch64.cpp +4262 -0
  57. data/ext/ggml/src/ggml-cpu/ggml-cpu-aarch64.h +8 -0
  58. data/ext/ggml/src/ggml-cpu/ggml-cpu-hbm.cpp +55 -0
  59. data/ext/ggml/src/ggml-cpu/ggml-cpu-hbm.h +8 -0
  60. data/ext/ggml/src/ggml-cpu/ggml-cpu-impl.h +386 -0
  61. data/ext/ggml/src/ggml-cpu/ggml-cpu-quants.c +10835 -0
  62. data/ext/ggml/src/ggml-cpu/ggml-cpu-quants.h +63 -0
  63. data/ext/ggml/src/ggml-cpu/ggml-cpu-traits.cpp +36 -0
  64. data/ext/ggml/src/ggml-cpu/ggml-cpu-traits.h +38 -0
  65. data/ext/ggml/src/ggml-cpu/ggml-cpu.c +14123 -0
  66. data/ext/ggml/src/ggml-cpu/ggml-cpu.cpp +622 -0
  67. data/ext/ggml/src/ggml-cpu/llamafile/sgemm.cpp +1884 -0
  68. data/ext/ggml/src/ggml-cpu/llamafile/sgemm.h +14 -0
  69. data/ext/ggml/src/ggml-cuda/vendors/cuda.h +14 -0
  70. data/ext/ggml/src/ggml-cuda/vendors/hip.h +186 -0
  71. data/ext/ggml/src/ggml-cuda/vendors/musa.h +134 -0
  72. data/ext/ggml/src/ggml-impl.h +556 -0
  73. data/ext/ggml/src/ggml-kompute/ggml-kompute.cpp +2251 -0
  74. data/ext/ggml/src/ggml-metal/ggml-metal-impl.h +288 -0
  75. data/ext/ggml/src/ggml-metal/ggml-metal.m +4884 -0
  76. data/ext/ggml/src/ggml-metal/ggml-metal.metal +6732 -0
  77. data/ext/ggml/src/ggml-opt.cpp +854 -0
  78. data/ext/ggml/src/ggml-quants.c +5238 -0
  79. data/ext/ggml/src/ggml-quants.h +100 -0
  80. data/ext/ggml/src/ggml-rpc/ggml-rpc.cpp +1406 -0
  81. data/ext/ggml/src/ggml-sycl/common.cpp +95 -0
  82. data/ext/ggml/src/ggml-sycl/concat.cpp +196 -0
  83. data/ext/ggml/src/ggml-sycl/conv.cpp +99 -0
  84. data/ext/ggml/src/ggml-sycl/convert.cpp +547 -0
  85. data/ext/ggml/src/ggml-sycl/dmmv.cpp +1023 -0
  86. data/ext/ggml/src/ggml-sycl/element_wise.cpp +1030 -0
  87. data/ext/ggml/src/ggml-sycl/ggml-sycl.cpp +4729 -0
  88. data/ext/ggml/src/ggml-sycl/im2col.cpp +126 -0
  89. data/ext/ggml/src/ggml-sycl/mmq.cpp +3031 -0
  90. data/ext/ggml/src/ggml-sycl/mmvq.cpp +1015 -0
  91. data/ext/ggml/src/ggml-sycl/norm.cpp +378 -0
  92. data/ext/ggml/src/ggml-sycl/outprod.cpp +56 -0
  93. data/ext/ggml/src/ggml-sycl/rope.cpp +276 -0
  94. data/ext/ggml/src/ggml-sycl/softmax.cpp +251 -0
  95. data/ext/ggml/src/ggml-sycl/tsembd.cpp +72 -0
  96. data/ext/ggml/src/ggml-sycl/wkv6.cpp +141 -0
  97. data/ext/ggml/src/ggml-threading.cpp +12 -0
  98. data/ext/ggml/src/ggml-threading.h +14 -0
  99. data/ext/ggml/src/ggml-vulkan/ggml-vulkan.cpp +8657 -0
  100. data/ext/ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp +593 -0
  101. data/ext/ggml/src/ggml.c +7694 -0
  102. data/ext/include/whisper.h +672 -0
  103. data/ext/metal-embed.mk +17 -0
  104. data/ext/metal.mk +6 -0
  105. data/ext/ruby_whisper.cpp +1608 -159
  106. data/ext/ruby_whisper.h +10 -0
  107. data/ext/scripts/get-flags.mk +38 -0
  108. data/ext/src/coreml/whisper-decoder-impl.h +146 -0
  109. data/ext/src/coreml/whisper-decoder-impl.m +201 -0
  110. data/ext/src/coreml/whisper-encoder-impl.h +142 -0
  111. data/ext/src/coreml/whisper-encoder-impl.m +197 -0
  112. data/ext/src/coreml/whisper-encoder.h +26 -0
  113. data/ext/src/openvino/whisper-openvino-encoder.cpp +108 -0
  114. data/ext/src/openvino/whisper-openvino-encoder.h +31 -0
  115. data/ext/src/whisper.cpp +7393 -0
  116. data/extsources.rb +6 -0
  117. data/lib/whisper/model/uri.rb +157 -0
  118. data/lib/whisper.rb +2 -0
  119. data/tests/helper.rb +7 -0
  120. data/tests/jfk_reader/.gitignore +5 -0
  121. data/tests/jfk_reader/extconf.rb +3 -0
  122. data/tests/jfk_reader/jfk_reader.c +68 -0
  123. data/tests/test_callback.rb +160 -0
  124. data/tests/test_error.rb +20 -0
  125. data/tests/test_model.rb +71 -0
  126. data/tests/test_package.rb +31 -0
  127. data/tests/test_params.rb +160 -0
  128. data/tests/test_segment.rb +83 -0
  129. data/tests/test_whisper.rb +211 -123
  130. data/whispercpp.gemspec +36 -0
  131. metadata +137 -11
  132. data/ext/ggml.c +0 -8616
  133. data/ext/ggml.h +0 -748
  134. data/ext/whisper.cpp +0 -4829
  135. data/ext/whisper.h +0 -402
data/ext/ggml/src/ggml-metal/ggml-metal-impl.h ADDED
@@ -0,0 +1,288 @@
1
+ #ifndef GGML_METAL_IMPL
2
+ #define GGML_METAL_IMPL
3
+
4
+ // kernel argument structs
5
+ //
6
+ // - element counters (e.g. ne00) typically use int32_t to reduce register usage
7
+ // however, be careful from int overflows when using those in the kernel implementation
8
+ //
9
+ // - strides (e.g. nb00) use uint64_t
10
+
11
+ typedef struct {
12
+ int32_t ne00;
13
+ int32_t ne01;
14
+ int32_t ne02;
15
+ int32_t ne03;
16
+ uint64_t nb00;
17
+ uint64_t nb01;
18
+ uint64_t nb02;
19
+ uint64_t nb03;
20
+ int32_t ne10;
21
+ int32_t ne11;
22
+ int32_t ne12;
23
+ int32_t ne13;
24
+ uint64_t nb10;
25
+ uint64_t nb11;
26
+ uint64_t nb12;
27
+ uint64_t nb13;
28
+ int32_t ne0;
29
+ int32_t ne1;
30
+ int32_t ne2;
31
+ int32_t ne3;
32
+ uint64_t nb0;
33
+ uint64_t nb1;
34
+ uint64_t nb2;
35
+ uint64_t nb3;
36
+ int32_t dim;
37
+ } ggml_metal_kargs_concat;
38
+
39
+ typedef struct {
40
+ int32_t ne00;
41
+ int32_t ne01;
42
+ int32_t ne02;
43
+ int32_t ne03;
44
+ uint64_t nb00;
45
+ uint64_t nb01;
46
+ uint64_t nb02;
47
+ uint64_t nb03;
48
+ int32_t ne10;
49
+ int32_t ne11;
50
+ int32_t ne12;
51
+ int32_t ne13;
52
+ uint64_t nb10;
53
+ uint64_t nb11;
54
+ uint64_t nb12;
55
+ uint64_t nb13;
56
+ int32_t ne0;
57
+ int32_t ne1;
58
+ int32_t ne2;
59
+ int32_t ne3;
60
+ uint64_t nb0;
61
+ uint64_t nb1;
62
+ uint64_t nb2;
63
+ uint64_t nb3;
64
+ uint64_t offs;
65
+ } ggml_metal_kargs_bin;
66
+
67
+ typedef struct {
68
+ int32_t ne00;
69
+ int32_t ne01;
70
+ int32_t ne02;
71
+ int32_t ne03;
72
+ uint64_t nb00;
73
+ uint64_t nb01;
74
+ uint64_t nb02;
75
+ uint64_t nb03;
76
+ int32_t ne0;
77
+ int32_t ne1;
78
+ int32_t ne2;
79
+ int32_t ne3;
80
+ uint64_t nb0;
81
+ uint64_t nb1;
82
+ uint64_t nb2;
83
+ uint64_t nb3;
84
+ } ggml_metal_kargs_repeat;
85
+
86
+ typedef struct {
87
+ int64_t ne00;
88
+ int64_t ne01;
89
+ int64_t ne02;
90
+ int64_t ne03;
91
+ uint64_t nb00;
92
+ uint64_t nb01;
93
+ uint64_t nb02;
94
+ uint64_t nb03;
95
+ int64_t ne0;
96
+ int64_t ne1;
97
+ int64_t ne2;
98
+ int64_t ne3;
99
+ uint64_t nb0;
100
+ uint64_t nb1;
101
+ uint64_t nb2;
102
+ uint64_t nb3;
103
+ } ggml_metal_kargs_cpy;
104
+
105
+ typedef struct {
106
+ int64_t ne10;
107
+ int64_t ne11;
108
+ int64_t ne12;
109
+ uint64_t nb10;
110
+ uint64_t nb11;
111
+ uint64_t nb12;
112
+ uint64_t nb13;
113
+ uint64_t nb1;
114
+ uint64_t nb2;
115
+ uint64_t nb3;
116
+ uint64_t offs;
117
+ bool inplace;
118
+ } ggml_metal_kargs_set;
119
+
120
+ typedef struct {
121
+ int32_t ne00;
122
+ int32_t ne01;
123
+ int32_t ne02;
124
+ int32_t ne03;
125
+ uint64_t nb00;
126
+ uint64_t nb01;
127
+ uint64_t nb02;
128
+ uint64_t nb03;
129
+ int32_t ne0;
130
+ int32_t ne1;
131
+ int32_t ne2;
132
+ int32_t ne3;
133
+ uint64_t nb0;
134
+ uint64_t nb1;
135
+ uint64_t nb2;
136
+ uint64_t nb3;
137
+ int32_t n_past;
138
+ int32_t n_dims;
139
+ int32_t n_ctx_orig;
140
+ float freq_base;
141
+ float freq_scale;
142
+ float ext_factor;
143
+ float attn_factor;
144
+ float beta_fast;
145
+ float beta_slow;
146
+ } ggml_metal_kargs_rope;
147
+
148
+ typedef struct {
149
+ int32_t ne01;
150
+ int32_t ne02;
151
+ int32_t ne03;
152
+ uint64_t nb01;
153
+ uint64_t nb02;
154
+ uint64_t nb03;
155
+ int32_t ne11;
156
+ int32_t ne_12_2; // assume K and V are same shape
157
+ int32_t ne_12_3;
158
+ uint64_t nb_12_1;
159
+ uint64_t nb_12_2;
160
+ uint64_t nb_12_3;
161
+ uint64_t nb31;
162
+ int32_t ne1;
163
+ int32_t ne2;
164
+ float scale;
165
+ float max_bias;
166
+ float m0;
167
+ float m1;
168
+ uint16_t n_head_log2;
169
+ float logit_softcap;
170
+ } ggml_metal_kargs_flash_attn_ext;
171
+
172
+ typedef struct {
173
+ int32_t ne00;
174
+ int32_t ne02;
175
+ uint64_t nb01;
176
+ uint64_t nb02;
177
+ uint64_t nb03;
178
+ int32_t ne12;
179
+ uint64_t nb10;
180
+ uint64_t nb11;
181
+ uint64_t nb12;
182
+ uint64_t nb13;
183
+ int32_t ne0;
184
+ int32_t ne1;
185
+ int16_t r2;
186
+ int16_t r3;
187
+ } ggml_metal_kargs_mul_mm;
188
+
189
+ typedef struct {
190
+ int32_t ne00;
191
+ int32_t ne01;
192
+ int32_t ne02;
193
+ uint64_t nb00;
194
+ uint64_t nb01;
195
+ uint64_t nb02;
196
+ uint64_t nb03;
197
+ int32_t ne10;
198
+ int32_t ne11;
199
+ int32_t ne12;
200
+ uint64_t nb10;
201
+ uint64_t nb11;
202
+ uint64_t nb12;
203
+ uint64_t nb13;
204
+ int32_t ne0;
205
+ int32_t ne1;
206
+ int16_t r2;
207
+ int16_t r3;
208
+ } ggml_metal_kargs_mul_mv;
209
+
210
+ typedef struct {
211
+ int32_t ne00;
212
+ int32_t ne01;
213
+ int32_t ne02;
214
+ uint64_t nb00;
215
+ uint64_t nb01;
216
+ uint64_t nb02;
217
+ uint64_t nb03;
218
+ int32_t ne10;
219
+ int32_t ne11;
220
+ int32_t ne12;
221
+ uint64_t nb10;
222
+ uint64_t nb11;
223
+ uint64_t nb12;
224
+ uint64_t nb13;
225
+ int32_t ne0;
226
+ int32_t ne1;
227
+ int16_t r2;
228
+ int16_t r3;
229
+ int16_t nsg;
230
+ int16_t nxpsg;
231
+ int16_t r1ptg;
232
+ } ggml_metal_kargs_mul_mv_ext;
233
+
234
+ typedef struct {
235
+ int32_t nei0;
236
+ int32_t nei1;
237
+ uint64_t nbi1;
238
+ int32_t ne00;
239
+ int32_t ne02;
240
+ uint64_t nb01;
241
+ uint64_t nb02;
242
+ int32_t ne11;
243
+ int32_t ne12;
244
+ int32_t ne13;
245
+ uint64_t nb10;
246
+ uint64_t nb11;
247
+ uint64_t nb12;
248
+ int32_t ne0;
249
+ int32_t ne1;
250
+ } ggml_metal_kargs_mul_mm_id;
251
+
252
+ typedef struct {
253
+ int32_t nei0;
254
+ int32_t nei1;
255
+ uint64_t nbi1;
256
+ int32_t ne00;
257
+ int32_t ne01;
258
+ int32_t ne02;
259
+ uint64_t nb00;
260
+ uint64_t nb01;
261
+ uint64_t nb02;
262
+ int32_t ne10;
263
+ int32_t ne11;
264
+ int32_t ne12;
265
+ int32_t ne13;
266
+ uint64_t nb10;
267
+ uint64_t nb11;
268
+ uint64_t nb12;
269
+ int32_t ne0;
270
+ int32_t ne1;
271
+ uint64_t nb1;
272
+ } ggml_metal_kargs_mul_mv_id;
273
+
274
+ typedef struct {
275
+ int32_t ne00;
276
+ int32_t ne00_4;
277
+ uint64_t nb01;
278
+ float eps;
279
+ } ggml_metal_kargs_norm;
280
+
281
+ typedef struct {
282
+ int32_t ne00;
283
+ int32_t ne00_4;
284
+ uint64_t nb01;
285
+ float eps;
286
+ } ggml_metal_kargs_rms_norm;
287
+
288
+ #endif // GGML_METAL_IMPL