llama_cpp 0.16.1 → 0.17.0
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/CHANGELOG.md +25 -0
- data/README.md +7 -12
- data/ext/llama_cpp/extconf.rb +2 -42
- data/ext/llama_cpp/llama_cpp.cpp +20 -0
- data/lib/llama_cpp/version.rb +3 -3
- data/sig/llama_cpp.rbs +5 -0
- metadata +2 -171
- data/vendor/include/.gitkeep +0 -0
- data/vendor/lib/.gitkeep +0 -0
- data/vendor/tmp/llama.cpp/LICENSE +0 -21
- data/vendor/tmp/llama.cpp/Makefile +0 -1116
- data/vendor/tmp/llama.cpp/ggml-alloc.c +0 -1041
- data/vendor/tmp/llama.cpp/ggml-alloc.h +0 -76
- data/vendor/tmp/llama.cpp/ggml-backend-impl.h +0 -153
- data/vendor/tmp/llama.cpp/ggml-backend.c +0 -2214
- data/vendor/tmp/llama.cpp/ggml-backend.h +0 -233
- data/vendor/tmp/llama.cpp/ggml-blas.cpp +0 -363
- data/vendor/tmp/llama.cpp/ggml-blas.h +0 -23
- data/vendor/tmp/llama.cpp/ggml-common.h +0 -1805
- data/vendor/tmp/llama.cpp/ggml-cuda/acc.cu +0 -47
- data/vendor/tmp/llama.cpp/ggml-cuda/arange.cu +0 -34
- data/vendor/tmp/llama.cpp/ggml-cuda/argsort.cu +0 -104
- data/vendor/tmp/llama.cpp/ggml-cuda/binbcast.cu +0 -280
- data/vendor/tmp/llama.cpp/ggml-cuda/clamp.cu +0 -34
- data/vendor/tmp/llama.cpp/ggml-cuda/concat.cu +0 -196
- data/vendor/tmp/llama.cpp/ggml-cuda/convert.cu +0 -686
- data/vendor/tmp/llama.cpp/ggml-cuda/cpy.cu +0 -490
- data/vendor/tmp/llama.cpp/ggml-cuda/diagmask.cu +0 -40
- data/vendor/tmp/llama.cpp/ggml-cuda/dmmv.cu +0 -674
- data/vendor/tmp/llama.cpp/ggml-cuda/fattn-tile-f16.cu +0 -319
- data/vendor/tmp/llama.cpp/ggml-cuda/fattn-tile-f32.cu +0 -312
- data/vendor/tmp/llama.cpp/ggml-cuda/fattn.cu +0 -345
- data/vendor/tmp/llama.cpp/ggml-cuda/getrows.cu +0 -178
- data/vendor/tmp/llama.cpp/ggml-cuda/im2col.cu +0 -104
- data/vendor/tmp/llama.cpp/ggml-cuda/mmq.cu +0 -88
- data/vendor/tmp/llama.cpp/ggml-cuda/mmvq.cu +0 -419
- data/vendor/tmp/llama.cpp/ggml-cuda/norm.cu +0 -221
- data/vendor/tmp/llama.cpp/ggml-cuda/pad.cu +0 -49
- data/vendor/tmp/llama.cpp/ggml-cuda/pool2d.cu +0 -94
- data/vendor/tmp/llama.cpp/ggml-cuda/quantize.cu +0 -112
- data/vendor/tmp/llama.cpp/ggml-cuda/rope.cu +0 -271
- data/vendor/tmp/llama.cpp/ggml-cuda/scale.cu +0 -31
- data/vendor/tmp/llama.cpp/ggml-cuda/softmax.cu +0 -206
- data/vendor/tmp/llama.cpp/ggml-cuda/sumrows.cu +0 -40
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-f16.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q8_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-f16.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q8_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-f16.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q8_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-f16.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q8_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-f16.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q8_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-f16.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q8_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs256-f16-f16.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-f16.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q8_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-f16.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q8_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-f16.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q8_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-f16.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q8_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-f16.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q8_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-f16.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q8_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-f16.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q8_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs256-f16-f16.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-f16.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q8_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqfloat-cpb16.cu +0 -10
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqfloat-cpb32.cu +0 -9
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb16.cu +0 -10
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb32.cu +0 -10
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb8.cu +0 -8
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q2_k.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q3_k.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q4_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q4_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q4_k.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q5_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q5_1.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q5_k.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q6_k.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q8_0.cu +0 -5
- data/vendor/tmp/llama.cpp/ggml-cuda/tsembd.cu +0 -47
- data/vendor/tmp/llama.cpp/ggml-cuda/unary.cu +0 -286
- data/vendor/tmp/llama.cpp/ggml-cuda/upscale.cu +0 -51
- data/vendor/tmp/llama.cpp/ggml-cuda.cu +0 -3069
- data/vendor/tmp/llama.cpp/ggml-cuda.h +0 -44
- data/vendor/tmp/llama.cpp/ggml-impl.h +0 -651
- data/vendor/tmp/llama.cpp/ggml-kompute.cpp +0 -2038
- data/vendor/tmp/llama.cpp/ggml-kompute.h +0 -46
- data/vendor/tmp/llama.cpp/ggml-metal.h +0 -66
- data/vendor/tmp/llama.cpp/ggml-metal.m +0 -3267
- data/vendor/tmp/llama.cpp/ggml-metal.metal +0 -6540
- data/vendor/tmp/llama.cpp/ggml-quants.c +0 -14380
- data/vendor/tmp/llama.cpp/ggml-quants.h +0 -133
- data/vendor/tmp/llama.cpp/ggml-rpc.cpp +0 -1173
- data/vendor/tmp/llama.cpp/ggml-rpc.h +0 -24
- data/vendor/tmp/llama.cpp/ggml-sycl.cpp +0 -17429
- data/vendor/tmp/llama.cpp/ggml-sycl.h +0 -49
- data/vendor/tmp/llama.cpp/ggml-vulkan-shaders.hpp +0 -140820
- data/vendor/tmp/llama.cpp/ggml-vulkan.cpp +0 -7271
- data/vendor/tmp/llama.cpp/ggml-vulkan.h +0 -29
- data/vendor/tmp/llama.cpp/ggml.c +0 -22589
- data/vendor/tmp/llama.cpp/ggml.h +0 -2452
- data/vendor/tmp/llama.cpp/llama.cpp +0 -18692
- data/vendor/tmp/llama.cpp/llama.h +0 -1143
- data/vendor/tmp/llama.cpp/scripts/get-flags.mk +0 -38
- data/vendor/tmp/llama.cpp/sgemm.cpp +0 -1030
- data/vendor/tmp/llama.cpp/sgemm.h +0 -14
- data/vendor/tmp/llama.cpp/unicode-data.cpp +0 -6983
- data/vendor/tmp/llama.cpp/unicode-data.h +0 -20
- data/vendor/tmp/llama.cpp/unicode.cpp +0 -796
- data/vendor/tmp/llama.cpp/unicode.h +0 -63
|
@@ -1,1116 +0,0 @@
|
|
|
1
|
-
# Define the default target now so that it is always the first target
|
|
2
|
-
BUILD_TARGETS = \
|
|
3
|
-
libllava.a \
|
|
4
|
-
llama-baby-llama \
|
|
5
|
-
llama-batched \
|
|
6
|
-
llama-batched-bench \
|
|
7
|
-
llama-bench \
|
|
8
|
-
llama-benchmark-matmult \
|
|
9
|
-
llama-cli \
|
|
10
|
-
llama-convert-llama2c-to-ggml \
|
|
11
|
-
llama-embedding \
|
|
12
|
-
llama-eval-callback \
|
|
13
|
-
llama-export-lora \
|
|
14
|
-
llama-finetune \
|
|
15
|
-
llama-gbnf-validator \
|
|
16
|
-
llama-gguf \
|
|
17
|
-
llama-gguf-split \
|
|
18
|
-
llama-gritlm \
|
|
19
|
-
llama-imatrix \
|
|
20
|
-
llama-infill \
|
|
21
|
-
llama-llava-cli \
|
|
22
|
-
llama-lookahead \
|
|
23
|
-
llama-lookup \
|
|
24
|
-
llama-lookup-create \
|
|
25
|
-
llama-lookup-merge \
|
|
26
|
-
llama-lookup-stats \
|
|
27
|
-
llama-parallel \
|
|
28
|
-
llama-passkey \
|
|
29
|
-
llama-perplexity \
|
|
30
|
-
llama-q8dot \
|
|
31
|
-
llama-quantize \
|
|
32
|
-
llama-quantize-stats \
|
|
33
|
-
llama-retrieval \
|
|
34
|
-
llama-save-load-state \
|
|
35
|
-
llama-server \
|
|
36
|
-
llama-simple \
|
|
37
|
-
llama-speculative \
|
|
38
|
-
llama-tokenize \
|
|
39
|
-
llama-train-text-from-scratch \
|
|
40
|
-
llama-vdot \
|
|
41
|
-
tests/test-c.o
|
|
42
|
-
|
|
43
|
-
# Binaries only useful for tests
|
|
44
|
-
TEST_TARGETS = \
|
|
45
|
-
tests/test-autorelease \
|
|
46
|
-
tests/test-backend-ops \
|
|
47
|
-
tests/test-double-float \
|
|
48
|
-
tests/test-grad0 \
|
|
49
|
-
tests/test-grammar-integration \
|
|
50
|
-
tests/test-grammar-parser \
|
|
51
|
-
tests/test-json-schema-to-grammar \
|
|
52
|
-
tests/test-llama-grammar \
|
|
53
|
-
tests/test-model-load-cancel \
|
|
54
|
-
tests/test-opt \
|
|
55
|
-
tests/test-quantize-fns \
|
|
56
|
-
tests/test-quantize-perf \
|
|
57
|
-
tests/test-rope \
|
|
58
|
-
tests/test-sampling \
|
|
59
|
-
tests/test-tokenizer-0 \
|
|
60
|
-
tests/test-tokenizer-1-bpe \
|
|
61
|
-
tests/test-tokenizer-1-spm
|
|
62
|
-
|
|
63
|
-
# Code coverage output files
|
|
64
|
-
COV_TARGETS = *.gcno tests/*.gcno *.gcda tests/*.gcda *.gcov tests/*.gcov lcov-report gcovr-report
|
|
65
|
-
|
|
66
|
-
ifndef UNAME_S
|
|
67
|
-
UNAME_S := $(shell uname -s)
|
|
68
|
-
endif
|
|
69
|
-
|
|
70
|
-
ifndef UNAME_P
|
|
71
|
-
UNAME_P := $(shell uname -p)
|
|
72
|
-
endif
|
|
73
|
-
|
|
74
|
-
ifndef UNAME_M
|
|
75
|
-
UNAME_M := $(shell uname -m)
|
|
76
|
-
endif
|
|
77
|
-
|
|
78
|
-
# In GNU make default CXX is g++ instead of c++. Let's fix that so that users
|
|
79
|
-
# of non-gcc compilers don't have to provide g++ alias or wrapper.
|
|
80
|
-
DEFCC := cc
|
|
81
|
-
DEFCXX := c++
|
|
82
|
-
ifeq ($(origin CC),default)
|
|
83
|
-
CC := $(DEFCC)
|
|
84
|
-
endif
|
|
85
|
-
ifeq ($(origin CXX),default)
|
|
86
|
-
CXX := $(DEFCXX)
|
|
87
|
-
endif
|
|
88
|
-
|
|
89
|
-
# Mac OS + Arm can report x86_64
|
|
90
|
-
# ref: https://github.com/ggerganov/whisper.cpp/issues/66#issuecomment-1282546789
|
|
91
|
-
ifeq ($(UNAME_S),Darwin)
|
|
92
|
-
ifndef LLAMA_NO_METAL
|
|
93
|
-
LLAMA_METAL := 1
|
|
94
|
-
endif
|
|
95
|
-
|
|
96
|
-
LLAMA_NO_OPENMP := 1
|
|
97
|
-
|
|
98
|
-
ifneq ($(UNAME_P),arm)
|
|
99
|
-
SYSCTL_M := $(shell sysctl -n hw.optional.arm64 2>/dev/null)
|
|
100
|
-
ifeq ($(SYSCTL_M),1)
|
|
101
|
-
# UNAME_P := arm
|
|
102
|
-
# UNAME_M := arm64
|
|
103
|
-
warn := $(warning Your arch is announced as x86_64, but it seems to actually be ARM64. Not fixing that can lead to bad performance. For more info see: https://github.com/ggerganov/whisper.cpp/issues/66\#issuecomment-1282546789)
|
|
104
|
-
endif
|
|
105
|
-
endif
|
|
106
|
-
endif
|
|
107
|
-
|
|
108
|
-
ifdef LLAMA_RPC
|
|
109
|
-
BUILD_TARGETS += rpc-server
|
|
110
|
-
endif
|
|
111
|
-
|
|
112
|
-
default: $(BUILD_TARGETS)
|
|
113
|
-
|
|
114
|
-
test: $(TEST_TARGETS)
|
|
115
|
-
@failures=0; \
|
|
116
|
-
for test_target in $(TEST_TARGETS); do \
|
|
117
|
-
if [ "$$test_target" = "tests/test-tokenizer-0" ]; then \
|
|
118
|
-
./$$test_target $(CURDIR)/models/ggml-vocab-llama-spm.gguf; \
|
|
119
|
-
./$$test_target $(CURDIR)/models/ggml-vocab-llama-bpe.gguf; \
|
|
120
|
-
./$$test_target $(CURDIR)/models/ggml-vocab-phi-3.gguf; \
|
|
121
|
-
./$$test_target $(CURDIR)/models/ggml-vocab-falcon.gguf; \
|
|
122
|
-
./$$test_target $(CURDIR)/models/ggml-vocab-bert-bge.gguf; \
|
|
123
|
-
./$$test_target $(CURDIR)/models/ggml-vocab-starcoder.gguf; \
|
|
124
|
-
./$$test_target $(CURDIR)/models/ggml-vocab-gpt-2.gguf; \
|
|
125
|
-
./$$test_target $(CURDIR)/models/ggml-vocab-refact.gguf; \
|
|
126
|
-
elif [ "$$test_target" = "tests/test-tokenizer-1-spm" ]; then \
|
|
127
|
-
continue; \
|
|
128
|
-
elif [ "$$test_target" = "tests/test-tokenizer-1-bpe" ]; then \
|
|
129
|
-
continue; \
|
|
130
|
-
else \
|
|
131
|
-
echo "Running test $$test_target..."; \
|
|
132
|
-
./$$test_target; \
|
|
133
|
-
fi; \
|
|
134
|
-
if [ $$? -ne 0 ]; then \
|
|
135
|
-
printf 'Test %s FAILED!\n\n' $$test_target; \
|
|
136
|
-
failures=$$(( failures + 1 )); \
|
|
137
|
-
else \
|
|
138
|
-
printf 'Test %s passed.\n\n' $$test_target; \
|
|
139
|
-
fi; \
|
|
140
|
-
done; \
|
|
141
|
-
if [ $$failures -gt 0 ]; then \
|
|
142
|
-
printf '\n%s tests failed.\n' $$failures; \
|
|
143
|
-
exit 1; \
|
|
144
|
-
fi
|
|
145
|
-
@echo 'All tests passed.'
|
|
146
|
-
|
|
147
|
-
all: $(BUILD_TARGETS) $(TEST_TARGETS)
|
|
148
|
-
|
|
149
|
-
coverage: ## Run code coverage
|
|
150
|
-
gcov -pb tests/*.cpp
|
|
151
|
-
|
|
152
|
-
lcov-report: coverage ## Generate lcov report
|
|
153
|
-
mkdir -p lcov-report
|
|
154
|
-
lcov --capture --directory . --output-file lcov-report/coverage.info
|
|
155
|
-
genhtml lcov-report/coverage.info --output-directory lcov-report
|
|
156
|
-
|
|
157
|
-
gcovr-report: coverage ## Generate gcovr report
|
|
158
|
-
mkdir -p gcovr-report
|
|
159
|
-
gcovr --root . --html --html-details --output gcovr-report/coverage.html
|
|
160
|
-
|
|
161
|
-
ifdef RISCV_CROSS_COMPILE
|
|
162
|
-
CC := riscv64-unknown-linux-gnu-gcc
|
|
163
|
-
CXX := riscv64-unknown-linux-gnu-g++
|
|
164
|
-
endif
|
|
165
|
-
|
|
166
|
-
#
|
|
167
|
-
# Compile flags
|
|
168
|
-
#
|
|
169
|
-
|
|
170
|
-
# keep standard at C11 and C++11
|
|
171
|
-
MK_CPPFLAGS = -I. -Icommon
|
|
172
|
-
MK_CFLAGS = -std=c11 -fPIC
|
|
173
|
-
MK_CXXFLAGS = -std=c++11 -fPIC
|
|
174
|
-
MK_NVCCFLAGS = -std=c++11
|
|
175
|
-
|
|
176
|
-
# -Ofast tends to produce faster code, but may not be available for some compilers.
|
|
177
|
-
ifdef LLAMA_FAST
|
|
178
|
-
MK_CFLAGS += -Ofast
|
|
179
|
-
HOST_CXXFLAGS += -Ofast
|
|
180
|
-
ifndef LLAMA_DEBUG
|
|
181
|
-
MK_NVCCFLAGS += -O3
|
|
182
|
-
endif # LLAMA_DEBUG
|
|
183
|
-
else
|
|
184
|
-
MK_CFLAGS += -O3
|
|
185
|
-
MK_CXXFLAGS += -O3
|
|
186
|
-
ifndef LLAMA_DEBUG
|
|
187
|
-
MK_NVCCFLAGS += -O3
|
|
188
|
-
endif # LLAMA_DEBUG
|
|
189
|
-
endif # LLAMA_FAST
|
|
190
|
-
|
|
191
|
-
ifndef LLAMA_NO_CCACHE
|
|
192
|
-
CCACHE := $(shell which ccache)
|
|
193
|
-
ifdef CCACHE
|
|
194
|
-
export CCACHE_SLOPPINESS = time_macros
|
|
195
|
-
$(info I ccache found, compilation results will be cached. Disable with LLAMA_NO_CCACHE.)
|
|
196
|
-
CC := $(CCACHE) $(CC)
|
|
197
|
-
CXX := $(CCACHE) $(CXX)
|
|
198
|
-
else
|
|
199
|
-
$(info I ccache not found. Consider installing it for faster compilation.)
|
|
200
|
-
endif # CCACHE
|
|
201
|
-
endif # LLAMA_NO_CCACHE
|
|
202
|
-
|
|
203
|
-
# clock_gettime came in POSIX.1b (1993)
|
|
204
|
-
# CLOCK_MONOTONIC came in POSIX.1-2001 / SUSv3 as optional
|
|
205
|
-
# posix_memalign came in POSIX.1-2001 / SUSv3
|
|
206
|
-
# M_PI is an XSI extension since POSIX.1-2001 / SUSv3, came in XPG1 (1985)
|
|
207
|
-
MK_CPPFLAGS += -D_XOPEN_SOURCE=600
|
|
208
|
-
|
|
209
|
-
# Somehow in OpenBSD whenever POSIX conformance is specified
|
|
210
|
-
# some string functions rely on locale_t availability,
|
|
211
|
-
# which was introduced in POSIX.1-2008, forcing us to go higher
|
|
212
|
-
ifeq ($(UNAME_S),OpenBSD)
|
|
213
|
-
MK_CPPFLAGS += -U_XOPEN_SOURCE -D_XOPEN_SOURCE=700
|
|
214
|
-
endif
|
|
215
|
-
|
|
216
|
-
# Data types, macros and functions related to controlling CPU affinity and
|
|
217
|
-
# some memory allocation are available on Linux through GNU extensions in libc
|
|
218
|
-
ifeq ($(UNAME_S),Linux)
|
|
219
|
-
MK_CPPFLAGS += -D_GNU_SOURCE
|
|
220
|
-
endif
|
|
221
|
-
|
|
222
|
-
# RLIMIT_MEMLOCK came in BSD, is not specified in POSIX.1,
|
|
223
|
-
# and on macOS its availability depends on enabling Darwin extensions
|
|
224
|
-
# similarly on DragonFly, enabling BSD extensions is necessary
|
|
225
|
-
ifeq ($(UNAME_S),Darwin)
|
|
226
|
-
MK_CPPFLAGS += -D_DARWIN_C_SOURCE
|
|
227
|
-
endif
|
|
228
|
-
ifeq ($(UNAME_S),DragonFly)
|
|
229
|
-
MK_CPPFLAGS += -D__BSD_VISIBLE
|
|
230
|
-
endif
|
|
231
|
-
|
|
232
|
-
# alloca is a non-standard interface that is not visible on BSDs when
|
|
233
|
-
# POSIX conformance is specified, but not all of them provide a clean way
|
|
234
|
-
# to enable it in such cases
|
|
235
|
-
ifeq ($(UNAME_S),FreeBSD)
|
|
236
|
-
MK_CPPFLAGS += -D__BSD_VISIBLE
|
|
237
|
-
endif
|
|
238
|
-
ifeq ($(UNAME_S),NetBSD)
|
|
239
|
-
MK_CPPFLAGS += -D_NETBSD_SOURCE
|
|
240
|
-
endif
|
|
241
|
-
ifeq ($(UNAME_S),OpenBSD)
|
|
242
|
-
MK_CPPFLAGS += -D_BSD_SOURCE
|
|
243
|
-
endif
|
|
244
|
-
|
|
245
|
-
ifdef LLAMA_SCHED_MAX_COPIES
|
|
246
|
-
MK_CPPFLAGS += -DGGML_SCHED_MAX_COPIES=$(LLAMA_SCHED_MAX_COPIES)
|
|
247
|
-
endif
|
|
248
|
-
|
|
249
|
-
ifdef LLAMA_DEBUG
|
|
250
|
-
MK_CFLAGS += -O0 -g
|
|
251
|
-
MK_CXXFLAGS += -O0 -g
|
|
252
|
-
MK_LDFLAGS += -g
|
|
253
|
-
MK_NVCCFLAGS += -O0 -g
|
|
254
|
-
|
|
255
|
-
ifeq ($(UNAME_S),Linux)
|
|
256
|
-
MK_CPPFLAGS += -D_GLIBCXX_ASSERTIONS
|
|
257
|
-
endif
|
|
258
|
-
else
|
|
259
|
-
MK_CPPFLAGS += -DNDEBUG
|
|
260
|
-
endif
|
|
261
|
-
|
|
262
|
-
ifdef LLAMA_SANITIZE_THREAD
|
|
263
|
-
MK_CFLAGS += -fsanitize=thread -g
|
|
264
|
-
MK_CXXFLAGS += -fsanitize=thread -g
|
|
265
|
-
MK_LDFLAGS += -fsanitize=thread -g
|
|
266
|
-
endif
|
|
267
|
-
|
|
268
|
-
ifdef LLAMA_SANITIZE_ADDRESS
|
|
269
|
-
MK_CFLAGS += -fsanitize=address -fno-omit-frame-pointer -g
|
|
270
|
-
MK_CXXFLAGS += -fsanitize=address -fno-omit-frame-pointer -g
|
|
271
|
-
MK_LDFLAGS += -fsanitize=address -fno-omit-frame-pointer -g
|
|
272
|
-
endif
|
|
273
|
-
|
|
274
|
-
ifdef LLAMA_SANITIZE_UNDEFINED
|
|
275
|
-
MK_CFLAGS += -fsanitize=undefined -g
|
|
276
|
-
MK_CXXFLAGS += -fsanitize=undefined -g
|
|
277
|
-
MK_LDFLAGS += -fsanitize=undefined -g
|
|
278
|
-
endif
|
|
279
|
-
|
|
280
|
-
ifdef LLAMA_SERVER_VERBOSE
|
|
281
|
-
MK_CPPFLAGS += -DSERVER_VERBOSE=$(LLAMA_SERVER_VERBOSE)
|
|
282
|
-
endif
|
|
283
|
-
|
|
284
|
-
ifdef LLAMA_SERVER_SSL
|
|
285
|
-
MK_CPPFLAGS += -DCPPHTTPLIB_OPENSSL_SUPPORT
|
|
286
|
-
MK_LDFLAGS += -lssl -lcrypto
|
|
287
|
-
endif
|
|
288
|
-
|
|
289
|
-
ifdef LLAMA_CODE_COVERAGE
|
|
290
|
-
MK_CXXFLAGS += -fprofile-arcs -ftest-coverage -dumpbase ''
|
|
291
|
-
endif
|
|
292
|
-
|
|
293
|
-
ifdef LLAMA_DISABLE_LOGS
|
|
294
|
-
MK_CPPFLAGS += -DLOG_DISABLE_LOGS
|
|
295
|
-
endif # LLAMA_DISABLE_LOGS
|
|
296
|
-
|
|
297
|
-
# warnings
|
|
298
|
-
WARN_FLAGS = -Wall -Wextra -Wpedantic -Wcast-qual -Wno-unused-function
|
|
299
|
-
MK_CFLAGS += $(WARN_FLAGS) -Wshadow -Wstrict-prototypes -Wpointer-arith -Wmissing-prototypes -Werror=implicit-int \
|
|
300
|
-
-Werror=implicit-function-declaration
|
|
301
|
-
MK_CXXFLAGS += $(WARN_FLAGS) -Wmissing-declarations -Wmissing-noreturn
|
|
302
|
-
|
|
303
|
-
ifeq ($(LLAMA_FATAL_WARNINGS),1)
|
|
304
|
-
MK_CFLAGS += -Werror
|
|
305
|
-
MK_CXXFLAGS += -Werror
|
|
306
|
-
endif
|
|
307
|
-
|
|
308
|
-
# this version of Apple ld64 is buggy
|
|
309
|
-
ifneq '' '$(findstring dyld-1015.7,$(shell $(CC) $(LDFLAGS) -Wl,-v 2>&1))'
|
|
310
|
-
MK_CPPFLAGS += -DHAVE_BUGGY_APPLE_LINKER
|
|
311
|
-
endif
|
|
312
|
-
|
|
313
|
-
# OS specific
|
|
314
|
-
# TODO: support Windows
|
|
315
|
-
ifneq '' '$(filter $(UNAME_S),Linux Darwin FreeBSD NetBSD OpenBSD Haiku)'
|
|
316
|
-
MK_CFLAGS += -pthread
|
|
317
|
-
MK_CXXFLAGS += -pthread
|
|
318
|
-
endif
|
|
319
|
-
|
|
320
|
-
# detect Windows
|
|
321
|
-
ifneq ($(findstring _NT,$(UNAME_S)),)
|
|
322
|
-
_WIN32 := 1
|
|
323
|
-
endif
|
|
324
|
-
|
|
325
|
-
# library name prefix
|
|
326
|
-
ifneq ($(_WIN32),1)
|
|
327
|
-
LIB_PRE := lib
|
|
328
|
-
endif
|
|
329
|
-
|
|
330
|
-
# Dynamic Shared Object extension
|
|
331
|
-
ifeq ($(_WIN32),1)
|
|
332
|
-
DSO_EXT := .dll
|
|
333
|
-
else ifeq ($(UNAME_S),Darwin)
|
|
334
|
-
DSO_EXT := .dylib
|
|
335
|
-
else
|
|
336
|
-
DSO_EXT := .so
|
|
337
|
-
endif
|
|
338
|
-
|
|
339
|
-
# Windows Sockets 2 (Winsock) for network-capable apps
|
|
340
|
-
ifeq ($(_WIN32),1)
|
|
341
|
-
LWINSOCK2 := -lws2_32
|
|
342
|
-
endif
|
|
343
|
-
|
|
344
|
-
ifdef LLAMA_GPROF
|
|
345
|
-
MK_CFLAGS += -pg
|
|
346
|
-
MK_CXXFLAGS += -pg
|
|
347
|
-
endif
|
|
348
|
-
ifdef LLAMA_PERF
|
|
349
|
-
MK_CPPFLAGS += -DGGML_PERF
|
|
350
|
-
endif
|
|
351
|
-
|
|
352
|
-
# Architecture specific
|
|
353
|
-
# TODO: probably these flags need to be tweaked on some architectures
|
|
354
|
-
# feel free to update the Makefile for your architecture and send a pull request or issue
|
|
355
|
-
|
|
356
|
-
ifndef RISCV
|
|
357
|
-
|
|
358
|
-
ifeq ($(UNAME_M),$(filter $(UNAME_M),x86_64 i686 amd64))
|
|
359
|
-
# Use all CPU extensions that are available:
|
|
360
|
-
MK_CFLAGS += -march=native -mtune=native
|
|
361
|
-
HOST_CXXFLAGS += -march=native -mtune=native
|
|
362
|
-
|
|
363
|
-
# Usage AVX-only
|
|
364
|
-
#MK_CFLAGS += -mfma -mf16c -mavx
|
|
365
|
-
#MK_CXXFLAGS += -mfma -mf16c -mavx
|
|
366
|
-
|
|
367
|
-
# Usage SSSE3-only (Not is SSE3!)
|
|
368
|
-
#MK_CFLAGS += -mssse3
|
|
369
|
-
#MK_CXXFLAGS += -mssse3
|
|
370
|
-
endif
|
|
371
|
-
|
|
372
|
-
ifneq '' '$(findstring mingw,$(shell $(CC) -dumpmachine))'
|
|
373
|
-
# The stack is only 16-byte aligned on Windows, so don't let gcc emit aligned moves.
|
|
374
|
-
# https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54412
|
|
375
|
-
# https://github.com/ggerganov/llama.cpp/issues/2922
|
|
376
|
-
MK_CFLAGS += -Xassembler -muse-unaligned-vector-move
|
|
377
|
-
MK_CXXFLAGS += -Xassembler -muse-unaligned-vector-move
|
|
378
|
-
|
|
379
|
-
# Target Windows 8 for PrefetchVirtualMemory
|
|
380
|
-
MK_CPPFLAGS += -D_WIN32_WINNT=0x602
|
|
381
|
-
endif
|
|
382
|
-
|
|
383
|
-
ifneq ($(filter aarch64%,$(UNAME_M)),)
|
|
384
|
-
# Apple M1, M2, etc.
|
|
385
|
-
# Raspberry Pi 3, 4, Zero 2 (64-bit)
|
|
386
|
-
# Nvidia Jetson
|
|
387
|
-
MK_CFLAGS += -mcpu=native
|
|
388
|
-
MK_CXXFLAGS += -mcpu=native
|
|
389
|
-
JETSON_RELEASE_INFO = $(shell jetson_release)
|
|
390
|
-
ifdef JETSON_RELEASE_INFO
|
|
391
|
-
ifneq ($(filter TX2%,$(JETSON_RELEASE_INFO)),)
|
|
392
|
-
JETSON_EOL_MODULE_DETECT = 1
|
|
393
|
-
CC = aarch64-unknown-linux-gnu-gcc
|
|
394
|
-
cxx = aarch64-unknown-linux-gnu-g++
|
|
395
|
-
endif
|
|
396
|
-
endif
|
|
397
|
-
endif
|
|
398
|
-
|
|
399
|
-
ifneq ($(filter armv6%,$(UNAME_M)),)
|
|
400
|
-
# Raspberry Pi 1, Zero
|
|
401
|
-
MK_CFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access
|
|
402
|
-
MK_CXXFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access
|
|
403
|
-
endif
|
|
404
|
-
|
|
405
|
-
ifneq ($(filter armv7%,$(UNAME_M)),)
|
|
406
|
-
# Raspberry Pi 2
|
|
407
|
-
MK_CFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access -funsafe-math-optimizations
|
|
408
|
-
MK_CXXFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access -funsafe-math-optimizations
|
|
409
|
-
endif
|
|
410
|
-
|
|
411
|
-
ifneq ($(filter armv8%,$(UNAME_M)),)
|
|
412
|
-
# Raspberry Pi 3, 4, Zero 2 (32-bit)
|
|
413
|
-
MK_CFLAGS += -mfp16-format=ieee -mno-unaligned-access
|
|
414
|
-
MK_CXXFLAGS += -mfp16-format=ieee -mno-unaligned-access
|
|
415
|
-
endif
|
|
416
|
-
|
|
417
|
-
ifneq ($(filter ppc64%,$(UNAME_M)),)
|
|
418
|
-
POWER9_M := $(shell grep "POWER9" /proc/cpuinfo)
|
|
419
|
-
ifneq (,$(findstring POWER9,$(POWER9_M)))
|
|
420
|
-
MK_CFLAGS += -mcpu=power9
|
|
421
|
-
MK_CXXFLAGS += -mcpu=power9
|
|
422
|
-
endif
|
|
423
|
-
endif
|
|
424
|
-
|
|
425
|
-
ifneq ($(filter ppc64le%,$(UNAME_M)),)
|
|
426
|
-
MK_CFLAGS += -mcpu=powerpc64le
|
|
427
|
-
MK_CXXFLAGS += -mcpu=powerpc64le
|
|
428
|
-
CUDA_POWER_ARCH = 1
|
|
429
|
-
endif
|
|
430
|
-
|
|
431
|
-
ifneq ($(filter loongarch64%,$(UNAME_M)),)
|
|
432
|
-
MK_CFLAGS += -mlasx
|
|
433
|
-
MK_CXXFLAGS += -mlasx
|
|
434
|
-
endif
|
|
435
|
-
|
|
436
|
-
else
|
|
437
|
-
MK_CFLAGS += -march=rv64gcv -mabi=lp64d
|
|
438
|
-
MK_CXXFLAGS += -march=rv64gcv -mabi=lp64d
|
|
439
|
-
endif
|
|
440
|
-
|
|
441
|
-
ifndef LLAMA_NO_ACCELERATE
|
|
442
|
-
# Mac OS - include Accelerate framework.
|
|
443
|
-
# `-framework Accelerate` works both with Apple Silicon and Mac Intel
|
|
444
|
-
ifeq ($(UNAME_S),Darwin)
|
|
445
|
-
MK_CPPFLAGS += -DGGML_USE_ACCELERATE -DGGML_USE_BLAS
|
|
446
|
-
MK_CPPFLAGS += -DACCELERATE_NEW_LAPACK
|
|
447
|
-
MK_CPPFLAGS += -DACCELERATE_LAPACK_ILP64
|
|
448
|
-
MK_LDFLAGS += -framework Accelerate
|
|
449
|
-
OBJS += ggml-blas.o
|
|
450
|
-
endif
|
|
451
|
-
endif # LLAMA_NO_ACCELERATE
|
|
452
|
-
|
|
453
|
-
ifndef LLAMA_NO_OPENMP
|
|
454
|
-
MK_CPPFLAGS += -DGGML_USE_OPENMP
|
|
455
|
-
MK_CFLAGS += -fopenmp
|
|
456
|
-
MK_CXXFLAGS += -fopenmp
|
|
457
|
-
endif # LLAMA_NO_OPENMP
|
|
458
|
-
|
|
459
|
-
ifdef LLAMA_OPENBLAS
|
|
460
|
-
MK_CPPFLAGS += -DGGML_USE_BLAS $(shell pkg-config --cflags-only-I openblas)
|
|
461
|
-
MK_CFLAGS += $(shell pkg-config --cflags-only-other openblas)
|
|
462
|
-
MK_LDFLAGS += $(shell pkg-config --libs openblas)
|
|
463
|
-
OBJS += ggml-blas.o
|
|
464
|
-
endif # LLAMA_OPENBLAS
|
|
465
|
-
|
|
466
|
-
ifdef LLAMA_OPENBLAS64
|
|
467
|
-
MK_CPPFLAGS += -DGGML_USE_BLAS $(shell pkg-config --cflags-only-I openblas64)
|
|
468
|
-
MK_CFLAGS += $(shell pkg-config --cflags-only-other openblas64)
|
|
469
|
-
MK_LDFLAGS += $(shell pkg-config --libs openblas64)
|
|
470
|
-
OBJS += ggml-blas.o
|
|
471
|
-
endif # LLAMA_OPENBLAS64
|
|
472
|
-
|
|
473
|
-
ifdef LLAMA_BLIS
|
|
474
|
-
MK_CPPFLAGS += -DGGML_USE_BLAS -I/usr/local/include/blis -I/usr/include/blis
|
|
475
|
-
MK_LDFLAGS += -lblis -L/usr/local/lib
|
|
476
|
-
OBJS += ggml-blas.o
|
|
477
|
-
endif # LLAMA_BLIS
|
|
478
|
-
|
|
479
|
-
ifndef LLAMA_NO_LLAMAFILE
|
|
480
|
-
MK_CPPFLAGS += -DGGML_USE_LLAMAFILE
|
|
481
|
-
OBJS += sgemm.o
|
|
482
|
-
endif
|
|
483
|
-
|
|
484
|
-
ifdef LLAMA_RPC
|
|
485
|
-
MK_CPPFLAGS += -DGGML_USE_RPC
|
|
486
|
-
OBJS += ggml-rpc.o
|
|
487
|
-
endif # LLAMA_RPC
|
|
488
|
-
|
|
489
|
-
ifdef LLAMA_CUBLAS
|
|
490
|
-
# LLAMA_CUBLAS is deprecated and will be removed in the future
|
|
491
|
-
LLAMA_CUDA := 1
|
|
492
|
-
endif
|
|
493
|
-
|
|
494
|
-
OBJS_CUDA_TEMP_INST = $(patsubst %.cu,%.o,$(wildcard ggml-cuda/template-instances/fattn-wmma*.cu))
|
|
495
|
-
OBJS_CUDA_TEMP_INST += $(patsubst %.cu,%.o,$(wildcard ggml-cuda/template-instances/mmq*.cu))
|
|
496
|
-
ifdef LLAMA_CUDA_FA_ALL_QUANTS
|
|
497
|
-
OBJS_CUDA_TEMP_INST += $(patsubst %.cu,%.o,$(wildcard ggml-cuda/template-instances/fattn-vec*.cu))
|
|
498
|
-
else
|
|
499
|
-
OBJS_CUDA_TEMP_INST += $(patsubst %.cu,%.o,$(wildcard ggml-cuda/template-instances/fattn-vec*q4_0-q4_0.cu))
|
|
500
|
-
OBJS_CUDA_TEMP_INST += $(patsubst %.cu,%.o,$(wildcard ggml-cuda/template-instances/fattn-vec*q8_0-q8_0.cu))
|
|
501
|
-
OBJS_CUDA_TEMP_INST += $(patsubst %.cu,%.o,$(wildcard ggml-cuda/template-instances/fattn-vec*f16-f16.cu))
|
|
502
|
-
endif # LLAMA_CUDA_FA_ALL_QUANTS
|
|
503
|
-
|
|
504
|
-
ifdef LLAMA_CUDA
|
|
505
|
-
ifneq ('', '$(wildcard /opt/cuda)')
|
|
506
|
-
CUDA_PATH ?= /opt/cuda
|
|
507
|
-
else
|
|
508
|
-
CUDA_PATH ?= /usr/local/cuda
|
|
509
|
-
endif
|
|
510
|
-
MK_CPPFLAGS += -DGGML_USE_CUDA -I$(CUDA_PATH)/include -I$(CUDA_PATH)/targets/$(UNAME_M)-linux/include -DGGML_CUDA_USE_GRAPHS
|
|
511
|
-
MK_LDFLAGS += -lcuda -lcublas -lculibos -lcudart -lcublasLt -lpthread -ldl -lrt -L$(CUDA_PATH)/lib64 -L/usr/lib64 -L$(CUDA_PATH)/targets/$(UNAME_M)-linux/lib -L/usr/lib/wsl/lib
|
|
512
|
-
OBJS += ggml-cuda.o
|
|
513
|
-
OBJS += $(patsubst %.cu,%.o,$(wildcard ggml-cuda/*.cu))
|
|
514
|
-
OBJS += $(OBJS_CUDA_TEMP_INST)
|
|
515
|
-
MK_NVCCFLAGS += -use_fast_math
|
|
516
|
-
ifdef LLAMA_FATAL_WARNINGS
|
|
517
|
-
MK_NVCCFLAGS += -Werror all-warnings
|
|
518
|
-
endif # LLAMA_FATAL_WARNINGS
|
|
519
|
-
ifndef JETSON_EOL_MODULE_DETECT
|
|
520
|
-
MK_NVCCFLAGS += --forward-unknown-to-host-compiler
|
|
521
|
-
endif # JETSON_EOL_MODULE_DETECT
|
|
522
|
-
ifdef LLAMA_DEBUG
|
|
523
|
-
MK_NVCCFLAGS += -lineinfo
|
|
524
|
-
endif # LLAMA_DEBUG
|
|
525
|
-
ifdef LLAMA_CUDA_DEBUG
|
|
526
|
-
MK_NVCCFLAGS += --device-debug
|
|
527
|
-
endif # LLAMA_CUDA_DEBUG
|
|
528
|
-
ifdef LLAMA_CUDA_NVCC
|
|
529
|
-
NVCC = $(CCACHE) $(LLAMA_CUDA_NVCC)
|
|
530
|
-
else
|
|
531
|
-
NVCC = $(CCACHE) nvcc
|
|
532
|
-
endif #LLAMA_CUDA_NVCC
|
|
533
|
-
ifdef CUDA_DOCKER_ARCH
|
|
534
|
-
MK_NVCCFLAGS += -Wno-deprecated-gpu-targets -arch=$(CUDA_DOCKER_ARCH)
|
|
535
|
-
else ifndef CUDA_POWER_ARCH
|
|
536
|
-
MK_NVCCFLAGS += -arch=native
|
|
537
|
-
endif # CUDA_DOCKER_ARCH
|
|
538
|
-
ifdef LLAMA_CUDA_FORCE_DMMV
|
|
539
|
-
MK_NVCCFLAGS += -DGGML_CUDA_FORCE_DMMV
|
|
540
|
-
endif # LLAMA_CUDA_FORCE_DMMV
|
|
541
|
-
ifdef LLAMA_CUDA_FORCE_MMQ
|
|
542
|
-
MK_NVCCFLAGS += -DGGML_CUDA_FORCE_MMQ
|
|
543
|
-
endif # LLAMA_CUDA_FORCE_MMQ
|
|
544
|
-
ifdef LLAMA_CUDA_DMMV_X
|
|
545
|
-
MK_NVCCFLAGS += -DGGML_CUDA_DMMV_X=$(LLAMA_CUDA_DMMV_X)
|
|
546
|
-
else
|
|
547
|
-
MK_NVCCFLAGS += -DGGML_CUDA_DMMV_X=32
|
|
548
|
-
endif # LLAMA_CUDA_DMMV_X
|
|
549
|
-
ifdef LLAMA_CUDA_MMV_Y
|
|
550
|
-
MK_NVCCFLAGS += -DGGML_CUDA_MMV_Y=$(LLAMA_CUDA_MMV_Y)
|
|
551
|
-
else ifdef LLAMA_CUDA_DMMV_Y
|
|
552
|
-
MK_NVCCFLAGS += -DGGML_CUDA_MMV_Y=$(LLAMA_CUDA_DMMV_Y) # for backwards compatibility
|
|
553
|
-
else
|
|
554
|
-
MK_NVCCFLAGS += -DGGML_CUDA_MMV_Y=1
|
|
555
|
-
endif # LLAMA_CUDA_MMV_Y
|
|
556
|
-
ifdef LLAMA_CUDA_F16
|
|
557
|
-
MK_NVCCFLAGS += -DGGML_CUDA_F16
|
|
558
|
-
endif # LLAMA_CUDA_F16
|
|
559
|
-
ifdef LLAMA_CUDA_DMMV_F16
|
|
560
|
-
MK_NVCCFLAGS += -DGGML_CUDA_F16
|
|
561
|
-
endif # LLAMA_CUDA_DMMV_F16
|
|
562
|
-
ifdef LLAMA_CUDA_KQUANTS_ITER
|
|
563
|
-
MK_NVCCFLAGS += -DK_QUANTS_PER_ITERATION=$(LLAMA_CUDA_KQUANTS_ITER)
|
|
564
|
-
else
|
|
565
|
-
MK_NVCCFLAGS += -DK_QUANTS_PER_ITERATION=2
|
|
566
|
-
endif
|
|
567
|
-
ifdef LLAMA_CUDA_PEER_MAX_BATCH_SIZE
|
|
568
|
-
MK_NVCCFLAGS += -DGGML_CUDA_PEER_MAX_BATCH_SIZE=$(LLAMA_CUDA_PEER_MAX_BATCH_SIZE)
|
|
569
|
-
else
|
|
570
|
-
MK_NVCCFLAGS += -DGGML_CUDA_PEER_MAX_BATCH_SIZE=128
|
|
571
|
-
endif # LLAMA_CUDA_PEER_MAX_BATCH_SIZE
|
|
572
|
-
ifdef LLAMA_CUDA_NO_PEER_COPY
|
|
573
|
-
MK_NVCCFLAGS += -DGGML_CUDA_NO_PEER_COPY
|
|
574
|
-
endif # LLAMA_CUDA_NO_PEER_COPY
|
|
575
|
-
ifdef LLAMA_CUDA_CCBIN
|
|
576
|
-
MK_NVCCFLAGS += -ccbin $(LLAMA_CUDA_CCBIN)
|
|
577
|
-
endif # LLAMA_CUDA_CCBIN
|
|
578
|
-
ifdef LLAMA_CUDA_FA_ALL_QUANTS
|
|
579
|
-
MK_NVCCFLAGS += -DGGML_CUDA_FA_ALL_QUANTS
|
|
580
|
-
endif # LLAMA_CUDA_FA_ALL_QUANTS
|
|
581
|
-
|
|
582
|
-
ifdef JETSON_EOL_MODULE_DETECT
|
|
583
|
-
define NVCC_COMPILE
|
|
584
|
-
$(NVCC) -I. -Icommon -D_XOPEN_SOURCE=600 -D_GNU_SOURCE -DNDEBUG -DGGML_USE_CUDA -I/usr/local/cuda/include -I/opt/cuda/include -I/usr/local/cuda/targets/aarch64-linux/include -std=c++11 -O3 $(NVCCFLAGS) $(CPPFLAGS) -Xcompiler "$(CUDA_CXXFLAGS)" -c $< -o $@
|
|
585
|
-
endef # NVCC_COMPILE
|
|
586
|
-
else
|
|
587
|
-
define NVCC_COMPILE
|
|
588
|
-
$(NVCC) $(NVCCFLAGS) $(CPPFLAGS) -Xcompiler "$(CUDA_CXXFLAGS)" -c $< -o $@
|
|
589
|
-
endef # NVCC_COMPILE
|
|
590
|
-
endif # JETSON_EOL_MODULE_DETECT
|
|
591
|
-
|
|
592
|
-
ggml-cuda/%.o: ggml-cuda/%.cu ggml.h ggml-common.h ggml-cuda/common.cuh
|
|
593
|
-
$(NVCC_COMPILE)
|
|
594
|
-
|
|
595
|
-
ggml-cuda.o: ggml-cuda.cu ggml-cuda.h ggml.h ggml-backend.h ggml-backend-impl.h ggml-common.h $(wildcard ggml-cuda/*.cuh)
|
|
596
|
-
$(NVCC_COMPILE)
|
|
597
|
-
endif # LLAMA_CUDA
|
|
598
|
-
|
|
599
|
-
ifdef LLAMA_VULKAN
|
|
600
|
-
MK_CPPFLAGS += -DGGML_USE_VULKAN
|
|
601
|
-
MK_LDFLAGS += -lvulkan
|
|
602
|
-
OBJS += ggml-vulkan.o
|
|
603
|
-
|
|
604
|
-
ifdef LLAMA_VULKAN_CHECK_RESULTS
|
|
605
|
-
MK_CPPFLAGS += -DGGML_VULKAN_CHECK_RESULTS
|
|
606
|
-
endif
|
|
607
|
-
|
|
608
|
-
ifdef LLAMA_VULKAN_DEBUG
|
|
609
|
-
MK_CPPFLAGS += -DGGML_VULKAN_DEBUG
|
|
610
|
-
endif
|
|
611
|
-
|
|
612
|
-
ifdef LLAMA_VULKAN_VALIDATE
|
|
613
|
-
MK_CPPFLAGS += -DGGML_VULKAN_VALIDATE
|
|
614
|
-
endif
|
|
615
|
-
|
|
616
|
-
ifdef LLAMA_VULKAN_RUN_TESTS
|
|
617
|
-
MK_CPPFLAGS += -DGGML_VULKAN_RUN_TESTS
|
|
618
|
-
endif
|
|
619
|
-
|
|
620
|
-
ggml-vulkan.o: ggml-vulkan.cpp ggml-vulkan.h
|
|
621
|
-
$(CXX) $(CXXFLAGS) -c $< -o $@
|
|
622
|
-
endif # LLAMA_VULKAN
|
|
623
|
-
|
|
624
|
-
ifdef LLAMA_HIPBLAS
|
|
625
|
-
ifeq ($(wildcard /opt/rocm),)
|
|
626
|
-
ROCM_PATH ?= /usr
|
|
627
|
-
AMDGPU_TARGETS ?= $(shell $(shell which amdgpu-arch))
|
|
628
|
-
else
|
|
629
|
-
ROCM_PATH ?= /opt/rocm
|
|
630
|
-
AMDGPU_TARGETS ?= $(shell $(ROCM_PATH)/llvm/bin/amdgpu-arch)
|
|
631
|
-
endif
|
|
632
|
-
HIPCC ?= $(CCACHE) $(ROCM_PATH)/bin/hipcc
|
|
633
|
-
LLAMA_CUDA_DMMV_X ?= 32
|
|
634
|
-
LLAMA_CUDA_MMV_Y ?= 1
|
|
635
|
-
LLAMA_CUDA_KQUANTS_ITER ?= 2
|
|
636
|
-
MK_CPPFLAGS += -DGGML_USE_HIPBLAS -DGGML_USE_CUDA
|
|
637
|
-
ifdef LLAMA_HIP_UMA
|
|
638
|
-
MK_CPPFLAGS += -DGGML_HIP_UMA
|
|
639
|
-
endif # LLAMA_HIP_UMA
|
|
640
|
-
MK_LDFLAGS += -L$(ROCM_PATH)/lib -Wl,-rpath=$(ROCM_PATH)/lib
|
|
641
|
-
MK_LDFLAGS += -L$(ROCM_PATH)/lib64 -Wl,-rpath=$(ROCM_PATH)/lib64
|
|
642
|
-
MK_LDFLAGS += -lhipblas -lamdhip64 -lrocblas
|
|
643
|
-
HIPFLAGS += $(addprefix --offload-arch=,$(AMDGPU_TARGETS))
|
|
644
|
-
HIPFLAGS += -DGGML_CUDA_DMMV_X=$(LLAMA_CUDA_DMMV_X)
|
|
645
|
-
HIPFLAGS += -DGGML_CUDA_MMV_Y=$(LLAMA_CUDA_MMV_Y)
|
|
646
|
-
HIPFLAGS += -DK_QUANTS_PER_ITERATION=$(LLAMA_CUDA_KQUANTS_ITER)
|
|
647
|
-
ifdef LLAMA_CUDA_FORCE_DMMV
|
|
648
|
-
HIPFLAGS += -DGGML_CUDA_FORCE_DMMV
|
|
649
|
-
endif # LLAMA_CUDA_FORCE_DMMV
|
|
650
|
-
ifdef LLAMA_CUDA_NO_PEER_COPY
|
|
651
|
-
HIPFLAGS += -DGGML_CUDA_NO_PEER_COPY
|
|
652
|
-
endif # LLAMA_CUDA_NO_PEER_COPY
|
|
653
|
-
OBJS += ggml-cuda.o
|
|
654
|
-
OBJS += $(patsubst %.cu,%.o,$(wildcard ggml-cuda/*.cu))
|
|
655
|
-
OBJS += $(OBJS_CUDA_TEMP_INST)
|
|
656
|
-
|
|
657
|
-
ggml-cuda.o: ggml-cuda.cu ggml-cuda.h ggml.h ggml-backend.h ggml-backend-impl.h ggml-common.h $(wildcard ggml-cuda/*.cuh)
|
|
658
|
-
$(HIPCC) $(CXXFLAGS) $(HIPFLAGS) -x hip -c -o $@ $<
|
|
659
|
-
|
|
660
|
-
ggml-cuda/%.o: ggml-cuda/%.cu ggml.h ggml-common.h ggml-cuda/common.cuh
|
|
661
|
-
$(HIPCC) $(CXXFLAGS) $(HIPFLAGS) -x hip -c -o $@ $<
|
|
662
|
-
|
|
663
|
-
endif # LLAMA_HIPBLAS
|
|
664
|
-
|
|
665
|
-
ifdef LLAMA_METAL
|
|
666
|
-
MK_CPPFLAGS += -DGGML_USE_METAL
|
|
667
|
-
MK_LDFLAGS += -framework Foundation -framework Metal -framework MetalKit
|
|
668
|
-
OBJS += ggml-metal.o
|
|
669
|
-
ifdef LLAMA_METAL_NDEBUG
|
|
670
|
-
MK_CPPFLAGS += -DGGML_METAL_NDEBUG
|
|
671
|
-
endif
|
|
672
|
-
ifdef LLAMA_METAL_EMBED_LIBRARY
|
|
673
|
-
MK_CPPFLAGS += -DGGML_METAL_EMBED_LIBRARY
|
|
674
|
-
OBJS += ggml-metal-embed.o
|
|
675
|
-
endif
|
|
676
|
-
endif # LLAMA_METAL
|
|
677
|
-
|
|
678
|
-
ifdef LLAMA_METAL
|
|
679
|
-
ggml-metal.o: ggml-metal.m ggml-metal.h ggml.h
|
|
680
|
-
$(CC) $(CFLAGS) -c $< -o $@
|
|
681
|
-
|
|
682
|
-
ifdef LLAMA_METAL_EMBED_LIBRARY
|
|
683
|
-
ggml-metal-embed.o: ggml-metal.metal ggml-common.h
|
|
684
|
-
@echo "Embedding Metal library"
|
|
685
|
-
@sed -e '/#include "ggml-common.h"/r ggml-common.h' -e '/#include "ggml-common.h"/d' < ggml-metal.metal > ggml-metal-embed.metal
|
|
686
|
-
$(eval TEMP_ASSEMBLY=$(shell mktemp))
|
|
687
|
-
@echo ".section __DATA, __ggml_metallib" > $(TEMP_ASSEMBLY)
|
|
688
|
-
@echo ".globl _ggml_metallib_start" >> $(TEMP_ASSEMBLY)
|
|
689
|
-
@echo "_ggml_metallib_start:" >> $(TEMP_ASSEMBLY)
|
|
690
|
-
@echo ".incbin \"ggml-metal-embed.metal\"" >> $(TEMP_ASSEMBLY)
|
|
691
|
-
@echo ".globl _ggml_metallib_end" >> $(TEMP_ASSEMBLY)
|
|
692
|
-
@echo "_ggml_metallib_end:" >> $(TEMP_ASSEMBLY)
|
|
693
|
-
@$(AS) $(TEMP_ASSEMBLY) -o $@
|
|
694
|
-
@rm -f ${TEMP_ASSEMBLY}
|
|
695
|
-
endif
|
|
696
|
-
endif # LLAMA_METAL
|
|
697
|
-
|
|
698
|
-
OBJS += ggml-alloc.o ggml-backend.o ggml-quants.o unicode.o unicode-data.o
|
|
699
|
-
COMMON_H_DEPS = common/common.h common/sampling.h common/log.h llama.h
|
|
700
|
-
COMMON_DEPS = common.o sampling.o grammar-parser.o build-info.o json-schema-to-grammar.o
|
|
701
|
-
|
|
702
|
-
ifndef LLAMA_NO_LLAMAFILE
|
|
703
|
-
sgemm.o: sgemm.cpp sgemm.h ggml.h
|
|
704
|
-
$(CXX) $(CXXFLAGS) -c $< -o $@
|
|
705
|
-
endif
|
|
706
|
-
|
|
707
|
-
ifdef LLAMA_RPC
|
|
708
|
-
ggml-rpc.o: ggml-rpc.cpp ggml-rpc.h
|
|
709
|
-
$(CXX) $(CXXFLAGS) -c $< -o $@
|
|
710
|
-
|
|
711
|
-
rpc-server.o: examples/rpc/rpc-server.cpp ggml-rpc.h
|
|
712
|
-
$(CXX) $(CXXFLAGS) -c $< -o $@
|
|
713
|
-
|
|
714
|
-
rpc-server: rpc-server.o ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
715
|
-
$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
|
|
716
|
-
endif # LLAMA_RPC
|
|
717
|
-
|
|
718
|
-
GF_CC := $(CC)
|
|
719
|
-
include scripts/get-flags.mk
|
|
720
|
-
|
|
721
|
-
# combine build flags with cmdline overrides
|
|
722
|
-
override CPPFLAGS := $(MK_CPPFLAGS) $(CPPFLAGS)
|
|
723
|
-
override CFLAGS := $(CPPFLAGS) $(MK_CFLAGS) $(GF_CFLAGS) $(CFLAGS)
|
|
724
|
-
BASE_CXXFLAGS := $(MK_CXXFLAGS) $(CXXFLAGS)
|
|
725
|
-
override CXXFLAGS := $(BASE_CXXFLAGS) $(HOST_CXXFLAGS) $(GF_CXXFLAGS) $(CPPFLAGS)
|
|
726
|
-
override NVCCFLAGS := $(MK_NVCCFLAGS) $(NVCCFLAGS)
|
|
727
|
-
override LDFLAGS := $(MK_LDFLAGS) $(LDFLAGS)
|
|
728
|
-
|
|
729
|
-
# identify CUDA host compiler
|
|
730
|
-
ifdef LLAMA_CUDA
|
|
731
|
-
GF_CC := $(NVCC) $(NVCCFLAGS) 2>/dev/null .c -Xcompiler
|
|
732
|
-
include scripts/get-flags.mk
|
|
733
|
-
CUDA_CXXFLAGS := $(BASE_CXXFLAGS) $(GF_CXXFLAGS) -Wno-pedantic
|
|
734
|
-
endif
|
|
735
|
-
|
|
736
|
-
ifdef LLAMA_CURL
|
|
737
|
-
override CXXFLAGS := $(CXXFLAGS) -DLLAMA_USE_CURL
|
|
738
|
-
override LDFLAGS := $(LDFLAGS) -lcurl
|
|
739
|
-
endif
|
|
740
|
-
|
|
741
|
-
#
|
|
742
|
-
# Print build information
|
|
743
|
-
#
|
|
744
|
-
|
|
745
|
-
$(info I llama.cpp build info: )
|
|
746
|
-
$(info I UNAME_S: $(UNAME_S))
|
|
747
|
-
$(info I UNAME_P: $(UNAME_P))
|
|
748
|
-
$(info I UNAME_M: $(UNAME_M))
|
|
749
|
-
$(info I CFLAGS: $(CFLAGS))
|
|
750
|
-
$(info I CXXFLAGS: $(CXXFLAGS))
|
|
751
|
-
$(info I NVCCFLAGS: $(NVCCFLAGS))
|
|
752
|
-
$(info I LDFLAGS: $(LDFLAGS))
|
|
753
|
-
$(info I CC: $(shell $(CC) --version | head -n 1))
|
|
754
|
-
$(info I CXX: $(shell $(CXX) --version | head -n 1))
|
|
755
|
-
ifdef LLAMA_CUDA
|
|
756
|
-
$(info I NVCC: $(shell $(NVCC) --version | tail -n 1))
|
|
757
|
-
CUDA_VERSION := $(shell $(NVCC) --version | grep -oP 'release (\K[0-9]+\.[0-9])')
|
|
758
|
-
ifeq ($(shell awk -v "v=$(CUDA_VERSION)" 'BEGIN { print (v < 11.7) }'),1)
|
|
759
|
-
ifndef CUDA_DOCKER_ARCH
|
|
760
|
-
ifndef CUDA_POWER_ARCH
|
|
761
|
-
$(error I ERROR: For CUDA versions < 11.7 a target CUDA architecture must be explicitly provided via environment variable CUDA_DOCKER_ARCH, e.g. by running "export CUDA_DOCKER_ARCH=compute_XX" on Unix-like systems, where XX is the minimum compute capability that the code needs to run on. A list with compute capabilities can be found here: https://developer.nvidia.com/cuda-gpus )
|
|
762
|
-
endif # CUDA_POWER_ARCH
|
|
763
|
-
endif # CUDA_DOCKER_ARCH
|
|
764
|
-
endif # eq ($(shell echo "$(CUDA_VERSION) < 11.7" | bc),1)
|
|
765
|
-
endif # LLAMA_CUDA
|
|
766
|
-
$(info )
|
|
767
|
-
|
|
768
|
-
ifdef LLAMA_CUBLAS
|
|
769
|
-
$(info !!!!)
|
|
770
|
-
$(info LLAMA_CUBLAS is deprecated and will be removed in the future. Use LLAMA_CUDA instead.)
|
|
771
|
-
$(info !!!!)
|
|
772
|
-
$(info )
|
|
773
|
-
endif
|
|
774
|
-
|
|
775
|
-
#
|
|
776
|
-
# Build library
|
|
777
|
-
#
|
|
778
|
-
|
|
779
|
-
ggml.o: ggml.c ggml.h ggml-cuda.h
|
|
780
|
-
$(CC) $(CFLAGS) -c $< -o $@
|
|
781
|
-
|
|
782
|
-
ggml-alloc.o: ggml-alloc.c ggml.h ggml-alloc.h
|
|
783
|
-
$(CC) $(CFLAGS) -c $< -o $@
|
|
784
|
-
|
|
785
|
-
ggml-backend.o: ggml-backend.c ggml.h ggml-backend.h
|
|
786
|
-
$(CC) $(CFLAGS) -c $< -o $@
|
|
787
|
-
|
|
788
|
-
ggml-quants.o: ggml-quants.c ggml.h ggml-quants.h ggml-common.h
|
|
789
|
-
$(CC) $(CFLAGS) -c $< -o $@
|
|
790
|
-
|
|
791
|
-
ggml-blas.o: ggml-blas.cpp ggml-blas.h
|
|
792
|
-
$(CXX) $(CXXFLAGS) -c $< -o $@
|
|
793
|
-
|
|
794
|
-
unicode.o: unicode.cpp unicode.h
|
|
795
|
-
$(CXX) $(CXXFLAGS) -c $< -o $@
|
|
796
|
-
|
|
797
|
-
unicode-data.o: unicode-data.cpp unicode-data.h
|
|
798
|
-
$(CXX) $(CXXFLAGS) -c $< -o $@
|
|
799
|
-
|
|
800
|
-
llama.o: llama.cpp unicode.h ggml.h ggml-alloc.h ggml-backend.h ggml-cuda.h ggml-metal.h llama.h
|
|
801
|
-
$(CXX) $(CXXFLAGS) -c $< -o $@
|
|
802
|
-
|
|
803
|
-
common.o: common/common.cpp $(COMMON_H_DEPS)
|
|
804
|
-
$(CXX) $(CXXFLAGS) -c $< -o $@
|
|
805
|
-
|
|
806
|
-
sampling.o: common/sampling.cpp $(COMMON_H_DEPS)
|
|
807
|
-
$(CXX) $(CXXFLAGS) -c $< -o $@
|
|
808
|
-
|
|
809
|
-
console.o: common/console.cpp common/console.h
|
|
810
|
-
$(CXX) $(CXXFLAGS) -c $< -o $@
|
|
811
|
-
|
|
812
|
-
grammar-parser.o: common/grammar-parser.cpp common/grammar-parser.h
|
|
813
|
-
$(CXX) $(CXXFLAGS) -c $< -o $@
|
|
814
|
-
|
|
815
|
-
json-schema-to-grammar.o: common/json-schema-to-grammar.cpp common/json-schema-to-grammar.h
|
|
816
|
-
$(CXX) $(CXXFLAGS) -c $< -o $@
|
|
817
|
-
|
|
818
|
-
train.o: common/train.cpp common/train.h
|
|
819
|
-
$(CXX) $(CXXFLAGS) -c $< -o $@
|
|
820
|
-
|
|
821
|
-
ngram-cache.o: common/ngram-cache.cpp common/ngram-cache.h
|
|
822
|
-
$(CXX) $(CXXFLAGS) -c $< -o $@
|
|
823
|
-
|
|
824
|
-
libllama.so: llama.o ggml.o $(OBJS)
|
|
825
|
-
$(CXX) $(CXXFLAGS) -shared -fPIC -o $@ $^ $(LDFLAGS)
|
|
826
|
-
|
|
827
|
-
libllama.a: llama.o ggml.o $(OBJS) $(COMMON_DEPS)
|
|
828
|
-
ar rcs libllama.a llama.o ggml.o $(OBJS) $(COMMON_DEPS)
|
|
829
|
-
|
|
830
|
-
|
|
831
|
-
lib: llama.o ggml.o $(OBJS)
|
|
832
|
-
$(CXX) $(CXXFLAGS) -shared -fPIC -o libllama$(DSO_EXT) $^ $(LDFLAGS)
|
|
833
|
-
ar rcs libllama.a $^
|
|
834
|
-
|
|
835
|
-
clean:
|
|
836
|
-
rm -vrf *.o tests/*.o *.so *.a *.dll *.dylib common/build-info.cpp *.dot $(COV_TARGETS) $(BUILD_TARGETS) $(TEST_TARGETS)
|
|
837
|
-
rm -vrf ggml-cuda/*.o
|
|
838
|
-
rm -vrf ggml-cuda/template-instances/*.o
|
|
839
|
-
|
|
840
|
-
#
|
|
841
|
-
# Examples
|
|
842
|
-
#
|
|
843
|
-
|
|
844
|
-
# $< is the first prerequisite, i.e. the source file.
|
|
845
|
-
# Explicitly compile this to an object file so that it can be cached with ccache.
|
|
846
|
-
# The source file is then filtered out from $^ (the list of all prerequisites) and the object file is added instead.
|
|
847
|
-
|
|
848
|
-
# Helper function that replaces .c, .cpp, and .cu file endings with .o:
|
|
849
|
-
GET_OBJ_FILE = $(patsubst %.c,%.o,$(patsubst %.cpp,%.o,$(patsubst %.cu,%.o,$(1))))
|
|
850
|
-
|
|
851
|
-
llama-cli: examples/main/main.cpp ggml.o llama.o $(COMMON_DEPS) console.o grammar-parser.o $(OBJS)
|
|
852
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
853
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
854
|
-
@echo
|
|
855
|
-
@echo '==== Run ./llama-cli -h for help. ===='
|
|
856
|
-
@echo
|
|
857
|
-
|
|
858
|
-
llama-infill: examples/infill/infill.cpp ggml.o llama.o $(COMMON_DEPS) console.o grammar-parser.o $(OBJS)
|
|
859
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
860
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
861
|
-
|
|
862
|
-
llama-simple: examples/simple/simple.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
863
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
864
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
865
|
-
|
|
866
|
-
llama-tokenize: examples/tokenize/tokenize.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
867
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
868
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
869
|
-
|
|
870
|
-
llama-batched: examples/batched/batched.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
871
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
872
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
873
|
-
|
|
874
|
-
llama-batched-bench: examples/batched-bench/batched-bench.cpp build-info.o ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
875
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
876
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
877
|
-
|
|
878
|
-
llama-quantize: examples/quantize/quantize.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
879
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
880
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
881
|
-
|
|
882
|
-
llama-quantize-stats: examples/quantize-stats/quantize-stats.cpp build-info.o ggml.o llama.o $(OBJS)
|
|
883
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
884
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
885
|
-
|
|
886
|
-
llama-perplexity: examples/perplexity/perplexity.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
887
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
888
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
889
|
-
|
|
890
|
-
llama-imatrix: examples/imatrix/imatrix.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
891
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
892
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
893
|
-
|
|
894
|
-
llama-embedding: examples/embedding/embedding.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
895
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
896
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
897
|
-
|
|
898
|
-
llama-gritlm: examples/gritlm/gritlm.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
899
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
900
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
901
|
-
|
|
902
|
-
llama-save-load-state: examples/save-load-state/save-load-state.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
903
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
904
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
905
|
-
|
|
906
|
-
llama-server: examples/server/server.cpp examples/server/utils.hpp examples/server/httplib.h common/json.hpp examples/server/colorthemes.css.hpp examples/server/style.css.hpp examples/server/theme-beeninorder.css.hpp examples/server/theme-ketivah.css.hpp examples/server/theme-mangotango.css.hpp examples/server/theme-playground.css.hpp examples/server/theme-polarnight.css.hpp examples/server/theme-snowstorm.css.hpp examples/server/index.html.hpp examples/server/index-new.html.hpp examples/server/index.js.hpp examples/server/completion.js.hpp examples/server/system-prompts.js.hpp examples/server/prompt-formats.js.hpp examples/server/json-schema-to-grammar.mjs.hpp common/stb_image.h ggml.o llama.o $(COMMON_DEPS) grammar-parser.o $(OBJS)
|
|
907
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
908
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h %.hpp $<,$^) -Iexamples/server $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS) $(LWINSOCK2)
|
|
909
|
-
|
|
910
|
-
# Portable equivalent of `cd examples/server/public && xxd -i $(notdir $<) ../$(notdir $<).hpp`:
|
|
911
|
-
examples/server/%.hpp: examples/server/public/% Makefile
|
|
912
|
-
@( export NAME=$(subst .,_,$(subst -,_,$(notdir $<))) && \
|
|
913
|
-
echo "unsigned char $${NAME}[] = {" && \
|
|
914
|
-
cat $< | od -v -t x1 -An | sed -E 's/([0-9a-fA-F]+)/0x\1, /g' && \
|
|
915
|
-
echo "};" && \
|
|
916
|
-
echo "unsigned int $${NAME}_len = $(shell cat $< | wc -c );" \
|
|
917
|
-
) > $@
|
|
918
|
-
|
|
919
|
-
llama-gguf: examples/gguf/gguf.cpp ggml.o $(OBJS)
|
|
920
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
921
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
922
|
-
|
|
923
|
-
llama-gguf-split: examples/gguf-split/gguf-split.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
924
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
925
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
926
|
-
|
|
927
|
-
llama-eval-callback: examples/eval-callback/eval-callback.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
928
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
929
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
930
|
-
|
|
931
|
-
llama-train-text-from-scratch: examples/train-text-from-scratch/train-text-from-scratch.cpp ggml.o llama.o $(COMMON_DEPS) train.o $(OBJS)
|
|
932
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
933
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
934
|
-
|
|
935
|
-
llama-convert-llama2c-to-ggml: examples/convert-llama2c-to-ggml/convert-llama2c-to-ggml.cpp ggml.o llama.o $(OBJS)
|
|
936
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
937
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
938
|
-
|
|
939
|
-
llama-bench: examples/llama-bench/llama-bench.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
940
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
941
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
942
|
-
|
|
943
|
-
libllava.a: examples/llava/llava.cpp examples/llava/llava.h examples/llava/clip.cpp examples/llava/clip.h common/stb_image.h common/base64.hpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
944
|
-
$(CXX) $(CXXFLAGS) -static -fPIC -c $< -o $@ -Wno-cast-qual
|
|
945
|
-
|
|
946
|
-
llama-llava-cli: examples/llava/llava-cli.cpp examples/llava/clip.h examples/llava/clip.cpp examples/llava/llava.h examples/llava/llava.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
947
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
948
|
-
$(CXX) $(CXXFLAGS) -c examples/llava/clip.cpp -o $(call GET_OBJ_FILE, examples/llava/clip.cpp) -Wno-cast-qual
|
|
949
|
-
$(CXX) $(CXXFLAGS) -c examples/llava/llava.cpp -o $(call GET_OBJ_FILE, examples/llava/llava.cpp)
|
|
950
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $< examples/llava/clip.cpp examples/llava/llava.cpp,$^) $(call GET_OBJ_FILE, $<) $(call GET_OBJ_FILE, examples/llava/clip.cpp) $(call GET_OBJ_FILE, examples/llava/llava.cpp) -o $@ $(LDFLAGS)
|
|
951
|
-
|
|
952
|
-
llama-baby-llama: examples/baby-llama/baby-llama.cpp ggml.o llama.o $(COMMON_DEPS) train.o $(OBJS)
|
|
953
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
954
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
955
|
-
|
|
956
|
-
llama-finetune: examples/finetune/finetune.cpp ggml.o llama.o $(COMMON_DEPS) train.o $(OBJS)
|
|
957
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
958
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
959
|
-
|
|
960
|
-
llama-export-lora: examples/export-lora/export-lora.cpp ggml.o common/common.h $(OBJS)
|
|
961
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
962
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
963
|
-
|
|
964
|
-
llama-retrieval: examples/retrieval/retrieval.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
965
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
966
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
967
|
-
|
|
968
|
-
llama-speculative: examples/speculative/speculative.cpp ggml.o llama.o $(COMMON_DEPS) grammar-parser.o $(OBJS)
|
|
969
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
970
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
971
|
-
|
|
972
|
-
llama-parallel: examples/parallel/parallel.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
973
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
974
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
975
|
-
|
|
976
|
-
llama-lookahead: examples/lookahead/lookahead.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
977
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
978
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
979
|
-
|
|
980
|
-
llama-lookup: examples/lookup/lookup.cpp ggml.o llama.o ngram-cache.o $(COMMON_DEPS) $(OBJS)
|
|
981
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
982
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
983
|
-
|
|
984
|
-
llama-lookup-create: examples/lookup/lookup-create.cpp ggml.o llama.o ngram-cache.o $(COMMON_DEPS) $(OBJS)
|
|
985
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
986
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
987
|
-
|
|
988
|
-
llama-lookup-merge: examples/lookup/lookup-merge.cpp ggml.o llama.o ngram-cache.o $(COMMON_DEPS) $(OBJS)
|
|
989
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
990
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
991
|
-
|
|
992
|
-
llama-lookup-stats: examples/lookup/lookup-stats.cpp ggml.o llama.o ngram-cache.o $(COMMON_DEPS) $(OBJS)
|
|
993
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
994
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
995
|
-
|
|
996
|
-
llama-passkey: examples/passkey/passkey.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
997
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
998
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
999
|
-
|
|
1000
|
-
llama-gbnf-validator: examples/gbnf-validator/gbnf-validator.cpp ggml.o llama.o $(COMMON_DEPS) grammar-parser.o $(OBJS)
|
|
1001
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1002
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1003
|
-
|
|
1004
|
-
ifeq ($(UNAME_S),Darwin)
|
|
1005
|
-
swift: examples/batched.swift
|
|
1006
|
-
(cd examples/batched.swift; make build)
|
|
1007
|
-
endif
|
|
1008
|
-
|
|
1009
|
-
common/build-info.cpp: $(wildcard .git/index) scripts/build-info.sh
|
|
1010
|
-
@sh scripts/build-info.sh "$(CC)" > $@.tmp
|
|
1011
|
-
@if ! cmp -s $@.tmp $@; then \
|
|
1012
|
-
mv $@.tmp $@; \
|
|
1013
|
-
else \
|
|
1014
|
-
rm $@.tmp; \
|
|
1015
|
-
fi
|
|
1016
|
-
|
|
1017
|
-
build-info.o: common/build-info.cpp
|
|
1018
|
-
$(CXX) $(CXXFLAGS) -c $(filter-out %.h,$^) -o $@
|
|
1019
|
-
|
|
1020
|
-
#
|
|
1021
|
-
# Tests
|
|
1022
|
-
#
|
|
1023
|
-
|
|
1024
|
-
tests: $(TEST_TARGETS)
|
|
1025
|
-
|
|
1026
|
-
llama-benchmark-matmult: examples/benchmark/benchmark-matmult.cpp build-info.o ggml.o $(OBJS)
|
|
1027
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1028
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1029
|
-
|
|
1030
|
-
run-benchmark-matmult: llama-benchmark-matmult
|
|
1031
|
-
./$@
|
|
1032
|
-
|
|
1033
|
-
.PHONY: run-benchmark-matmult swift
|
|
1034
|
-
|
|
1035
|
-
llama-vdot: pocs/vdot/vdot.cpp ggml.o $(OBJS)
|
|
1036
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1037
|
-
$(CXX) $(CXXFLAGS) $(filter-out $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1038
|
-
|
|
1039
|
-
llama-q8dot: pocs/vdot/q8dot.cpp ggml.o $(OBJS)
|
|
1040
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1041
|
-
$(CXX) $(CXXFLAGS) $(filter-out $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1042
|
-
|
|
1043
|
-
tests/test-llama-grammar: tests/test-llama-grammar.cpp ggml.o grammar-parser.o $(OBJS)
|
|
1044
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1045
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1046
|
-
|
|
1047
|
-
tests/test-grammar-parser: tests/test-grammar-parser.cpp ggml.o llama.o grammar-parser.o $(OBJS)
|
|
1048
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1049
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1050
|
-
|
|
1051
|
-
tests/test-grammar-integration: tests/test-grammar-integration.cpp ggml.o llama.o grammar-parser.o $(OBJS)
|
|
1052
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1053
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1054
|
-
|
|
1055
|
-
tests/test-double-float: tests/test-double-float.cpp ggml.o $(OBJS)
|
|
1056
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1057
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1058
|
-
|
|
1059
|
-
tests/test-json-schema-to-grammar: tests/test-json-schema-to-grammar.cpp json-schema-to-grammar.o ggml.o llama.o grammar-parser.o $(OBJS)
|
|
1060
|
-
$(CXX) $(CXXFLAGS) -Iexamples/server -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1061
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1062
|
-
|
|
1063
|
-
tests/test-grad0: tests/test-grad0.cpp ggml.o $(OBJS)
|
|
1064
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1065
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1066
|
-
|
|
1067
|
-
tests/test-opt: tests/test-opt.cpp ggml.o $(OBJS)
|
|
1068
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1069
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1070
|
-
|
|
1071
|
-
tests/test-quantize-fns: tests/test-quantize-fns.cpp ggml.o $(OBJS)
|
|
1072
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1073
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1074
|
-
|
|
1075
|
-
tests/test-quantize-perf: tests/test-quantize-perf.cpp ggml.o $(OBJS)
|
|
1076
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1077
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1078
|
-
|
|
1079
|
-
tests/test-sampling: tests/test-sampling.cpp ggml.o llama.o $(OBJS)
|
|
1080
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1081
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1082
|
-
|
|
1083
|
-
tests/test-tokenizer-0: tests/test-tokenizer-0.cpp ggml.o llama.o $(COMMON_DEPS) console.o $(OBJS)
|
|
1084
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1085
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1086
|
-
|
|
1087
|
-
tests/test-tokenizer-1-bpe: tests/test-tokenizer-1-bpe.cpp ggml.o llama.o $(COMMON_DEPS) console.o $(OBJS)
|
|
1088
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1089
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1090
|
-
|
|
1091
|
-
tests/test-tokenizer-1-spm: tests/test-tokenizer-1-spm.cpp ggml.o llama.o $(COMMON_DEPS) console.o $(OBJS)
|
|
1092
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1093
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1094
|
-
|
|
1095
|
-
tests/test-rope: tests/test-rope.cpp ggml.o $(OBJS)
|
|
1096
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1097
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1098
|
-
|
|
1099
|
-
tests/test-c.o: tests/test-c.c llama.h
|
|
1100
|
-
$(CC) $(CFLAGS) -c $(filter-out %.h,$^) -o $@
|
|
1101
|
-
|
|
1102
|
-
tests/test-backend-ops: tests/test-backend-ops.cpp ggml.o $(OBJS)
|
|
1103
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1104
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1105
|
-
|
|
1106
|
-
tests/test-model-load-cancel: tests/test-model-load-cancel.cpp ggml.o llama.o tests/get-model.cpp $(COMMON_DEPS) $(OBJS)
|
|
1107
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1108
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1109
|
-
|
|
1110
|
-
tests/test-autorelease: tests/test-autorelease.cpp ggml.o llama.o tests/get-model.cpp $(COMMON_DEPS) $(OBJS)
|
|
1111
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1112
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|
|
1113
|
-
|
|
1114
|
-
tests/test-chat-template: tests/test-chat-template.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
|
|
1115
|
-
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
|
|
1116
|
-
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
|