00da2a5f4e
neox is updated
2023-05-17 14:56:54 +08:00
90fe9096b4
clean and refactoring pass before supporting newer models for different arch
2023-05-17 11:23:29 +08:00
60ee00428b
updated lite
2023-05-17 10:26:36 +08:00
d8d39f1ba8
Merge branch 'master' into concedo_experimental
...
# Conflicts:
# Makefile
2023-05-17 10:07:43 +08:00
f561fe5a4a
switch back to ofast for c
2023-05-17 10:04:54 +08:00
504a2aa874
Merge remote-tracking branch 'fixmake/concedo' into concedo_experimental
2023-05-17 10:01:57 +08:00
327763c21b
Merge remote-tracking branch 'occam/opencl-dev' into concedo_experimental
2023-05-17 10:01:22 +08:00
2b2646931b
convert.py: Support models which are stored in a single pytorch_model.bin ( #1469 )
...
* Support models in a single pytorch_model.bin
* Remove spurious line with typo
2023-05-17 00:04:35 +02:00
42627421ec
~7% faster Q5_1 AVX2 code ( #1477 )
2023-05-16 18:36:47 +00:00
de10afa80f
Fix tensor load to device
...
Co-authored-by: Johannes Gäßler <johannesg@5d6.de >
2023-05-16 18:49:49 +02:00
f29c25e7a1
hacky fix for linux cublas build
2023-05-16 12:29:04 -04:00
417711be46
add more QOL
2023-05-17 00:11:28 +08:00
9560655409
define default model path once, sync path with readme ( #1366 )
2023-05-16 17:46:34 +02:00
94ef3e81cf
inc allocation
2023-05-16 23:32:35 +08:00
954b87eb05
working checkpoint
2023-05-16 22:33:21 +08:00
84c1bc7822
Merge remote-tracking branch 'occam/opencl-dev' into concedo_experimental
2023-05-16 22:00:15 +08:00
b3ff66d87f
Fix error in convert f16 to f32 kernel call
2023-05-16 13:05:33 +02:00
8e394f9913
progress
2023-05-16 17:29:55 +08:00
196fbba527
Merge branch 'opencl-dev2' into concedo_experimental
...
# Conflicts:
# CMakeLists.txt
2023-05-16 17:04:33 +08:00
2a5ee023ad
Add alternate include path for openblas ( #1476 )
...
In some linux distributions (fedora, for example), the include path for openblas is located at '/usr/local/include'
2023-05-16 10:30:15 +02:00
554340f565
revert library back first
2023-05-16 15:45:05 +08:00
342d346c13
Generate dequant_mul_mat kernels from simple templates
2023-05-16 07:42:01 +02:00
e4e6994353
Not working, don't use. testing a merge
2023-05-16 12:33:24 +08:00
1747c598fa
Fix CMakeLists.txt
2023-05-15 19:51:23 +02:00
d43b243b9a
off static
2023-05-15 22:17:04 +08:00
96c28dda4d
export symbols
2023-05-15 20:38:21 +08:00
72836d4eac
fixing more compile issues
2023-05-15 20:10:54 +08:00
6504150fac
just testing cublas
2023-05-15 20:01:22 +08:00
fce2e7e518
up version
2023-05-15 14:53:13 +08:00
466cd21368
test cmakefile for cublas.
2023-05-15 14:50:38 +08:00
923184f2e8
Merge branch 'master' into concedo_experimental
...
# Conflicts:
# ggml.h
2023-05-15 10:55:15 +08:00
63d20469b8
fix get_num_physical_cores() ( #1436 )
...
* fix get_num_physical_cores()
had been broken on complex topologies because "cpu cores" in /proc/cpuinfo is per-"physical id"
* Add spaces to maintain consistent formatting
---------
Co-authored-by: slaren <ddevesa@gmail.com >
2023-05-15 04:25:42 +02:00
b5c9295eef
benchmark-matmul: fix clang-tidy issues, report results in GFLOPS ( #1458 )
...
* benchmark-matmul: fix command line parsing, replace macros with functions, report results in GFLOPS
2023-05-14 22:46:00 +02:00
5a74dc1536
Add remaining dequant_mul_mat functions
2023-05-14 22:19:54 +02:00
883e587a04
Fix dequant_mul_mat kernel
2023-05-14 21:26:28 +02:00
8795403de3
Fix bugs in dequant_mul_mat code
2023-05-14 21:14:05 +02:00
eb363627fd
cuda : deduplicated dequantization code ( #1453 )
2023-05-14 21:53:23 +03:00
79b2d5b69d
ggml : alternative fix for race condition bug in non-inplace ggml_compute_forward_diag_mask_f32 ( #1454 )
...
* fix race condition bug in non-inplace ggml_compute_forward_diag_mask_f32
memcpy needs to be synchronized across threads to avoid race conditions.
=> do it in INIT phase
* remove trailing whitespace
* Update ggml.c
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com >
2023-05-14 18:55:02 +03:00
13c351ad72
ggml : various fixes ( #1450 )
...
- `ggml_rope()`
- `ggml_diag_mask_inf()` multi-threaded
- compatibility with scratch buffers
2023-05-14 18:22:50 +03:00
c77966524a
Refactor OpenCL code to work more like the CUDA code, add missing functions
2023-05-14 17:01:46 +02:00
82bc517b9a
Move back to C++ for OpenCL
2023-05-14 17:00:37 +02:00
c81dd58e76
Merge commit 'f954edda935a70a14cf0cc45ecc7fe7d60cf3e4b' into archive_lib
...
# Conflicts:
# ggml.c
2023-05-14 18:34:56 +08:00
60f8c361ca
ggml : add AVX support based on AVX2 code ( #1430 )
2023-05-14 10:03:51 +00:00
b692e4d2a4
wip
2023-05-14 17:21:07 +08:00
601a033475
ggml : add GGML_QNT_VERSION to track quantization format changes
...
https://github.com/ggerganov/ggml/issues/150#issuecomment-1546625668
2023-05-14 10:20:19 +03:00
e01e373e63
Merge branch 'master' into concedo_experimental
...
# Conflicts:
# Makefile
# ggml.c
# llama.cpp
2023-05-14 11:34:41 +08:00
9cd5b9a769
up ver
2023-05-14 11:10:26 +08:00
8a5fe628df
recognize q8_0 as an older format as the new clblast doesnt work correctly with it
2023-05-14 11:06:23 +08:00
49d6334dc1
try fix kernel
2023-05-14 00:41:26 +08:00
e05455f852
fixed wrong sized struct from legacy q8_1, fixed opencl varsize arrays
2023-05-13 23:56:08 +08:00
Concedo
Tom Jobbins
Ilya Kurdyukov
0cc4m
horenbergerb
András Salamon
sandyiscool
zrm
slaren
Johannes Gäßler
xaedes
Georgi Gerganov
katsu560