 Srihari-mcw
|
3d82dbcbce
ggml : block interleaving support for Q4_K quantization for x86 AVX2 architecture (#12332)
|
10 bulan lalu |
 Bartowski
|
732b5fbf5e
convert : avoid calls to tokenizer.added_tokens_decoder (#12473)
|
10 bulan lalu |
 fairydreaming
|
568013d0cd
context : clear sets containing encoder output sequence ids before storing new values (#12470)
|
10 bulan lalu |
 Gaurav Garg
|
517b5ddbf0
CUDA: Improve flash decoding kernel GPU occupancy for BS=1 case (#12183)
|
10 bulan lalu |
 Jeff Bolz
|
a9b59288e2
vulkan: optimize iq1 coopmat2 dequant functions (#12427)
|
10 bulan lalu |
 Guus Waals
|
0fd8487b14
Fix visionOS build and add CI (#12415)
|
10 bulan lalu |
 Sigbjørn Skjæret
|
108e53c2f1
llama : add support for GPT2, Bloom and CodeShell tied word embeddings (#12456)
|
10 bulan lalu |
|
Sigbjørn Skjæret
|
a686171ea7
convert : Support chat_template.json (#12460)
|
10 bulan lalu |
|
Jeff Bolz
|
c446b2edd2
vulkan: Submit once enough matmul work has been recorded (#12406)
|
10 bulan lalu |
 lhez
|
d84635b1b0
opencl: improve profiling (#12442)
|
10 bulan lalu |
 Georgi Gerganov
|
75422e8bc4
graph : normalize Q, K, V shapes + sync cross attention (#12449)
|
10 bulan lalu |
 R0CKSTAR
|
bb115d2bf7
musa: override warp_size of musa device to 32 (#12445)
|
10 bulan lalu |
 Xuan-Son Nguyen
|
29fff308c7
llama : support converting Mistral Small text-only (#12450)
|
10 bulan lalu |
|
Georgi Gerganov
|
c6af2161b2
speculative : fix seg fault in certain cases (#12454)
|
10 bulan lalu |
|
Xuan-Son Nguyen
|
99aa304fb9
llama : add support for EXAONE tied word embeddings (#12451)
|
10 bulan lalu |
|
Georgi Gerganov
|
8551c44d84
context : always use non-causal attention for encoder graphs (#12447)
|
10 bulan lalu |
 Łukasz Ślusarczyk
|
35cae5ba05
SYCL: using graphs is configurable by environment variable and compile option (#12371)
|
10 bulan lalu |
|
Georgi Gerganov
|
810e0af3f5
server : fix warmup draft cache type (#12446)
|
10 bulan lalu |
 Prajwal B Mehendarkar
|
eba92d64c3
cmake : fix PowerPC build (#12241)
|
10 bulan lalu |
 fj-y-saito
|
d9a14523bb
ggml : add SVE support for q6_K_q8_K (#12361)
|
10 bulan lalu |
 0cc4m
|
fd123cfead
Vulkan: Default to 1GB allocations instead of 4GB to avoid fragmentation and driver issues (#12434)
|
10 bulan lalu |
|
Łukasz Ślusarczyk
|
a53f7f7b88
fixed compilation warnings in ggml-sycl (#12424)
|
10 bulan lalu |
 Molly Sophia
|
7dfad387e3
llama: Add support for RWKV v7 architecture (#12412)
|
10 bulan lalu |
|
Sigbjørn Skjæret
|
60c902926c
docs : bring llama-cli conversation/template docs up-to-date (#12426)
|
10 bulan lalu |
|
Gaurav Garg
|
b1b132efcb
cuda : enable CUDA Graph on CUDA Toolkit < 12.x (#12394)
|
10 bulan lalu |
|
Guus Waals
|
01e8f2138b
ggml-vulkan: remove unused find_program(glslc) (#12416)
|
10 bulan lalu |
|
Jeff Bolz
|
484a8ab513
vulkan: Add N/2 and N/4 optimized paths in coopmat2 shader (#12312)
|
10 bulan lalu |
 Daniele
|
cf2270e4d3
vulkan: subgroup size tuning (#12087)
|
10 bulan lalu |
|
Jeff Bolz
|
f07690c930
vulkan: use fp32 in coopmat2 q4_k dequant function (#12309)
|
10 bulan lalu |
|
Jeff Bolz
|
891c63956d
vulkan: Pad N dimension of B matrix for coopmat2 perf, to avoid bounds checking (#12273)
|
10 bulan lalu |
