COMMITS
/ tests/test-backend-ops.cpp December 4, 2024
J
vulkan: Implement "fast divide" (mul+shift) for unary ops like copy (#10642)
Jeff Bolz committed
December 3, 2024
P
ggml : add `GGML_PAD_REFLECT_1D` operation (ggml/1034)
PAB committed
G
metal : small-batch mat-mul kernels (#10581)
Georgi Gerganov committed
December 2, 2024
P
November 29, 2024
G
ggml : fix I8MM Q4_1 scaling factor conversion (#10562)
Georgi Gerganov committed
November 26, 2024
J
vulkan: fix group_norm (#10496)
Jeff Bolz committed
November 25, 2024
D
ggml : add support for dynamic loading of backends (#10469)
Diego Devesa committed
November 21, 2024
D
cuda : optimize argmax (#10441)
Diego Devesa committed
November 20, 2024
J
ggml-opt: fix data corruption (ggml/1022)
Johannes Gäßler committed
November 19, 2024
J
vulkan: Optimize soft_max (#10301)
Jeff Bolz committed
November 16, 2024
J
ggml: new optimization interface (ggml/988)
Johannes Gäßler committed
November 13, 2024
J
vulkan: Optimize contiguous copies (#10254)
Jeff Bolz committed
November 8, 2024
G
metal : optimize FA kernels (#10171)
Georgi Gerganov committed
November 7, 2024
Z
November 6, 2024
G
metal : add BF16 support (#8439)
Georgi Gerganov committed
November 3, 2024
D
ggml : move CPU backend to a separate file (#10144)
Diego Devesa committed
October 24, 2024
J
CUDA: fix MMQ for non-contiguous src0, add tests (#10021)
Johannes Gäßler committed
October 18, 2024
J
CUDA: fix 1D im2col, add tests (ggml/993)
Johannes Gäßler committed
October 23, 2024
J
metal : add POOL2D and fix IM2COL (#9943)
Jun Hee Yoo committed
October 8, 2024
D
ggml : fix BLAS with unsupported types (#9775)
Diego Devesa committed
October 7, 2024
D
ggml : add backend registry / device interfaces to BLAS backend (#9752)
Diego Devesa committed
October 3, 2024
J
ggml/ex: calculate accuracy in graph, adapt MNIST (ggml/980)
Johannes Gäßler committed
October 2, 2024
D
ggml-backend : add device and backend reg interfaces (#9707)
Diego Devesa committed
September 30, 2024
J
test: fix OPT_STEP_ADAMW for test-backend-ops (ggml/974)
Johannes Gäßler committed
September 29, 2024
J
ggml: fix gradient allocation logic (ggml/966)
Johannes Gäßler committed
September 28, 2024
S
test-backend-ops : use flops for some performance tests (#9657)
slaren committed
September 22, 2024
J
CUDA: enable Gemma FA for HIP/Pascal (#9581)
Johannes Gäßler committed
M
RWKV v6: RWKV_WKV op CUDA implementation (#9454)
Molly Sophia committed
September 20, 2024
J
ggml/examples: add backend support for numerical optimization (ggml/949)
Johannes Gäßler committed
September 12, 2024
G
ggml : hide ggml_object, ggml_cgraph, ggml_hash_set (#9408)
Georgi Gerganov committed
September 8, 2024
G
metal : update support condition for im2col + fix warning (#0)
Georgi Gerganov committed
September 3, 2024
J
tests: add gradient tests for all backends (ggml/932)
Johannes Gäßler committed
August 28, 2024
S
ggml : fix cont with transposed tensors when one dimension is 1 (ggml/934)
Salvatore Mesoraca committed
September 6, 2024
C
ggml-quants : ternary packing for TriLMs and BitNet b1.58 (#8151)
compilade committed
August 27, 2024
G
sync : ggml
Georgi Gerganov committed
August 26, 2024
G
ggml : add SSM Metal kernels (#8546)
Georgi Gerganov committed
S
metal : gemma2 flash attention support (#9159)
slaren committed
August 24, 2024
J
CPU/CUDA: Gemma 2 FlashAttention support (#8542)
Johannes Gäßler committed
August 20, 2024
Z
[SYCL] Fix SYCL `im2col` and `convert` Overflow with Large Dims (#9052)
zhentaoyu committed
August 6, 2024
M
ggml : add epsilon as a parameter for group_norm (#8818)
Molly Sophia committed
August 5, 2024
0
August 2, 2024
M
cann: Fix ggml_cann_im2col for 1D im2col (#8819)
Mengqing Cao committed
August 1, 2024
S
cuda : fix dmmv cols requirement to 2*GGML_CUDA_DMMV_X (#8800)
slaren committed
July 27, 2024
S
ggml : reduce hash table reset cost (#8698)
slaren committed
July 19, 2024
S
ggml : fix quant dot product with odd number of blocks (#8549)
slaren committed
July 17, 2024
H
[CANN] Add Ascend NPU backend (#6035)
hipudding committed
July 8, 2024
G
tests : fix whitespace (#0)
Georgi Gerganov committed
July 2, 2024
J
feat: cuda implementation for `ggml_conv_transpose_1d` (ggml/854)
John Balis committed
S
cuda : update supports_op for matrix multiplication (#8245)
slaren committed