Commits

Branch selector

 tortoise

User selector

All users

Datepicker

All time

Commit History

Commits on Jul 19, 2024

• crude working version of ggml_conv_1d float 32 support

  balisujohn

  committed
  e6846fc

  Copy full SHA for e6846fc

• some changes towards float32 conv1d

  balisujohn

  committed
  bbccd9a

  Copy full SHA for bbccd9a

Commits on Jul 1, 2024

• patched getrows

  balisujohn

  committed
  e9ac38f

  Copy full SHA for e9ac38f

Commits on Jun 25, 2024

• op count merge

  balisujohn

  committed
  5e92393

  Copy full SHA for 5e92393

Commits on Jun 22, 2024

• Merge branch 'dev-unfold-1d' into exp-mixture-6

  balisujohn

  committed
  490925e

  Copy full SHA for 490925e

• cleanup

  balisujohn

  committed
  d1f4020

  Copy full SHA for d1f4020

• added cuda kernel for ggml_unfold_1d and verified it works with existing tests

  balisujohn

  committed
  07a49f7

  Copy full SHA for 07a49f7

Commits on Jun 21, 2024

• CPU version of ggml_unfold_1d seems to work with basic tests added

  balisujohn

  committed
  dc3c545

  Copy full SHA for dc3c545

• partial draft of ggml_unfold_1d, some logic still placeholder

  balisujohn

  committed
  dc6b2d4

  Copy full SHA for dc6b2d4

Commits on Jun 20, 2024

• Merge branch 'dev-pad-prefix' into exp-mixture-5

  balisujohn

  committed
  c04ee15

  Copy full SHA for c04ee15

Commits on Jun 19, 2024

• adds ggml_pad_ext and appropriately extends tests

  balisujohn

  committed
  1865953

  Copy full SHA for 1865953

• now supports up to 4d prefix andpostfix padding with CPU, and up to 3d prefix and postfix padding with cuda backend

  balisujohn

  committed
  7810e6f

  Copy full SHA for 7810e6f

Commits on Jun 18, 2024

• removed unnecessary arguments

  balisujohn

  committed
  1ea1df1

  Copy full SHA for 1ea1df1

• working version of ggml_pad with option for prefix padding

  balisujohn

  committed
  ddeaf15

  Copy full SHA for ddeaf15

Commits on Jun 15, 2024

• fixed accumulator bug

  balisujohn

  committed
  a9b38d2

  Copy full SHA for a9b38d2

Commits on Jun 14, 2024

• removed some unnecessary arguments and fixed a bug where test failure would cause subsequent tests to fail

  balisujohn

  committed
  b22da17

  Copy full SHA for b22da17

• Merge branch 'dev-conv-transpose-1d-cuda' into exp-mixture-4

  balisujohn

  committed
  b9e20a4

  Copy full SHA for b9e20a4

• added tests for ggml_pad_reflect_1d, also added assert to make sure pad is shorter than existing padded dimension

  balisujohn

  committed
  272e57d

  Copy full SHA for 272e57d

Commits on Jun 13, 2024

• restored test-conv-transpose.c

  balisujohn

  committed
  152e04e

  Copy full SHA for 152e04e

• removed use cuda hardcoding

  balisujohn

  committed
  7eff0ab

  Copy full SHA for 7eff0ab

• added a test for large tensors

  balisujohn

  committed
  f3bb758

  Copy full SHA for f3bb758

• working with all old and new conv1d tests

  balisujohn

  committed
  53a4fcf

  Copy full SHA for 53a4fcf

Commits on Jun 12, 2024

• initial draft appears to work with stride other than 1

  balisujohn

  committed
  f35d3ec

  Copy full SHA for f35d3ec

Commits on Jun 11, 2024

• working for different input and output channel counts, added test for variable stride

  balisujohn

  committed
  f6883de

  Copy full SHA for f6883de

Commits on Jun 10, 2024

• conv transpose 1d passing test for 1d input and kernel

  balisujohn

  committed
  70de8b7

  Copy full SHA for 70de8b7

Commits on Jun 6, 2024

• initial draft of ggml_pad_reflect_1d

  balisujohn

  committed
  5dafdf0

  Copy full SHA for 5dafdf0

Commits on May 29, 2024

• sync : llama.cpp

  Show description for 2aae01f
  ggerganov

  committed
  2aae01f

  Copy full SHA for 2aae01f

• examples : adapt to new ggml_concat (#0)

  ggerganov

  committed
  f911bed

  Copy full SHA for f911bed

• ggml : fix typo in ggml.c (llama/7603)

  

  jeffzhou2000

  authored and

  ggerganov

  committed
  2ae2db6

  Copy full SHA for 2ae2db6

• Align GEMM dispatch (llama/7566)

  Show description for b1c8a17
  

  airMeng

  authored and

  ggerganov

  committed
  b1c8a17

  Copy full SHA for b1c8a17

• sycl : fix assert (llama/7563)

  ggerganov

  committed
  c052218

  Copy full SHA for c052218

• vulkan: properly initialize vulkan devices for LLAMA_SPLIT_MODE_NONE (llama/7552)

  

  Adriankhl

  authored and

  ggerganov

  committed
  df8d6dd

  Copy full SHA for df8d6dd

• rpc : resource management rework (llama/7562)

  Show description for 0a0070b
  

  rgerganov

  authored and

  ggerganov

  committed
  0a0070b

  Copy full SHA for 0a0070b

• fix ggml_sycl_mul_mat_id() to match the change of api (llama/7436)

  Show description for ef03773
  

  arthw

  authored and

  ggerganov

  committed
  ef03773

  Copy full SHA for ef03773

• ggml : generalize GGML_OP_CONCAT (llama/7563)

  Show description for abf0f9f
  ggerganov

  committed
  abf0f9f

  Copy full SHA for abf0f9f

Pagination

Previous

Next