Commits

Breadcrumbs

History for

ond061bf9

User selector

All users

Datepicker

All time

Commit History

Commits on Jan 5, 2024

• ggml : add error handling to graph_compute (whisper/1714)

  

  finnvoor

  authored and

  ggerganov

  committed
  1bf681f

  Copy full SHA for 1bf681f

Commits on Jan 3, 2024

• cuda : simplify expression

  Show description for 7bed7eb
  

  ggerganov

  and

  slaren

  committed
  7bed7eb

  Copy full SHA for 7bed7eb

• cuda : mark I16 and I32 ops as unsupported

  Show description for d55356d
  ggerganov

  committed
  d55356d

  Copy full SHA for d55356d

Commits on Dec 30, 2023

• CUDA: fixed tensor cores not being used on RDNA3 (ggerganov#4697)

  JohannesGaessler

  authored
  39d8bc7

  Copy full SHA for 39d8bc7

Commits on Dec 29, 2023

• CUDA: fix tensor core logic for Pascal and HIP (ggerganov#4682)

  JohannesGaessler

  authored
  a20f3c7

  Copy full SHA for a20f3c7

• cuda: fix vmm oom issue on NVIDIA AGX Orin (ggerganov#4687)

  Show description for 91bb39c
  hydai

  authored
  91bb39c

  Copy full SHA for 91bb39c

• ggml : fix some mul mat cases + add tests for src1 F16 (ggml/669)

  Show description for afc8c19
  

  3 peopleauthored and

  ggerganov

  committed
  afc8c19

  Copy full SHA for afc8c19

Commits on Dec 26, 2023

• cuda : fix vmm pool with multi GPU (ggerganov#4620)

  Show description for dc68f00
  slaren

  authored
  dc68f00

  Copy full SHA for dc68f00

• Fix new CUDA10 compilation errors (ggerganov#4635)

  FantasyGmm

  authored
  77465da

  Copy full SHA for 77465da

Commits on Dec 24, 2023

• cuda : improve cuda pool efficiency using virtual memory (ggerganov#4606)

  Show description for 5bf3953
  

  slaren

  and

  JohannesGaessler

  authored
  5bf3953

  Copy full SHA for 5bf3953

Commits on Dec 23, 2023

• fallback to CPU buffer if host buffer alloc fails (ggerganov#4610)

  slaren

  authored
  708e179

  Copy full SHA for 708e179

• CUDA: fixed row rounding for 0 tensor splits (ggerganov#4594)

  JohannesGaessler

  authored
  e0a4002

  Copy full SHA for e0a4002

Commits on Dec 22, 2023

• sync : ggml (fix im2col) (ggerganov#4591)

  Show description for ba66175
  

  ggerganov

  and

  leejet

  authored
  ba66175

  Copy full SHA for ba66175

• cuda : fix jetson compile error (ggerganov#4560)

  Show description for a558769
  

  FantasyGmm

  and

  ggerganov

  authored
  a558769

  Copy full SHA for a558769

• Fix CudaMemcpy direction (ggerganov#4599)

  Ttl

  authored
  6724ef1

  Copy full SHA for 6724ef1

• llama : fix platforms without mmap (ggerganov#4578)

  Show description for 48b7ff1
  slaren

  authored
  48b7ff1

  Copy full SHA for 48b7ff1

Commits on Dec 21, 2023

• ggml : change ggml_scale to take a float instead of tensor (ggerganov#4573)

  Show description for afefa31
  ggerganov

  authored
  afefa31

  Copy full SHA for afefa31

• llama : initial ggml-backend integration (ggerganov#4520)

  Show description for d232aca
  

  slaren

  and

  ggerganov

  authored
  d232aca

  Copy full SHA for d232aca

• cuda : ROCm AMD Unified Memory Architecture (UMA) handling (ggerganov#4449)

  Show description for 0f630fb
  ekg

  authored
  0f630fb

  Copy full SHA for 0f630fb

• ggml-cuda: Fix HIP build by adding define for __trap (ggerganov#4569)

  Show description for 562cf22
  arlo-phoenix

  authored
  562cf22

  Copy full SHA for 562cf22

• CUDA: mul_mat_id always on GPU for batches >= 32 (ggerganov#4553)

  JohannesGaessler

  authored
  9154494

  Copy full SHA for 9154494

• cuda : better error message for ggml_get_rows (ggerganov#4561)

  Show description for 66f35a2
  

  bobqianic

  and

  ggerganov

  authored
  66f35a2

  Copy full SHA for 66f35a2

• cuda : replace asserts in wrong architecture checks with __trap (ggerganov#4556)

  Show description for 1398823
  slaren

  authored
  1398823

  Copy full SHA for 1398823

• Fix access violation in ggml_cuda_free_data if tensor->extra is NULL (ggerganov#4554)

  LoganDark

  authored
  1d7a191

  Copy full SHA for 1d7a191

Commits on Dec 20, 2023

• CUDA: Faster Mixtral prompt processing (ggerganov#4538)

  Show description for 799fc22
  

  JohannesGaessler

  and

  slaren

  authored
  799fc22

  Copy full SHA for 799fc22

Commits on Dec 18, 2023

• ggml-cuda: Fix HIP build (ggerganov#4528)

  Show description for a7aee47
  arlo-phoenix

  authored
  a7aee47

  Copy full SHA for a7aee47

• llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (ggerganov#4490)

  Show description for b9e74f9
  

  4 peopleauthored
  b9e74f9

  Copy full SHA for b9e74f9

Commits on Dec 14, 2023

• ggml : use ggml_row_size where possible (ggerganov#4472)

  Show description for 6744dbe
  slaren

  authored
  6744dbe

  Copy full SHA for 6744dbe

Commits on Dec 13, 2023

• sync : ggml (SD ops, tests, kernels) (ggerganov#4444)

  Show description for 4d98d9a
  

  ggerganov

  and

  slaren

  authored
  4d98d9a

  Copy full SHA for 4d98d9a

• llama : add Mixtral support (ggerganov#4406)

  Show description for 799a1cb
  

  3 peopleauthored
  799a1cb

  Copy full SHA for 799a1cb

Commits on Dec 7, 2023

• sync : ggml (new ops, tests, backend, etc.) (ggerganov#4359)

  Show description for fe680e3
  

  ggerganov

  and

  slaren

  authored
  fe680e3

  Copy full SHA for fe680e3

• llama : per-layer KV cache + quantum K cache (ggerganov#4309)

  Show description for bcc0eb4
  

  ggerganov

  and

  slaren

  authored
  bcc0eb4

  Copy full SHA for bcc0eb4

Commits on Dec 1, 2023

• ggml : add ggml_soft_max_ext (ggerganov#4256)

  Show description for ef47ec1
  ggerganov

  authored
  ef47ec1

  Copy full SHA for ef47ec1

Commits on Nov 24, 2023

• ggml-cuda : support stablelm rope (ggerganov#4156)

  Show description for 8a052c1
  

  slaren

  and

  ggerganov

  authored
  8a052c1

  Copy full SHA for 8a052c1

Commits on Nov 23, 2023

• Fix incorrect format strings and uninitialized variables. (ggerganov#4133)

  Show description for 55978ce
  haohui

  authored
  55978ce

  Copy full SHA for 55978ce

Pagination

Previous

Next