Commits

Branch selector

 master

User selector

slaren

Datepicker

All time

Commit History

Commits on Jan 26, 2024

• cuda : fix tensor size calculation for non-split buffer (ggerganov#5145)

  slaren

  authored
  62fead3

• ggml-alloc : add 10% margin to the buffer sizes (ggerganov#5149)

  slaren

  authored
  15b4538

Commits on Jan 24, 2024

• llama : pre-allocate input tensors in a separate buffer (ggerganov#5100)

  slaren

  authored
  1387ea2

Commits on Jan 22, 2024

• llama : fix not enough space in buffer with Qwen (ggerganov#5086)

  slaren

  authored
  011e8ec

Commits on Jan 20, 2024

• llama : run all KQV ops on the CPU with no KV offload (ggerganov#5049)

  Show description for 6df465a
  slaren

  authored
  6df465a

Commits on Jan 18, 2024

• llama : fix mlock with no-mmap with Metal (ggerganov#5025)

  slaren

  authored
  96d7f56

Commits on Jan 12, 2024

• backend_sched : fix assignments

  Show description for fa5c1fb
  

  slaren

  authored and

  ggerganov

  committed
  fa5c1fb

• llama : ggml-backend integration (ggerganov#4766)

  Show description for e7e4df0
  

  3 peopleauthored
  e7e4df0

Commits on Jan 7, 2024

• llama-bench : add no-kv-offload parameter (ggerganov#4812)

  slaren

  authored
  226460c

Commits on Dec 26, 2023

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

  Show description for dc68f00
  slaren

  authored
  dc68f00

Commits on Dec 24, 2023

• simplify bug issue template (ggerganov#4623)

  slaren

  authored
  b9f4795

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

  Show description for 5bf3953
  

  slaren

  and

  JohannesGaessler

  authored
  5bf3953

Commits on Dec 23, 2023

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

  slaren

  authored
  708e179

Commits on Dec 22, 2023

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

  Show description for 48b7ff1
  slaren

  authored
  48b7ff1

Commits on Dec 21, 2023

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

  Show description for d232aca
  

  slaren

  and

  ggerganov

  authored
  d232aca

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

  Show description for 1398823
  slaren

  authored
  1398823

Commits on Dec 17, 2023

• finetune : keep allocs alive until all allocations are done (ggerganov#4486)

  slaren

  authored
  4566863

Commits on Dec 16, 2023

• lora : add support for non-llama models (ggerganov#3333)

  Show description for c6c4fc0
  slaren

  authored
  c6c4fc0

Commits on Dec 15, 2023

• ggml : group mul_mat_id rows by matrix (cpu only) (ggerganov#4480)

  Show description for ee4725a
  slaren

  authored
  ee4725a

Commits on Dec 14, 2023

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

  Show description for 6744dbe
  slaren

  authored
  6744dbe

• ggml : remove n_dims from ggml_tensor (ggerganov#4469)

  Show description for cafcd4f
  slaren

  authored
  cafcd4f

Commits on Dec 13, 2023

• llama : add Mixtral support (ggerganov#4406)

  Show description for 799a1cb
  

  3 peopleauthored
  799a1cb

Commits on Nov 30, 2023

• convert.py : fix llama/llama2 conversion due to vocab_size=-1 (ggerganov#4258)

  slaren

  authored
  f4d973c

Commits on Nov 24, 2023

• llama : set metal log callback correctly (ggerganov#4204)

  slaren

  authored
  e9c13ff

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

  Show description for 8a052c1
  

  slaren

  and

  ggerganov

  authored
  8a052c1

Commits on Nov 19, 2023

• gguf-py : export chat templates (ggerganov#4125)

  Show description for e937066
  slaren

  authored
  e937066

Commits on Nov 17, 2023

• llama : increase max nodes (ggerganov#4115)

  slaren

  authored
  bbecf3f

• llama : add functions to get the model's metadata (ggerganov#4013)

  Show description for e85bb1a
  slaren

  authored
  e85bb1a

Commits on Nov 15, 2023

• ggml-cuda : increase max graph size (ggerganov#4084)

  slaren

  authored
  1cf2850

Commits on Nov 8, 2023

• ggml-alloc : fix backend assignments of views (ggerganov#3982)

  slaren

  authored
  875fb42

Commits on Nov 5, 2023

• ggml-cuda : fix f16 mul mat (ggerganov#3961)

  Show description for 2833a6f
  slaren

  authored
  2833a6f

• cuda : revert CUDA pool stuff (ggerganov#3944)

  Show description for 48ade94
  slaren

  authored
  48ade94

Commits on Nov 3, 2023

• ggml-cuda : move row numbers to x grid dim in mmv kernels (ggerganov#3921)

  slaren

  authored
  abb77e7

Commits on Nov 2, 2023

• cmake : disable LLAMA_NATIVE by default (ggerganov#3906)

  slaren

  authored
  21958bb

Commits on Nov 1, 2023

• ggml-cuda : compute ptrs for cublasGemmBatchedEx in a kernel (ggerganov#3891)

  Show description for d02e98c
  slaren

  authored
  d02e98c

Pagination

PreviousNext