Code

Bump tqdm from 4.66.2 to 4.66.3#21

Code

Closed

dependabot[bot] wants to merge 35 commits intomasternomic-ai/llama.cpp:masterfrom dependabot/pip/tqdm-4.66.3nomic-ai/llama.cpp:dependabot/pip/tqdm-4.66.3Copy head branch name to clipboard

+589-205

Commits

Commits on Jul 18, 2024

• llama : expose llama_load_model_from_file_gpt4all

  cebtenzzre

  committed
  506d2fe

  Copy full SHA for 506d2fe

• kompute : fix ggml_vk_device leaks

  cebtenzzre

  committed
  88289b5

  Copy full SHA for 88289b5

• kompute : fix c++11 compatibility

  cebtenzzre

  committed
  3f7c4b9

  Copy full SHA for 3f7c4b9

• kompute : enable Pascal GPUs

  cebtenzzre

  committed
  9d5207b

  Copy full SHA for 9d5207b

• llama : wrap llama_new_context_with_model in try/catch

  Show description for b6891bc
  cebtenzzre

  committed
  b6891bc

  Copy full SHA for b6891bc

• kompute : add missing call to ggml_backend_kompute_device_unref

  cebtenzzre

  committed
  b80287e

  Copy full SHA for b80287e

• kompute : fix ggml_vk_allocate failure control flow

  Show description for dc7a50b
  cebtenzzre

  committed
  dc7a50b

  Copy full SHA for dc7a50b

• kompute : disable GPU offload for Mixtral

  Show description for c5014f6
  cebtenzzre

  committed
  c5014f6

  Copy full SHA for c5014f6

• kompute : do not list Intel GPUs as they are unsupported (#14)

  cebtenzzre

  committed
  c76f5c3

  Copy full SHA for c76f5c3

• kompute : make partial tensor copies faster by syncing less data (#15)

  cebtenzzre

  committed
  6ff4387

  Copy full SHA for 6ff4387

• kompute : disable LLAMA_SPLIT_LAYER after ggerganov#5321

  cebtenzzre

  committed
  12dcddc

  Copy full SHA for 12dcddc

• kompute : add gemma, phi-2, qwen2, and stablelm to whitelist

  cebtenzzre

  committed
  82b50e5

  Copy full SHA for 82b50e5

• kompute : enable GPU support for 10 more model architectures

  Show description for a76f5f4
  cebtenzzre

  committed
  a76f5f4

  Copy full SHA for a76f5f4

• llama : fix -Wunused-const-variable warning for non-Kompute build

  cebtenzzre

  committed
  877851b

  Copy full SHA for 877851b

• llama : expose model name and architecture via API

  cebtenzzre

  committed
  729d661

  Copy full SHA for 729d661

• kompute : put device with most VRAM first, not least

  cebtenzzre

  committed
  2b8cb26

  Copy full SHA for 2b8cb26

• vulkan : make ggml_vk_instance_init static

  cebtenzzre

  committed
  6e0b5d9

  Copy full SHA for 6e0b5d9

• vulkan : don't filter devices by default, don't abort if none

  cebtenzzre

  committed
  aea0abe

  Copy full SHA for aea0abe

• vulkan : implement ggml_vk_available_devices

  cebtenzzre

  committed
  535c7b1

  Copy full SHA for 535c7b1

• vulkan : guard against multiple initialization

  Show description for 2a91dbf
  cebtenzzre

  committed
  2a91dbf

  Copy full SHA for 2a91dbf

• rocm : symlink source files so CUDA can be built in the same project

  cebtenzzre

  committed
  ad1ab57

  Copy full SHA for ad1ab57

• cuda : implement ggml_cuda_available_devices

  cebtenzzre

  committed
  09058b1

  Copy full SHA for 09058b1

• kompute : update submodule for install fix

  cebtenzzre

  committed
  b0ccbe1

  Copy full SHA for b0ccbe1

• kompute : fix leaks in ggml_vk_current_device

  cebtenzzre

  committed
  74a41c6

  Copy full SHA for 74a41c6

• kompute : fix use-after-free in ggml_vk_get_device

  cebtenzzre

  committed
  f10326c

  Copy full SHA for f10326c

• llama : replace ngl=0 hack with llama_model_using_gpu

  cebtenzzre

  committed
  e5c0df7

  Copy full SHA for e5c0df7

• llama : use the correct buffer type when we choose not to load on GPU

  Show description for 159235e
  cebtenzzre

  committed
  159235e

  Copy full SHA for 159235e

• kompute : update for leak fixes, cleanup changes, shaderFloat16

  cebtenzzre

  committed
  c301b42

  Copy full SHA for c301b42

• kompute : plug a few memory leaks

  cebtenzzre

  committed
  7d402b3

  Copy full SHA for 7d402b3

• common : Kompute supports --main-gpu, do not warn

  cebtenzzre

  committed
  48a830c

  Copy full SHA for 48a830c

• kompute : fix dangling references in ggml_vk_graph_kompute

  Show description for 6e0ad3c
  cebtenzzre

  committed
  6e0ad3c

  Copy full SHA for 6e0ad3c

• kompute : avoid freeing device/instance until absolutely necessary

  Show description for c3d5264
  cebtenzzre

  committed
  c3d5264

  Copy full SHA for c3d5264

• kompute : update ggml_vk_supports_op to fix false pos/neg

  Show description for 561d0ce
  cebtenzzre

  committed
  561d0ce

  Copy full SHA for 561d0ce

• kompute : fix missing unref on allocation failure

  cebtenzzre

  committed
  cd13f44

  Copy full SHA for cd13f44

• Bump tqdm from 4.66.2 to 4.66.3

  Show description for 3d4c558
  dependabot[bot]

  authored
  3d4c558

  Copy full SHA for 3d4c558