Commit graph

  • c4d1c26710 Merge branch 'master' into add_gpt2_support manikbhandari 2023-12-26 07:23:04 -05:00
  • de8e496437
    Update comment for AdamW implementation reference. (#4604) b1702 WillCorticesAI 2023-12-26 05:42:08 -05:00
  • 77465dad48
    Fix new CUDA10 compilation errors (#4635) b1701 FantasyGmm 2023-12-26 18:38:36 +08:00
  • ff8b994161
    remove unused variable wonjun Jang 2023-12-26 19:25:32 +09:00
  • 9f297f81ad
    Add byte token type to hf format wonjun Jang 2023-12-26 19:21:36 +09:00
  • ae6bebccb1
    nix: respect config.cudaCapabilities Someone Serge 2023-12-26 04:33:24 +00:00
  • 4522c47a22
    nix: fewer rebuilds Someone Serge 2023-12-26 04:05:51 +00:00
  • dd0e12c741
    nix: cuda: reduce runtime closure Someone Serge 2023-12-26 04:02:11 +00:00
  • a16f58997c
    nix: cmakeFlags: explicit OFF bools Someone Serge 2023-12-26 03:58:02 +00:00
  • 12d4a68efe
    ci: add the (Nix) flakestry workflow Someone Serge 2023-12-25 17:05:21 +00:00
  • e3b1ba27c2
    flake.nix: overlay: expose both the namespace and the default attribute Someone Serge 2023-12-25 17:03:19 +00:00
  • a629371245
    nix: migrate to pname+version Someone Serge 2023-12-25 17:02:36 +00:00
  • a28c9acca3
    flake.nix: use flake-parts Someone Serge 2023-12-25 16:18:52 +00:00
  • d08690af65
    flake.nix: avoid re-evaluating nixpkgs too many times Someone Serge 2023-12-25 15:27:58 +00:00
  • 04bc417466
    nix: add missing meta attributes Someone Serge 2023-12-25 17:32:44 +00:00
  • a07407c98d
    nix: move meta down to follow Nixpkgs style more closely Someone Serge 2023-12-25 17:29:50 +00:00
  • 69c56bc0f4
    flake.nix: add maintainers Someone Serge 2023-12-25 17:29:02 +00:00
  • 0fa62c1ab9
    *.nix: nixfmt Someone Serge 2023-12-25 16:23:56 +00:00
  • 0259941791
    nix: pass most darwin frameworks unconditionally Someone Serge 2023-12-24 19:36:30 +00:00
  • eab1c125b9
    nix: unclutter darwin support Someone Serge 2023-12-24 19:35:32 +00:00
  • 0607e24ec2
    flake.nix: use finalPackage instead of passing it manually Someone Serge 2023-12-24 18:15:25 +00:00
  • 8364cf4d0b
    flake.nix: rewrite Philip Taron 2023-12-22 12:33:09 -08:00
  • 5006b23099
    CUDA 11.4 for Github CI (#582) henk717 2023-12-26 04:23:43 +01:00
  • d3e73df66a
    flake.lock: update to hotfix CUDA::cuda_driver Someone Serge 2023-12-26 03:09:04 +00:00
  • 272ddcdde4
    Fix new CUDA10 compilation errors FantasyGmm 2023-12-26 10:00:25 +08:00
  • da9fc775a3 fix scratch buffer size, re-enable vmm pool for all devices slaren 2023-12-26 01:28:39 +01:00
  • 23c6dd677b only use vmm for the main device slaren 2023-12-25 22:49:06 +01:00
  • 0dcc1a77d7 style fixes slaren 2023-12-25 21:23:31 +01:00
  • 561f1f9500 remove unnecessary inlines slaren 2023-12-25 20:56:23 +01:00
  • 692887fbe4 consolidate error checking in ggml_cuda_set_device slaren 2023-12-25 20:53:23 +01:00
  • 32304d796f use cuda_pool_alloc in ggml_cuda_op_mul_mat slaren 2023-12-25 20:44:10 +01:00
  • 865d042d56 use cudaMemcpy3DPeerAsync slaren 2023-12-25 19:07:32 +01:00
  • d98f4ab00e
    nix: passthru: expose the use* flags for inspection Someone Serge 2023-12-25 18:09:28 +00:00
  • 42c7bbed76
    nix: clarify the meaning of "broken" and "badPlatforms" Someone Serge 2023-12-25 17:59:16 +00:00
  • 12115d62da
    nix: clarify the interpretation of meta.maintainers Someone Serge 2023-12-25 17:45:19 +00:00
  • 5dfe9ba4b6
    nix: add missing meta attributes Someone Serge 2023-12-25 17:32:44 +00:00
  • 0e614077c3
    nix: move meta down to follow Nixpkgs style more closely Someone Serge 2023-12-25 17:29:50 +00:00
  • 0925e7ee7c
    flake.nix: add maintainers Someone Serge 2023-12-25 17:29:02 +00:00
  • 34f2477854
    *.nix: nixfmt Someone Serge 2023-12-25 16:23:56 +00:00
  • e8d02f4664
    ci: add the flakestry workflow Someone Serge 2023-12-25 17:05:21 +00:00
  • 869e1e0961
    flake.nix: overlay: expose both the namespace and the default attribute Someone Serge 2023-12-25 17:03:19 +00:00
  • cc37631f7c
    flake.nix: migrate to pname+version Someone Serge 2023-12-25 17:02:36 +00:00
  • 286bb392a5
    flake.nix: use flake-parts Someone Serge 2023-12-25 16:18:52 +00:00
  • 4b12b0579f
    flake.nix: avoid re-evaluating nixpkgs too many times Someone Serge 2023-12-25 15:27:58 +00:00
  • c8d39878d9
    *.nix: nixfmt Someone Serge 2023-12-25 16:23:56 +00:00
  • a206137f92
    Adding Emeltal reference to UI list (#4629) b1700 Paul Tsochantaris 2023-12-25 16:09:53 +00:00
  • c2d87b6545 increase multiuser default Concedo 2023-12-25 23:49:45 +08:00
  • 78a9d206d3 randomize horde genkey Concedo 2023-12-25 22:47:21 +08:00
  • e692c2d887 CUDA: fixed peer access toggle synchronization JohannesGaessler 2023-12-22 15:21:36 +01:00
  • fbdd6fd739 Adding Emeltal reference to UI list Paul Tsochantaris 2023-12-25 13:23:41 +00:00
  • 1659cd1bc4 fix mixtral slaren 2023-12-25 13:24:36 +01:00
  • cc64f2cad1 Merge branch 'master' into concedo_experimental Concedo 2023-12-25 18:47:21 +08:00
  • d089842447 update: readme namtranase 2023-12-25 17:24:56 +07:00
  • 44f4ce2272 Merge branch 'master' of https://github.com/namtranase/llama.cpp namtranase 2023-12-25 17:13:50 +07:00
  • 13f60c417d uppdate: remove use_awq arg namtranase 2023-12-25 17:13:06 +07:00
  • 293395e0f5 Merge commit '708e179e85' into concedo_experimental Concedo 2023-12-25 16:48:15 +08:00
  • b0583f71d3
    Merge branch 'master' into phi-1 teleprint-me 2023-12-24 21:42:01 -05:00
  • fea0239e24 flake.nix: pass most darwin frameworks unconditionally Someone Serge 2023-12-24 19:36:30 +00:00
  • d73272f18b flake.nix: unclutter darwin support Someone Serge 2023-12-24 19:35:32 +00:00
  • b28426a952 fixup! flake.nix: rewrite Someone Serge 2023-12-24 18:24:28 +00:00
  • 76f348aabb flake.nix: use finalPackage instead of passing it manually Someone Serge 2023-12-24 18:15:25 +00:00
  • b9f47952ff
    simplify bug issue template (#4623) b1699 slaren 2023-12-24 21:01:12 +01:00
  • c01f8f3cc7 simplify bug issue template slaren 2023-12-24 19:16:52 +01:00
  • 6f35a4a6e9 better error checking slaren 2023-12-24 19:04:36 +01:00
  • a76cadad48 use recommended granularity instead of minimum slaren 2023-12-24 18:36:43 +01:00
  • 2c3fbf980c hip slaren 2023-12-24 18:03:50 +01:00
  • 32dc09aab3 cuda : fix vmm pool with multi GPU slaren 2023-12-24 17:53:14 +01:00
  • 753be377b6
    llama : add PLaMo model (#3557) b1698 Shintarou Okada 2023-12-24 22:35:49 +09:00
  • 5bf3953d7e
    cuda : improve cuda pool efficiency using virtual memory (#4606) b1697 slaren 2023-12-24 14:34:22 +01:00
  • 1949c95598
    plamo : fix tensor names for correct GPU offload Georgi Gerganov 2023-12-24 15:33:31 +02:00
  • eedd434575
    llama : remove obsolete KQ_scale Georgi Gerganov 2023-12-24 15:30:12 +02:00
  • bd0d9039ec better approach to multiuser check Concedo 2023-12-24 20:03:33 +08:00
  • 532cb9b99c act_size -> actual_size slaren 2023-12-24 12:26:35 +01:00
  • 3ad45fc3a8 ensure allocations are always aligned slaren 2023-12-24 12:26:04 +01:00
  • 3081c4e768
    Update ggml-cuda.cu slaren 2023-12-24 12:13:55 +01:00
  • b9c5a6e74a
    Update ggml-cuda.cu slaren 2023-12-24 12:12:08 +01:00
  • 5acc9e50f5 cuda : fallback to CPU on host buffer alloc fail slaren 2023-12-24 11:59:15 +01:00
  • ab6ad5e6e8 minor slaren 2023-12-24 11:58:36 +01:00
  • 307481f28d update okada 2023-12-24 18:57:53 +09:00
  • 602cc71b06 Merge branch 'master' into add_pfnet_plamo_13b_2 okada 2023-12-24 18:45:18 +09:00
  • 700f7c600a fix style okada 2023-12-24 18:15:20 +09:00
  • 26340a1902 remove plamo_llm_build_kqv and use llm_build_kqv okada 2023-12-24 18:14:12 +09:00
  • bc24c9334c prevent prompt leakage during usage of check endpoint when genkey is provided in multiuser mode Concedo 2023-12-24 17:08:43 +08:00
  • db1b18dc97 shuffle attn_q.weight and attn_output.weight for broadcasting okada 2023-12-24 17:58:55 +09:00
  • cbf07e7178
    flake.nix: rewrite Philip Taron 2023-12-22 12:33:09 -08:00
  • 1cdfdb34e0 fixed hpx runtime initialization and finalization ct-clmsn 2023-12-23 22:35:47 -05:00
  • e1eb3d14f1 initial import of hpx support ct-clmsn 2023-12-23 22:15:52 -05:00
  • 776f5e29cd initial import of hpx support ct-clmsn 2023-12-23 21:59:34 -05:00
  • d3f155733f updated thread support ct-clmsn 2023-12-23 20:49:12 -05:00
  • 348d565aa5
    Merge branch 'master' into phi-1 teleprint-me 2023-12-23 19:10:57 -05:00
  • 16ad7f74ee Update workflow crasm 2023-12-23 18:55:55 -05:00
  • 91f318c228 Enforce "compatible release" syntax instead of == crasm 2023-12-23 18:27:43 -05:00
  • dc209c09a8 Fail on "==" being used for package requirements (but can be suppressed) crasm 2023-12-23 18:06:24 -05:00
  • bc1b0d5351 Add flake app to run openai proxy paretoOptimalDev 2023-12-23 16:50:23 -06:00
  • b6a9efb0e0 Move requirements into ./requirements crasm 2023-12-23 17:18:21 -05:00
  • 26e97b5861 Merge remote-tracking branch 'origin/master' into sl/cuda-virt-pool slaren 2023-12-23 19:26:38 +01:00
  • d8883623a6 minor slaren 2023-12-23 19:25:29 +01:00
  • 6fe9da0f8a ggml : fix msvc warnings slaren 2023-12-23 19:20:59 +01:00
  • 5eb626225d llama : fix msvc warnings slaren 2023-12-23 19:13:52 +01:00
  • 9f5ac6d2d2 more hip crap slaren 2023-12-23 18:05:35 +01:00