repo
stringclasses
15 values
fix_commit
stringlengths
40
40
buggy_commit
stringlengths
40
40
message
stringlengths
3
64.3k
files
listlengths
1
300
timestamp
timestamp[s]date
2013-03-13 20:45:00
2026-04-11 07:48:46
ggml-org/llama.cpp
63908b631a2cef6491c2b6ce1c504f046660183f
42b12b560886dc2093b17af11c97ef6d276a3b97
cmake: fix Mach-O current version number (#17877) PR #17091 set the VERSION of various libraries to 0.0.abcd, where abcd is the LLAMA_BUILD_NUMBER. That build number is too large to fit in the Mach-O 'current version' field's 'micro' part, which only goes up to 255. This just sets the Mach-O current version to 0 to ge...
[ { "path": "src/CMakeLists.txt", "patch": "@@ -139,6 +139,7 @@ add_library(llama\n set_target_properties(llama PROPERTIES\n VERSION ${LLAMA_INSTALL_VERSION}\n SOVERSION 0\n+ MACHO_CURRENT_VERSION 0 # keep macOS linker from seeing oversized version number\n )\n \n target_include_directories(llama P...
2025-12-09T11:17:41
huggingface/transformers
663b4cf121d2ab284bb24dd8174e53a79c404d15
dc7364824d3020a261068618e31e190ba2b473fb
tiny fix for deepseekocr support [vllm] (#42423) * tiny fix * fix usage * only need the config * remove that too * fixup
[ { "path": "src/transformers/models/deepseek_v2/configuration_deepseek_v2.py", "patch": "@@ -101,6 +101,9 @@ class DeepseekV2Config(PreTrainedConfig):\n Number of selected groups per token for expert selection.\n topk_method (`str`, *optional*, defaults to `\"greedy\"`):\n The...
2025-11-27T14:25:16
vuejs/vue
58a39dfa0e8c4a51959e9a84369dad8fbca0e6ac
0c703e34d1a2083d9f162fcf0885deefb803182e
fix(model): correctly set select v-model initial value on patch (#6910)
[ { "path": "src/platforms/web/runtime/directives/model.js", "patch": "@@ -6,6 +6,8 @@\n import { isTextInputType } from 'web/util/element'\n import { looseEqual, looseIndexOf } from 'shared/util'\n import { warn, isAndroid, isIE9, isIE, isEdge } from 'core/util/index'\n+import { mergeVNodeHook } from 'core/v...
2017-11-03T02:03:19
denoland/deno
5b5f84c8488dab766e8687c649c0a11b89b4006a
01f99e8133a4efd351e65cbaa8a6b7eb1e5b49eb
fix(ext/crypto): fix P-521 EC key export for all formats (#32412) ## Summary - **Fixed P-521 key generation** to output PKCS8 DER (matching P-256/P-384 behavior) instead of raw scalar bytes, which broke JWK private, JWK public (from private handle), and SPKI exports - **Added P-521 JWK private export arm** in the `ex...
[ { "path": "ext/crypto/export_key.rs", "patch": "@@ -410,7 +410,23 @@ fn export_key_ec(\n Err(SharedError::ExpectedValidPublicECKey.into())\n }\n }\n- _ => Err(ExportKeyError::UnsupportedNamedCurve),\n+ EcNamedCurve::P521 => {\n+ let ec_key = p521::SecretK...
2026-03-03T09:50:53
ollama/ollama
703684a82a150ba59b300ca9c1ec678bd9b4ab27
6459377ae081021a0ec2b5b408c6ac4503de01b2
server: replace blob prefix separator from ':' to '-' (#3146) This fixes issues with blob file names that contain ':' characters to be rejected by file systems that do not support them.
[ { "path": "server/fixblobs.go", "patch": "@@ -0,0 +1,26 @@\n+package server\n+\n+import (\n+\t\"os\"\n+\t\"path/filepath\"\n+\t\"strings\"\n+)\n+\n+// fixBlobs walks the provided dir and replaces (\":\") to (\"-\") in the file\n+// prefix. (e.g. sha256:1234 -> sha256-1234)\n+func fixBlobs(dir string) error ...
2024-03-15T03:18:06
ggml-org/llama.cpp
4e842d5120f0349fe4c014a1cc8b9a55c5715c66
ca709e427b1c5cf0134aaabc9bbc0183f32d1df0
console: allow using arrow left/right, home/end keys and history mode (#17836) * console: allow using arrow left/right to edit the line (with UTF-8 support) * console: fix arrow keys on Windows using private-use Unicode * console: add Home/End key support for Windows and Linux * console: add basic Up/Down history n...
[ { "path": "common/console.cpp", "patch": "@@ -1,6 +1,11 @@\n #include \"console.h\"\n #include <vector>\n #include <iostream>\n+#include <cassert>\n+#include <cstddef>\n+#include <cctype>\n+#include <cwctype>\n+#include <cstdint>\n \n #if defined(_WIN32)\n #define WIN32_LEAN_AND_MEAN\n@@ -35,9 +40,26 @@\n \...
2025-12-09T10:53:59
vuejs/vue
0c703e34d1a2083d9f162fcf0885deefb803182e
bdcc05a0d40eaef0e0f76b9d975a04fe0e97f5e0
fix(v-model): v-if / v-else not working with :type + v-model (#6955) fix #6918
[ { "path": "src/platforms/web/compiler/modules/model.js", "patch": "@@ -29,6 +29,8 @@ function preTransformNode (el: ASTElement, options: CompilerOptions) {\n const typeBinding: any = getBindingAttr(el, 'type')\n const ifCondition = getAndRemoveAttr(el, 'v-if', true)\n const ifConditionExtr...
2017-11-02T21:09:29
huggingface/transformers
dc7364824d3020a261068618e31e190ba2b473fb
e1ab70974d87f0aac43a92cf8cdf489ac2e6d485
[`FA`] Cleanup loading logic (#41427) * fix * style * fix kernels loading as well * fix typing * refactor CB loading logic as well * fix base fa logic * rename * properly lazy load paged fa * fix * check if ci is crashing again * fix fallback * style * allow varlen only, e.g. for metal kernel * fixup new ...
[ { "path": "examples/pytorch/continuous_batching.py", "patch": "@@ -172,7 +172,7 @@ def batch_generate(\n \n # Model parameters\n parser.add_argument(\"--sliding-window\", type=int, default=0)\n- parser.add_argument(\"--attn\", type=str, default=\"kernels-community/flash-attn\", help=\"Attention i...
2025-11-27T14:23:13
denoland/deno
01f99e8133a4efd351e65cbaa8a6b7eb1e5b49eb
88bd1bc158d4e9860afc16429ceeee6bd4440eb6
fix: validate cron schedule when using cron socket (#32406)
[ { "path": "ext/cron/socket.rs", "patch": "@@ -490,6 +490,11 @@ impl CronHandler for SocketCronHandler {\n return Err(CronError::RejectedError(reason.clone()));\n }\n \n+ spec\n+ .cron_schedule\n+ .parse::<saffron::Cron>()\n+ .map_err(|_| CronError::InvalidCron)?;\n+\n let (in...
2026-03-03T09:50:33
ollama/ollama
8546dd3d72867be55c7439ecc58bc0790a0f556b
87100be5e0a12a9b058f331b50509e041c266738
.github: fix model and feature request yml (#3155)
[ { "path": ".github/ISSUE_TEMPLATE/10_model_request.yml", "patch": "@@ -1,6 +1,5 @@\n name: Model request\n description: Request a new model for the library\n-title: \"\"\n labels: [mr]\n body:\n - type: markdown", "additions": 0, "deletions": 1, "language": "YAML" }, { "path": ".github...
2024-03-14T22:26:06
ggml-org/llama.cpp
ca709e427b1c5cf0134aaabc9bbc0183f32d1df0
0cdce38a97bd05462416272fbb912f4d7ecd2940
CANN: add support for partial RoPE and Vision mode (#17543) * cann: add support for partial RoPE and Vision mode Add support for two important RoPE variants: partial rotation (rope_dims < ne0) and Vision mode rotation. 1. Support for partial RoPE (rope_dims < ne0): - Split tensor into head (first rope_dims dimens...
[ { "path": "ggml/src/ggml-cann/aclnn_ops.cpp", "patch": "@@ -2251,12 +2251,12 @@ static void aclnn_rope_cache_init(ggml_backend_cann_context & ctx,\n int sections[4],\n bool mrope_used,\n ...
2025-12-09T09:53:23
vuejs/vue
bdcc05a0d40eaef0e0f76b9d975a04fe0e97f5e0
c24f3e4208cd045832002ee9916559f6fe0dc2b5
chore: fix creative Tim missing from the sponsors list (#6939)
[ { "path": "README.md", "patch": "@@ -143,8 +143,7 @@ Funds donated via Patreon goes directly to support Evan You's full-time work on\n <a href=\"https://opencollective.com/vuejs/tiers/goldsponsors/1/website\" target=\"_blank\"><img src=\"https://opencollective.com/vuejs/tiers/goldsponsors/1/avatar.svg\"></a...
2017-11-02T21:05:30
huggingface/transformers
352a2e0c120370f7db3b40caa4c60f0efca4e737
7094f1eca16aebf61aff876d82ddda3e873533ac
Benchmark simplification (#42408) * Renames * Added the timestamps to request * Better rename for prompt_ids * Merged the two timing functions * Style * Remove the first timestamp for generate timing * Fix nit in comment * Re-introduce timestamps * Now upload two versions of the results: full and summarized *...
[ { "path": "benchmark_v2/framework/benchmark_runner.py", "patch": "@@ -10,6 +10,7 @@\n from queue import Queue\n from typing import Any\n \n+import numpy as np\n import torch\n from datasets import Dataset\n from huggingface_hub import HfApi\n@@ -208,10 +209,11 @@ def run_benchmark(\n self.logger...
2025-11-27T10:51:00
ggml-org/llama.cpp
0cdce38a97bd05462416272fbb912f4d7ecd2940
e39502e74b46c54b359ddbbefc9bba4f98704906
CUDA: fix FP16 overflow in tile FA kernel (#17875)
[ { "path": "ggml/src/ggml-cuda/fattn-tile.cuh", "patch": "@@ -564,6 +564,12 @@ static __device__ __forceinline__ void flash_attn_tile_iter(\n for (int i_KQ_0 = 0; i_KQ_0 < nbatch_fa; i_KQ_0 += np*warp_size) {\n const int i_KQ = i_KQ_0 + (threadIdx.y % np)*warp_size + threadIdx.x;\n \n+#if...
2025-12-09T08:34:02
vuejs/vue
c24f3e4208cd045832002ee9916559f6fe0dc2b5
f2e00f756fb540fb09ce3414289c652ce172d85c
fix(events): properly $off array of events (#6949)
[ { "path": "src/core/instance/events.js", "patch": "@@ -92,7 +92,7 @@ export function eventsMixin (Vue: Class<Component>) {\n if (!cbs) {\n return vm\n }\n- if (arguments.length === 1) {\n+ if (!fn) {\n vm._events[event] = null\n return vm\n }", "additions": 1, "de...
2017-11-02T20:58:21
ollama/ollama
291c663865fe9765aa4c541e617948a54d400590
da20786e3e22a9f5a7223081ccfec200cfbd5255
fix: clip memory leak
[ { "path": "llm/patches/05-fix-clip-free.diff", "patch": "@@ -0,0 +1,45 @@\n+From 9192432daf90b1bfec75577434a99b4ea70d54c8 Mon Sep 17 00:00:00 2001\n+From: Michael Yang <mxyng@pm.me>\n+Date: Thu, 14 Mar 2024 12:09:50 -0700\n+Subject: [PATCH] fix clip free\n+\n+---\n+ examples/llava/clip.cpp | 4 ++++\n+ ex...
2024-03-14T19:45:46
ggml-org/llama.cpp
e39502e74b46c54b359ddbbefc9bba4f98704906
1d2a1ab73d927644d58f44235209affab7e4043e
llama : add token matching support to llama-grammar (#17816) * llama : add token support to llama-grammar * fix inverse token comment * refactor trigger_patterns to replay tokens instead of the entire string * add token documentation * fix test-llama-grammar * improve test cases for tokens
[ { "path": "grammars/README.md", "patch": "@@ -67,6 +67,30 @@ Parentheses `()` can be used to group sequences, which allows for embedding alte\n - `{m,n}` repeats the precedent symbol or sequence at between `m` and `n` times (included)\n - `{0,n}` repeats the precedent symbol or sequence at most `n` times (i...
2025-12-09T06:32:57
huggingface/transformers
7c8d72bde8fe7c362aa3294f4244a8dd67af0abf
a33943ec696ca80f84e4c64ca0aa96b1b0990281
[MODEL] Nanochat implementation (#41634) * first draft on modelling * add modelling to auto * add to init * [WIP] add scripts for conversion * hack at attention and rotary using logit comparison unitl it works * update conversion script * fix test * tody up decoding inputs * fix all naming nanogpt >> nanochat ...
[ { "path": "docs/source/en/_toctree.yml", "patch": "@@ -620,6 +620,8 @@\n title: MVP\n - local: model_doc/myt5\n title: myt5\n+ - local: model_doc/nanochat\n+ title: NanoChat\n - local: model_doc/nemotron\n title: Nemotron\n - local: model_doc/nllb", ...
2025-11-27T09:21:38
vuejs/vue
f2e00f756fb540fb09ce3414289c652ce172d85c
350f5788d39741d690d49aea91c084f2db764992
fix: special case for static muted attribute in firefox fix #6887
[ { "path": "src/compiler/parser/index.js", "patch": "@@ -556,6 +556,13 @@ function processAttrs (el) {\n }\n }\n addAttr(el, name, JSON.stringify(value))\n+ // #6887 firefox doesn't update muted state if set via attribute\n+ // even immediately after element creation\n+ if ...
2017-11-02T20:55:05
denoland/deno
291532f0e5723c2498afc50e19a906ba9a5e0ee2
f6d5b957cb5d5ad398bd580e44c963588c5b36e3
fix(ext/node): support RSA PSS padding option in crypto sign/verify (#32269) ## Summary - Pass the `padding` option from `crypto.createSign().sign()` / `crypto.createVerify().verify()` through to the Rust signing ops - Handle `RSA_PKCS1_PSS_PADDING` (6) for regular RSA keys by using PSS signing/verification instead of...
[ { "path": "ext/node/polyfills/internal/crypto/sig.ts", "patch": "@@ -66,6 +66,10 @@ export interface VerifyKeyObjectInput extends SigningOptions {\n key: KeyObject;\n }\n \n+function getPadding(options) {\n+ return getIntOption(\"padding\", options);\n+}\n+\n function getSaltLength(options) {\n return ...
2026-03-03T09:28:52
huggingface/transformers
a33943ec696ca80f84e4c64ca0aa96b1b0990281
1fe7cfab8e5c2dc5d5ca3b1f62ec5c44ae6b41ce
Fix an edge case for `get_encoder()` (#42295) * fix and edge case * ci must turn green * ci must turn green - second try
[ { "path": "src/transformers/modeling_utils.py", "patch": "@@ -2030,7 +2030,11 @@ def get_encoder(self, modality: Optional[str] = None):\n return getattr(self, name)\n \n if self.base_model is not self and hasattr(self.base_model, \"get_encoder\"):\n- return self.base_model...
2025-11-27T09:09:18
vuejs/vue
8a784d8d2333f0a05569f6c11c5a0fb0ab3a164e
6f6e5c88af95d4b50ba70e4d983e1add7f8fd254
fix(weex): stop trim css units in richtext component (#6927) + Remove the `trimCSSUnit` method. + Modify the test cases to support css units. + Add flow type annotations.
[ { "path": "src/platforms/weex/runtime/components/richtext.js", "patch": "@@ -1,55 +1,44 @@\n-function getVNodeType (vnode) {\n+/* @flow */\n+\n+function getVNodeType (vnode: VNode): string {\n if (!vnode.tag) {\n return ''\n }\n return vnode.tag.replace(/vue\\-component\\-(\\d+\\-)?/, '')\n }\n \n...
2017-10-31T00:28:54
huggingface/transformers
1fe7cfab8e5c2dc5d5ca3b1f62ec5c44ae6b41ce
5458d81d0f9995d85d9ff8a3c7b86de3b1808318
[fp8] fix scales param name (#42434) * fix * up * up
[ { "path": "src/transformers/integrations/finegrained_fp8.py", "patch": "@@ -409,14 +409,14 @@ def __init__(self, config, block_size, device):\n # gate_up tiles: ceil(Wg_out/bo) x ceil(Wg_in/bi)\n gu_scale_o = _ceil_div(Wg_out, bo)\n gu_scale_i = _ceil_div(Wg_in, bi)\n- ...
2025-11-27T09:09:05
denoland/deno
f6d5b957cb5d5ad398bd580e44c963588c5b36e3
984300a647b6a6f663c3c14708fa60b818205e82
perf(install): split tarball extraction into decompress + write, limit write concurrency (#32408) Split tarball extraction into a CPU-bound decompression phase and an I/O-bound filesystem write phase. Gate the write phase with a semaphore to reduce fs contention. This is on a create-next-app project. ``` ❯ hyperfine ...
[ { "path": "Cargo.lock", "patch": "@@ -2812,6 +2812,7 @@ dependencies = [\n \"tar\",\n \"tempfile\",\n \"thiserror 2.0.12\",\n+ \"tokio\",\n \"url\",\n ]\n ", "additions": 1, "deletions": 0, "language": "Unknown" }, { "path": "libs/npm_cache/Cargo.toml", "patch": "@@ -40,6 +40,7 @...
2026-03-03T09:23:38
vuejs/vue
e9c8e12397f1e89382f91b1557cc7f7d4139e651
d1d8b58ab6a0b0698a86aff716c0fd94b7cdd654
chore: fix sponsors from open collective (#6958) * Fix sponsors from open collective Also added Silver and Bronze sponsors. Fixes https://github.com/opencollective/opencollective/issues/579 * Update README.md only gold and above are listed here.
[ { "path": "README.md", "patch": "@@ -16,7 +16,7 @@\n Vue.js is an MIT-licensed open source project. It's an independent project with its ongoing development made possible entirely thanks to the support by these awesome [backers](https://github.com/vuejs/vue/blob/dev/BACKERS.md). If you'd like to join them, ...
2017-10-30T14:37:28
huggingface/transformers
5458d81d0f9995d85d9ff8a3c7b86de3b1808318
55b1400b74b7329f050f8e2bf7ca8534f2019295
Fix processor usage + add chat_template support to TTS pipeline, and shift common chat template logic to base class. (#42326) * Fix processor usage and add chat_template support to TTS pipeline. * Fallback to tokenizer for musicgen. * Fallback to tokenizer for musicgen. * Make style * style/quality after update? ...
[ { "path": "src/transformers/models/bark/modeling_bark.py", "patch": "@@ -651,8 +651,10 @@ def generate(\n ) # size: 10048\n \n # take the generated semantic tokens\n- semantic_output = semantic_output[:, max_input_semantic_length + 1 :]\n-\n+ if kwargs.get(\"return_dict_in_gen...
2025-11-27T09:04:39
ggml-org/llama.cpp
2fa51c19b028180b35d316e9ed06f5f0f7ada2c1
951520ddb05402bb8844509a7683d1a9a517dfc6
model-conversion : add token ids to prompt token output [no ci] (#17863) This commit adds the token ids to the printed prompt outputs. The motivation for this is that is can be useful to see the actual token ids alongside the token strings for debugging.
[ { "path": "examples/model-conversion/logits.cpp", "patch": "@@ -144,7 +144,7 @@ int main(int argc, char ** argv) {\n return 1;\n }\n std::string s(buf, n);\n- printf(\"%s\", s.c_str());\n+ printf(\"%s (%d)\", s.c_str(), id);\n }\n printf(\"\\n\");\n ", "...
2025-12-08T16:13:08
denoland/deno
984300a647b6a6f663c3c14708fa60b818205e82
f5f84cd58b7ae4602d82016a0e58cfeff010564e
fix(ext/node): CBC decipher setAutoPadding(false) incorrectly strips bytes (#32403) ## Summary - `Decipheriv.final()` unconditionally stripped PKCS7 padding from decrypted output regardless of the `#autoPadding` flag, corrupting data when `setAutoPadding(false)` was used - Made the PKCS7 padding stripping conditional...
[ { "path": "ext/node/polyfills/internal/crypto/cipher.ts", "patch": "@@ -485,7 +485,13 @@ export class Decipheriv extends Transform implements Cipher {\n throw new Error(\"Invalid final block size\");\n }\n \n- buf = buf.subarray(0, 16 - buf.at(-1)); // Padded in Pkcs7 mode\n+ if (this.#autoP...
2026-03-03T09:09:00
ollama/ollama
82b0c7c27e50c17b6fa5418caf56d32da28a9049
34d00f90b1cf9db0c59c953b3128f58d25073b36
Fix iGPU detection for linux This fixes a few bugs in the new sysfs discovery logic. iGPUs are now correctly identified by their <1G VRAM reported. the sysfs IDs are off by one compared to what HIP wants due to the CPU being reported in amdgpu, but HIP only cares about GPUs.
[ { "path": "gpu/amd_common.go", "patch": "@@ -40,19 +40,17 @@ func amdSetVisibleDevices(ids []int, skip map[int]interface{}) {\n \t// TODO - does sort order matter?\n \tdevices := []string{}\n \tfor i := range ids {\n-\t\tslog.Debug(fmt.Sprintf(\"i=%d\", i))\n \t\tif _, skipped := skip[i]; skipped {\n-\t\t\t...
2024-03-12T23:57:19
vuejs/vue
c3553196b8b15a71f982bd5e04c61be52e87c828
4ad9a56b229b156e633f3d575cd0e99ba5e474d9
fix: clean up target variables to avoid memory leaks (#6932) fix #6931
[ { "path": "src/core/instance/events.js", "patch": "@@ -19,7 +19,7 @@ export function initEvents (vm: Component) {\n }\n }\n \n-let target: Component\n+let target: any\n \n function add (event, fn, once) {\n if (once) {\n@@ -40,6 +40,7 @@ export function updateComponentListeners (\n ) {\n target = vm\n...
2017-10-27T01:08:10
ggml-org/llama.cpp
f896d2c34f7bb502c13986830b3ed7d85aac67d9
e4e9c4329c088d3aa97b8c242e18ff79bfe66248
server: improve speed of speculative decoding (#17808) * server: improve speed of speculative decoding * fix small draft case * add link to the PR * server : fix generation time measurement * server : fix draft acceptance logs (add SRV_CNT, SLT_CNT macros) * server : add comment * add PR to docs --------- Co-a...
[ { "path": "tools/server/README-dev.md", "patch": "@@ -81,6 +81,7 @@ For detailed instructions, see the [test documentation](./tests/README.md).\n - Separation of HTTP logic into dedicated files: https://github.com/ggml-org/llama.cpp/pull/17216\n - Large-scale code base split into smaller files: https://gith...
2025-12-08T13:35:28
denoland/deno
f5f84cd58b7ae4602d82016a0e58cfeff010564e
507336286804bbbc2b76e421e346563af9875253
fix: process level environment variables should take precedence over env files (#32407)
[ { "path": "cli/util/watch_env_tracker.rs", "patch": "@@ -114,6 +114,20 @@ impl WatchEnvTracker {\n let key_os = OsString::from(key);\n let value_os = OsString::from(value);\n \n+ // Process-level env vars should always take precedence over env files.\n+ ...
2026-03-03T09:05:49
vuejs/vue
4ad9a56b229b156e633f3d575cd0e99ba5e474d9
1cf02efda206185cb72bbaafb00037fa6269e3f3
fix(ssr): should not optimize root if conditions fix #6907
[ { "path": "src/server/optimizing-compiler/optimizer.js", "patch": "@@ -55,7 +55,7 @@ function walk (node: ASTNode, isRoot?: boolean) {\n if (node.ifConditions) {\n for (let i = 1, l = node.ifConditions.length; i < l; i++) {\n const block = node.ifConditions[i].block\n- walk(block)\n...
2017-10-26T16:43:57
ollama/ollama
53c107e20e183e3fd59f046275a6a95d6f6d4ddc
51578d857319a1e807e746dc7ab1b1f30a444958
chore: fix typo (#3073) Signed-off-by: racerole <jiangyifeng@outlook.com>
[ { "path": "llm/ext_server/ext_server.cpp", "patch": "@@ -115,7 +115,7 @@ void llama_server_init(ext_server_params *sparams, ext_server_resp_t *err) {\n llama_numa_init(params.numa);\n \n if (!llama->load_model(params)) { \n- // an error occured that was not thrown\n+ // an error occurred that wa...
2024-03-12T18:09:22
ggml-org/llama.cpp
636fc17a376dacc01da20d508e6986a299b1f819
51e0c2d917c21826585e84be1c27f75147325de0
Fix Kimi-K2 tool-call parsing issues (#17376) * Fix kimi-k2 parsing * fix template & add more tests for kimi-k2 * Another fix for Kimi-K2 chat template. * enable allow_toolcall_in_think for Kimi-K2 * Refine key-value separator and value end format * Enable tool call in think for kimi-k2 * allow_toolcall_in_think...
[ { "path": "common/chat-parser-xml-toolcall.cpp", "patch": "@@ -724,16 +724,10 @@ inline void parse_msg_with_xml_tool_calls(common_chat_msg_parser & builder, cons\n if (reasoning_unclosed) {\n if (auto pos = content.find(end_think); pos == std::string::npos && builder.pos() != builder.inp...
2025-12-08T13:32:04
huggingface/transformers
55b1400b74b7329f050f8e2bf7ca8534f2019295
1ae4d917ed3badbdb1ffc167e0529f5a6d3c080d
Any to any pipeline and auto-mapping (#40884) * initial commit * fix tests * fix copies, tests and rename pipe * another rename * fix copies again * activate pipeline mixin in some models * audio loading * typo * fix the test * stupid typo in filename * fix copies * docs * forgot * fix pipe tests * fix c...
[ { "path": "docs/source/en/_toctree.yml", "patch": "@@ -302,6 +302,8 @@\n title: Image tasks with IDEFICS\n - local: tasks/image_text_to_text\n title: Image-text-to-text\n+ - local: tasks/any_to_any\n+ title: Any-to-any\n - local: tasks/video_text_to_text\n tit...
2025-11-27T07:57:23
denoland/deno
507336286804bbbc2b76e421e346563af9875253
df9c814ffad815fb22f3be5565c6c70bdbb90385
fix(ext/node): implement KeyObject.equals, fix createCipheriv with KeyObject/null IV (#32409) ## Summary Builds on top of #32267 (must be merged first). Implements `KeyObject.prototype.equals()` and fixes `createCipheriv`/`createDecipheriv` to accept `KeyObject` keys and `null` IVs. Closes #27211 Closes #27295 Close...
[ { "path": "ext/node/polyfills/internal/crypto/cipher.ts", "patch": "@@ -22,6 +22,7 @@ import {\n op_node_decipheriv_decrypt,\n op_node_decipheriv_final,\n op_node_decipheriv_set_aad,\n+ op_node_export_secret_key,\n op_node_private_decrypt,\n op_node_private_encrypt,\n op_node_public_encrypt,\n@...
2026-03-03T09:00:38
vuejs/vue
1cf02efda206185cb72bbaafb00037fa6269e3f3
cfd73c2386623341fdbb3ac636c4baf84ea89c2c
fix: also clone component slot children during deepClone fix #6891, fix #6915
[ { "path": "src/core/vdom/patch.js", "patch": "@@ -673,9 +673,12 @@ export function createPatchFunction (backend) {\n // create an empty node and replace it\n oldVnode = emptyNodeAt(oldVnode)\n }\n+\n // replacing existing element\n const oldElm = oldVnode.elm\n ...
2017-10-26T16:00:19
ollama/ollama
51578d857319a1e807e746dc7ab1b1f30a444958
b5fcd9d3aa6b05fe09f0a604bde9314d5d8d27ea
fix gpu_info_cuda.c compile warning (#3077)
[ { "path": "gpu/gpu_info_cuda.c", "patch": "@@ -155,8 +155,8 @@ void cuda_check_vram(cuda_handle_t h, mem_info_t *resp) {\n }\n }\n \n- LOG(h.verbose, \"[%d] CUDA totalMem %ld\\n\", i, memInfo.total);\n- LOG(h.verbose, \"[%d] CUDA usedMem %ld\\n\", i, memInfo.used);\n+ LOG(h.verbose, \"[%d...
2024-03-12T18:08:40
ggml-org/llama.cpp
5814b4dce18f9c5cbebef175e381a7b0ff147d72
79d61896d35f37b79f432ae935698c5459ba8a41
cuda: optimize SOLVE_TRI using registers and FMAF (#17703) * ggml-cuda: optimize solve_tri_f32_fast and fix stride handling - Switch from using shared memory for the RHS/solution matrix to a register-based approach (x_low, x_high), reducing shared memory pressure and bank conflicts. - Implement explicit `fmaf` instr...
[ { "path": "ggml/src/ggml-cuda/solve_tri.cu", "patch": "@@ -3,7 +3,6 @@\n #include \"solve_tri.cuh\"\n \n #define MAX_N_FAST 64\n-#define MAX_K_FAST 32\n \n // ======================\n // Fast Kernel (n <= 64, k <= 32) - Warp-based parallel reduction\n@@ -48,65 +47,58 @@ static __global__ void solve_tri_f32_...
2025-12-08T09:41:08
huggingface/transformers
1ae4d917ed3badbdb1ffc167e0529f5a6d3c080d
06d52fe106dc6922815dc25bfa156a9d78b16959
Clean bnb integration using weight converter (#42426) * clean bnb * style * let's go * style * torch * Apply style fixes --------- Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
[ { "path": "src/transformers/core_model_loading.py", "patch": "@@ -674,9 +674,6 @@ def convert_and_load_state_dict_in_model(\n \n renamings = [entry for entry in weight_mapping if isinstance(entry, WeightRenaming)]\n converters = [entry for entry in weight_mapping if isinstance(entry, WeightConverter...
2025-11-26T20:28:07
denoland/deno
df9c814ffad815fb22f3be5565c6c70bdbb90385
7f6915423cd8b642ffc8d4fecf4f64ff0b6dd7f3
fix(compile): store self extracting dir in hidden dir beside executable (#32329)
[ { "path": "cli/rt/binary.rs", "patch": "@@ -179,7 +179,7 @@ fn choose_and_create_extraction_dir(\n \n // try next to the executable first\n if let Some(exe_dir) = current_exe.parent() {\n- let dir = exe_dir.join(format!(\"{}.fs\", exe_name)).join(hash_str);\n+ let dir = exe_dir.join(format!(\".{}\...
2026-03-03T08:50:08
vuejs/vue
cfd73c2386623341fdbb3ac636c4baf84ea89c2c
83ed92608d81349e1cac2e481ed079e51a490b2b
fix: handle encoded tabs and newlines in attributes for Chrome a[href] and IE/Edge fix #6828, fix #6916
[ { "path": "flow/compiler.js", "patch": "@@ -14,7 +14,8 @@ declare type CompilerOptions = {\n preserveWhitespace?: boolean;\n isFromDOM?: boolean;\n shouldDecodeTags?: boolean;\n- shouldDecodeNewlines?: boolean;\n+ shouldDecodeNewlines?: boolean;\n+ shouldDecodeNewlinesForHref?: boolean;\n \n // ...
2017-10-26T14:01:12
ollama/ollama
b80661e8c78e115ed9b41391c87fdb7f1a7f69ec
6d3adfbea21699db7770eb608264d5e16b8663ee
relay load model errors to the client (#3065)
[ { "path": "llm/dyn_ext_server.go", "patch": "@@ -149,7 +149,7 @@ func newDynExtServer(library, model string, adapters, projectors []string, opts\n \n \tslog.Info(\"Initializing llama server\")\n \tslog.Debug(fmt.Sprintf(\"server params: %+v\", sparams))\n-\tinitResp := newExtServerResp(128)\n+\tinitResp := ...
2024-03-11T20:48:27
denoland/deno
7f6915423cd8b642ffc8d4fecf4f64ff0b6dd7f3
f7bcc9694483e4869d39c318aa25dfdb90ddd912
fix(ext/crypto): handle wrong algorithm in subtle.importKey without panicking (#32410) ## Summary - Fixes #30243. Passing a valid PKCS#8 DER key with the wrong algorithm (e.g., an RSA key imported as ECDSA) to `crypto.subtle.importKey` caused a panic due to an `.unwrap()` on a `try_into()` conversion in `import_key_ec...
[ { "path": "ext/crypto/import_key.rs", "patch": "@@ -686,7 +686,7 @@ fn import_key_ec(\n .parameters\n .ok_or(ImportKeyError::MalformedParameters)?\n .try_into()\n- .unwrap();\n+ .map_err(|_| ImportKeyError::MalformedParameters)?;\n \n let pk_named_curve = match na...
2026-03-03T08:33:14
vuejs/vue
83ed92608d81349e1cac2e481ed079e51a490b2b
ee0e8b5ce1d64120b2b8d6426d783f6f0a683395
fix: $set should respect properties on prototype chain fix #6845
[ { "path": "src/core/observer/index.js", "patch": "@@ -196,7 +196,7 @@ export function set (target: Array<any> | Object, key: any, val: any): any {\n target.splice(key, 1, val)\n return val\n }\n- if (hasOwn(target, key)) {\n+ if (key in target && !(key in Object.prototype)) {\n target[key] =...
2017-10-26T11:52:36
ggml-org/llama.cpp
08f9d3cc1d40169082ccf2416ebadb8a5afca9d9
0a540f9abd98915edb99fed47d80078ed8d2f343
Vulkan: improve mul_mat_vec_iq1_m (#16907) * Optimize Vulkan shader for matrix-vector multiplication * Revert changes on compute_outputs and main Refactor compute_outputs to handle remaining rows correctly. * Fix trailing whitespace
[ { "path": "ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq1_m.comp", "patch": "@@ -7,35 +7,85 @@ layout(local_size_x_id = 0, local_size_y = 1, local_size_z = 1) in;\n \n FLOAT_TYPE temp[NUM_COLS][NUM_ROWS];\n \n-void calc_superblock(const uint a_offset, const uint b_offset, const uint ib32, const uint i,...
2025-12-07T17:40:42
huggingface/transformers
06d52fe106dc6922815dc25bfa156a9d78b16959
d2825a0219113897eef6487a7bdfd3012388ff3d
Simplify and standardize processor tests (#41773) * remove attributes and add all missing sub processors to their auto classes * remove all mentions of .attributes * cleanup * fix processor tests * fix modular * remove last attributes * fixup * fixes after merge * fix wrong tokenizer in auto florence2 * fix m...
[ { "path": "docs/source/en/model_doc/layoutxlm.md", "patch": "@@ -70,6 +70,12 @@ data for the model.\n As LayoutXLM's architecture is equivalent to that of LayoutLMv2, one can refer to [LayoutLMv2's documentation page](layoutlmv2) for all tips, code examples and notebooks.\n </Tip>\n \n+\n+## LayoutXLMConfig...
2025-11-26T17:40:37
denoland/deno
f7bcc9694483e4869d39c318aa25dfdb90ddd912
9becb837c6a425124d4423985a415c5c7404e706
fix(ext/node): support Ed25519, X25519, and P-521 keys in X509Certificate.publicKey (#32267) ## Summary - Fixes `X509Certificate.prototype.publicKey` throwing "unsupported x509 public key type" for **Ed25519** and **X25519** certificates - Fixes `X509Certificate.prototype.publicKey` throwing "unsupported ec named cur...
[ { "path": "Cargo.lock", "patch": "@@ -2724,6 +2724,7 @@ dependencies = [\n \"p224\",\n \"p256\",\n \"p384\",\n+ \"p521\",\n \"pbkdf2\",\n \"pkcs8\",\n \"rand 0.8.5\",", "additions": 1, "deletions": 0, "language": "Unknown" }, { "path": "Cargo.toml", "patch": "@@ -398,7 +398,7 @...
2026-03-02T16:37:47
vuejs/vue
98ea0a3b48e37719f278c10a8ee5fb94d7d5db4e
604230fe953f864be5dc70bd7d34f64ae43e4f7e
fix(types): improve typing for better completion (#6886)
[ { "path": "types/vue.d.ts", "patch": "@@ -61,7 +61,7 @@ export interface Vue {\n $createElement: CreateElement;\n }\n \n-export type CombinedVueInstance<Instance extends Vue, Data, Methods, Computed, Props> = Instance & Data & Methods & Computed & Props;\n+export type CombinedVueInstance<Instance extends ...
2017-10-25T04:45:15
ggml-org/llama.cpp
d9e03db1e701e34ed0b764615025110041729864
db97837385edfbc772230debbd49e5efae843a71
sycl: add missing BF16 conversion support for Intel oneAPI (#17780) * sycl: add missing BF16 conversion support for Intel oneAPI * Fix Line 645: Trailing whitespace
[ { "path": "ggml/src/ggml-sycl/convert.cpp", "patch": "@@ -2,6 +2,13 @@\n #include \"dequantize.hpp\"\n #include \"presets.hpp\"\n \n+#if defined(__INTEL_LLVM_COMPILER)\n+ #if __has_include(<sycl/ext/oneapi/bfloat16.hpp>)\n+ #include <sycl/ext/oneapi/bfloat16.hpp>\n+ #define GGML_SYCL_HAS_BF...
2025-12-07T01:18:18
ollama/ollama
9ea492f1ce255e81b08f25b781b3104bebd5765e
ac64cd4ef97de04333ff21185aa730082cfbc67f
convert: fix shape
[ { "path": "convert/convert.go", "patch": "@@ -103,9 +103,9 @@ func ReadSafeTensors(fn string, offset uint64) ([]llm.Tensor, uint64, error) {\n \t\t\treturn []llm.Tensor{}, 0, err\n \t\t}\n \n-\t\tshape := [4]uint64{1, 1, 1, 1}\n-\t\tfor cnt, s := range data.Shape {\n-\t\t\tshape[cnt] = uint64(s)\n+\t\tshape...
2024-03-10T17:41:40
huggingface/transformers
d2825a0219113897eef6487a7bdfd3012388ff3d
0a1d5ad3be78897bccd2cc778d9cd036843c5d5c
Bart config doesn't need generation parameters (#42337) * fix bart and mvp * nits and fixes * style --------- Co-authored-by: vasqu <antonprogamer@gmail.com>
[ { "path": "src/transformers/generation/configuration_utils.py", "patch": "@@ -436,6 +436,13 @@ def __init__(self, **kwargs):\n self._commit_hash = kwargs.pop(\"_commit_hash\", None)\n self.transformers_version = kwargs.pop(\"transformers_version\", __version__)\n \n+ # Ensure backward...
2025-11-26T16:33:19
vuejs/vue
604230fe953f864be5dc70bd7d34f64ae43e4f7e
048e9407795ec7420d4b2962cbd50f769d3432fb
fix(keep-alive): higher priority for exclude than include (#6905)
[ { "path": "src/core/components/keep-alive.js", "patch": "@@ -87,8 +87,8 @@ export default {\n // check pattern\n const name: ?string = getComponentName(componentOptions)\n if (name && (\n- (this.include && !matches(this.include, name)) ||\n- (this.exclude && matches(this.excl...
2017-10-25T04:44:54
denoland/deno
9becb837c6a425124d4423985a415c5c7404e706
5fb8a4d498c07e0bccef1f5fcfade14501ba45a3
fix(node/crypto): handle empty input in decipher final with auto_pad=false (#31958) ## Summary Fixes a panic in `node:crypto` when using `createDecipheriv` with `setAutoPadding(false)` (denoland/deno#31957). Two cases caused panics in `DecipherContext::final`: 1. **Empty input** — `GenericArray::from_slice` panicked...
[ { "path": "ext/node_crypto/cipher.rs", "patch": "@@ -752,10 +752,13 @@ impl Decipher {\n Ok(())\n }\n (Aes128Cbc(mut decryptor), false) => {\n- decryptor.decrypt_block_b2b_mut(\n- GenericArray::from_slice(input),\n- GenericArray::from_mut_slice(output),\n- ...
2026-03-02T16:33:46
ggml-org/llama.cpp
db97837385edfbc772230debbd49e5efae843a71
017761daf5f543e3f064d956784a9c9421dc0b66
vulkan: perf_logger improvements (#17672) * vulkan: perf_logger improvements - Move perf_logger from device to ctx. - Add an env var to control the frequency we dump the stats. If you set a very large value, it just dumps when the ctx is destroyed. - Add a fusion info string to the tracking, only log one item per fus...
[ { "path": "ggml/src/ggml-vulkan/ggml-vulkan.cpp", "patch": "@@ -777,11 +777,6 @@ struct vk_device_struct {\n std::unique_ptr<vk_memory_logger> memory_logger;\n #endif\n \n- // for GGML_VK_PERF_LOGGER\n- std::unique_ptr<vk_perf_logger> perf_logger;\n- vk::QueryPool query_pool;\n- int32_t num_...
2025-12-06T17:46:46
ollama/ollama
41b00b9856b6708e5af165064c6ee0e556f600ab
c2a8ed48e7f77dc23682d0615f8eb87af88fef7c
fix `03-locale.diff`
[ { "path": "llm/patches/03-locale.diff", "patch": "@@ -4,7 +4,7 @@ index b19616e8..519b9602 100644\n +++ b/llama.cpp\n @@ -9938,7 +9938,7 @@ struct llm_tokenizer_wpm {\n }\n-\n+ \n uint32_t to_lower(uint32_t code) {\n - static const std::locale locale(\"en_US.UTF-8\");\n + static cons...
2024-03-10T23:21:05
vuejs/vue
b6c384dd78b56bd247e6a34d5aea0d3903f5b7fd
9caed00d20f37c750e39db4ec86d278b453f0e5d
fix(core): static trees should be cached on options (#6826) (#6837) * test(once): failing test for #6826 * fix(core): static trees should be cached on options (#6826)
[ { "path": "src/core/instance/render-helpers/render-static.js", "patch": "@@ -11,8 +11,8 @@ export function renderStatic (\n ): VNode | Array<VNode> {\n // static trees can be rendered once and cached on the contructor options\n // so every instance shares the same cached trees\n- const renderFns = this...
2017-10-18T04:26:55
ollama/ollama
f9cd55c70b28944531361d595147a59d5f5b5c3f
ac64cd4ef97de04333ff21185aa730082cfbc67f
disable gpu for certain model architectures and fix divide-by-zero on memory estimation
[ { "path": "llm/llm.go", "patch": "@@ -6,6 +6,7 @@ import (\n \t\"log/slog\"\n \t\"os\"\n \t\"runtime\"\n+\t\"slices\"\n \n \t\"github.com/jmorganca/ollama/api\"\n \t\"github.com/jmorganca/ollama/gpu\"\n@@ -19,6 +20,10 @@ type LLM interface {\n \tClose()\n }\n \n+var cpuOnlyFamilies = []string{\n+\t\"mamba\"...
2024-03-09T20:51:38
denoland/deno
5fb8a4d498c07e0bccef1f5fcfade14501ba45a3
968baf725a6d28cd739fd61b1f447fe3de89b622
fix(npm): peerDependency resolution leading to multiple versions being installed + hanging (#32358) Essentially, this PR rewrites our dependency resolution code to use a two-phase approach, which fixes duplicate peer dependencies or hanging installation. Peer deps were resolved inline during BFS traversal, making res...
[ { "path": "tests/specs/npm/info_peer_deps/main_info.out", "patch": "@@ -4,11 +4,11 @@ dependencies: 6 unique\n size: [WILDCARD]\n \n file:///[WILDCARD]/main.ts (171B)\n-├─┬ npm:/@denotest/peer-dep-test-child@1.0.0_@denotest+peer-dep-test-peer@1.0.0 ([WILDCARD])\n+├─┬ npm:/@denotest/peer-dep-test-child@1.0.0...
2026-03-02T15:33:10
ggml-org/llama.cpp
017761daf5f543e3f064d956784a9c9421dc0b66
c42712b056fb2cf03902ef57fd314c531e356965
ggml-zendnn : add ZenDNN backend for AMD CPUs (#17690) * ggml-zennn: add ZenDNN backend support * ggml-zendnn : address ZenDNN backend review fixes and suggestions * docs : apply blockquote syntax to ZenDNN docs --------- Co-authored-by: Manoj Kumar <mkumar@zettabolt.com>
[ { "path": "README.md", "patch": "@@ -276,6 +276,7 @@ Instructions for adding support for new models: [HOWTO-add-model.md](docs/develo\n | [MUSA](docs/build.md#musa) | Moore Threads GPU |\n | [CUDA](docs/build.md#cuda) | Nvidia GPU |\n | [HIP](docs/build.md#hip) | AMD GPU |\n+| [ZenDNN](docs/build.md#zendnn)...
2025-12-06T16:13:33
denoland/deno
968baf725a6d28cd739fd61b1f447fe3de89b622
826878191feedbf83147bbf6726a8e2984704545
fix(ext/node): support DER-encoded keys in publicEncrypt/privateDecrypt (#32291) ## Summary - `publicEncrypt`, `privateDecrypt`, and `privateEncrypt` only accepted PEM-encoded (text) keys. When passed DER (binary) encoded keys, they failed with `invalid utf-8 sequence of 1 bytes from index 1` because the code tried t...
[ { "path": "ext/node_crypto/lib.rs", "patch": "@@ -33,6 +33,8 @@ use rsa::Oaep;\n use rsa::Pkcs1v15Encrypt;\n use rsa::RsaPrivateKey;\n use rsa::RsaPublicKey;\n+use rsa::pkcs1::DecodeRsaPrivateKey;\n+use rsa::pkcs1::DecodeRsaPublicKey;\n use rsa::pkcs8::DecodePrivateKey;\n use rsa::pkcs8::DecodePublicKey;\n ...
2026-03-02T15:17:41
vuejs/vue
9caed00d20f37c750e39db4ec86d278b453f0e5d
471de4a31d229e681cc9dce18632b5bcab944c77
fix(typing): relax $options type for TS2.6+ (#6819)
[ { "path": "types/vue.d.ts", "patch": "@@ -22,7 +22,7 @@ export interface CreateElement {\n \n export interface Vue {\n readonly $el: HTMLElement;\n- readonly $options: ComponentOptions<this>;\n+ readonly $options: ComponentOptions<Vue>;\n readonly $parent: Vue;\n readonly $root: Vue;\n readonly $c...
2017-10-15T18:26:08
ollama/ollama
18979ad4a1d40d04e3b981a477fa6323a40304b6
8e0ef931d893c225689253a19b4915408e674561
convert: fix default shape
[ { "path": "convert/convert.go", "patch": "@@ -103,7 +103,7 @@ func ReadSafeTensors(fn string, offset uint64) ([]llm.Tensor, uint64, error) {\n \t\t\treturn []llm.Tensor{}, 0, err\n \t\t}\n \n-\t\tshape := [4]uint64{0, 0, 0, 0}\n+\t\tshape := [4]uint64{1, 1, 1, 1}\n \t\tfor cnt, s := range data.Shape {\n \t\...
2024-03-08T23:40:16
ggml-org/llama.cpp
c42712b056fb2cf03902ef57fd314c531e356965
09c7c50e64c98adac452d090406b6e5f6c320a41
server: support multiple generations from one prompt (OAI "n" option) (#17775) * backend support * server: support multiple generations from one prompt (OAI "n" option) * fix invalid batch * format oai * clean up * disable ctx shift * add test * update comments * fix style * add n_cmpl to docs [no ci] * allo...
[ { "path": "tools/server/README.md", "patch": "@@ -493,6 +493,8 @@ Note for `multimodal_data` in JSON object prompts. This should be an array of st\n `n_keep`: Specify the number of tokens from the prompt to retain when the context size is exceeded and tokens need to be discarded. The number excludes the BOS...
2025-12-06T14:54:38
huggingface/transformers
96c41be562d43785d07744924c2f1e96bc7d6884
f77950605e2f2f27613b829559b133c2b9fd8912
fixed json syntax error for zero2 configuration file found in deepspeed.md (#42406)
[ { "path": "docs/source/en/deepspeed.md", "patch": "@@ -196,7 +196,7 @@ ZeRO-2 shards the optimizer and gradient states across GPUs. This stage is prima\n \"overlap_comm\": true,\n \"reduce_scatter\": true,\n \"reduce_bucket_size\": 5e8,\n- \"contiguous_gradients\": true\n+ ...
2025-11-26T14:32:52
vuejs/vue
a11d6f37c2cededd2b3ec4c74942815bf16e34e0
4e0c48511d49f331fde31fc87b6ca428330f32d1
chore: fix flow
[ { "path": "src/core/util/next-tick.js", "patch": "@@ -86,7 +86,7 @@ export function withMacroTask (fn: Function): Function {\n })\n }\n \n-export function nextTick (cb?: Function, ctx?: Object): ?Promise {\n+export function nextTick (cb?: Function, ctx?: Object) {\n let _resolve\n callbacks.push(() =>...
2017-10-13T19:56:01
denoland/deno
8c7409882b13fa20ca1a9739d6e4605c2d464241
9a246baac9cb8c2bb4422e18fcea8b1fd880d7ef
fix(test): dedupe discovered workspace test modules (#32380) This PR should address #25949. I fixed duplicate test module execution caused by overlapping discovery entries (e.g., workspace + `deno test .`). IMHO this is a suboptimal but safer fix. The cleaner solution would be to make discovery non-overlapping at me...
[ { "path": "cli/tools/test/mod.rs", "patch": "@@ -129,6 +129,10 @@ pub enum TestMode {\n }\n \n impl TestMode {\n+ fn union(self, other: Self) -> Self {\n+ if self == other { self } else { Self::Both }\n+ }\n+\n /// Returns `true` if the test mode indicates that code snippet extraction is\n /// need...
2026-03-02T14:36:29
ggml-org/llama.cpp
09c7c50e64c98adac452d090406b6e5f6c320a41
f334b79494008e8eac6b41edbb62639226394d82
ggml : add circular tiling support to pad, for Vulkan, CUDA, and CPU (used for making seamless textures) (#16985) * Feat: Added vulkan circular tiling support * Feat: Added cpu circular * Feat: Added cuda kernels * Added tests * Added tests * Removed non-pad operations * Removed unneded changes * removed backen...
[ { "path": "ggml/include/ggml.h", "patch": "@@ -2196,6 +2196,15 @@ extern \"C\" {\n int p2,\n int p3);\n \n+ // pad each dimension with values on the other side of the torus (looping around)\n+ GGML_API struct ggml_tensor * ggml_pad_circular(\n+...
2025-12-06T14:07:02
vuejs/vue
4e0c48511d49f331fde31fc87b6ca428330f32d1
6658b81a3b9e9d1d4285b4b985cddcdee930dae8
fix: further adjust nextTick strategy fix #6813
[ { "path": "src/core/util/env.js", "patch": "@@ -1,7 +1,4 @@\n /* @flow */\n-/* globals MessageChannel */\n-\n-import { handleError } from './error'\n \n // can we use __proto__?\n export const hasProto = '__proto__' in {}\n@@ -62,88 +59,6 @@ export const hasSymbol =\n typeof Symbol !== 'undefined' && isNa...
2017-10-13T19:37:52
huggingface/transformers
cb739f8610959c25faa798a0b6030ea930dcfe00
a95d9975c8e9ba6366bb13e0b75d133c0cc49d98
[core] fix mxfp4 (#42382) * initial commit * fix import * fix * add ops * style * decouple dequantize & deserialize logic * up
[ { "path": "src/transformers/integrations/mxfp4.py", "patch": "@@ -18,13 +18,19 @@\n if is_torch_available():\n import torch\n from torch import nn\n+from typing import Optional\n+\n+from ..core_model_loading import ConversionOps\n+\n \n if is_accelerate_available():\n from accelerate import init...
2025-11-26T11:16:33
ollama/ollama
0cebc79cba778d3d504405c3973dd9f2507b1ab0
0e4669b04f1553e46221d8e8e789a91b664015da
fix: allow importing a model from name reference (#3005)
[ { "path": "server/images.go", "patch": "@@ -10,6 +10,7 @@ import (\n \t\"errors\"\n \t\"fmt\"\n \t\"io\"\n+\t\"io/fs\"\n \t\"log\"\n \t\"log/slog\"\n \t\"net/http\"\n@@ -322,9 +323,12 @@ func CreateModel(ctx context.Context, name, modelFileDir string, commands []pars\n \n \t\t\tggufName, err := convertSafet...
2024-03-08T17:27:47
denoland/deno
9a246baac9cb8c2bb4422e18fcea8b1fd880d7ef
e7269534ae4fe4d08258d562738d0cb1aa275733
fix: fix some stack frame dimming, and make the console log more aligned with cli formatting (#32399)
[ { "path": "ext/web/01_console.js", "patch": "@@ -369,6 +369,12 @@ const kArrayExtrasType = 2;\n const coreModuleRegExp = new SafeRegExp(\n /^ {4}at (?:[^/\\\\(]+ \\(|)node:(.+):\\d+:\\d+\\)?$/,\n );\n+const extModuleRegExp = new SafeRegExp(\n+ /^ {4}at (?:[^/\\\\(]+ \\(|)ext:.+:\\d+:\\d+\\)?$/,\n+);\n+co...
2026-03-02T14:21:46
ggml-org/llama.cpp
f334b79494008e8eac6b41edbb62639226394d82
a28e3c7567ae2f064d663db7019ca2d5da73df55
HIP: fix RDNA3 FP16/BF16 matrix multiplication (#17817)
[ { "path": "ggml/src/ggml-cuda/mmf.cu", "patch": "@@ -160,9 +160,9 @@ bool ggml_cuda_should_use_mmf(enum ggml_type type, int cc, int warp_size, const\n case GGML_TYPE_F32:\n return ampere_mma_available(cc);\n case GGML_TYPE_F16:\n- return volta_mma_available(cc) || turi...
2025-12-06T12:45:36
huggingface/transformers
a95d9975c8e9ba6366bb13e0b75d133c0cc49d98
9c1082af99e1b290f9e17199ca0a9291dcbb0a1a
Many small fixes for the CI (#42364)
[ { "path": "src/transformers/utils/generic.py", "patch": "@@ -49,6 +49,12 @@\n _is_torch_available = True\n \n \n+# required for @can_return_tuple decorator to work with torchdynamo\n+_is_mlx_available = False\n+if is_mlx_available():\n+ _is_mlx_available = True\n+\n+\n # vendored from distutils.util\...
2025-11-26T10:37:38
denoland/deno
e7269534ae4fe4d08258d562738d0cb1aa275733
70497bbf6a7e673b7f9f89b7a112c488fd20e11d
fix(ext/node): make `execPath` named export a real string (#32396) ## Summary - The named ESM export `execPath` from `node:process` was a frozen object with `String.prototype` that pretended to be a string but had `typeof === "object"`. This broke libraries like execa that check `typeof execPath === "string"`, causin...
[ { "path": "ext/node/polyfills/process.ts", "patch": "@@ -480,20 +480,7 @@ function uncaughtExceptionHandler(err: any, origin: string) {\n process.emit(\"uncaughtException\", err, origin);\n }\n \n-export let execPath: string = Object.freeze({\n- __proto__: String.prototype,\n- toString() {\n- execPat...
2026-03-02T14:15:58
ggml-org/llama.cpp
e31b5c55c3dbbf5c8e47c8dcb17ba92250f82303
21f24f27a9bd58e35ba2ba9a3366bfc660b1f095
webui: Fix context available value in Multi-model Router mode (#17804) * fix: Use context size from `/props?model=...` in ROUTER mode * chore: update webui build output
[ { "path": "tools/server/webui/src/lib/stores/chat.svelte.ts", "patch": "@@ -2,7 +2,11 @@ import { DatabaseService, ChatService } from '$lib/services';\n import { conversationsStore } from '$lib/stores/conversations.svelte';\n import { config } from '$lib/stores/settings.svelte';\n import { contextSize, isRo...
2025-12-06T12:23:29
huggingface/transformers
c12dfddba94c04fb6a5f5be4be879af51e354e2f
47e795b874b5393802d5bb6ab17bda45183ef711
[SAM3] Fix precompute vision_embeds or text_embeds for inference (#42407) Fix precompute vision embeds or text_embeds for inference
[ { "path": "docs/source/en/model_doc/sam3.md", "patch": "@@ -256,6 +256,104 @@ SAM3 also provides semantic segmentation alongside instance masks:\n >>> print(f\"Semantic segmentation: {semantic_seg.shape}\")\n ```\n \n+### Efficient Multi-Prompt Inference on Single Image\n+\n+When running multiple text promp...
2025-11-25T23:06:40
denoland/deno
70497bbf6a7e673b7f9f89b7a112c488fd20e11d
fd9def342d39cb660de815f760745ea0aacda4e5
fix(ext/node): make `setAutoPadding(false)` a no-op for GCM ciphers (#32290) ## Summary - GCM is a stream cipher mode that doesn't use block padding. In Node.js, `setAutoPadding()` is a no-op for GCM modes, but Deno was throwing `setAutoPadding(false) not supported for Aes128Gcm/Aes256Gcm yet` when called on GCM deci...
[ { "path": "ext/node_crypto/cipher.rs", "patch": "@@ -505,12 +505,6 @@ pub enum DecipherError {\n #[error(\"Failed to authenticate data\")]\n DataAuthenticationFailed,\n #[class(type)]\n- #[error(\"setAutoPadding(false) not supported for Aes128Gcm yet\")]\n- SetAutoPaddingFalseAes128GcmUnsupported,\n...
2026-03-02T13:54:58
ollama/ollama
2cb74e23fb709f83cec1f8bccb79c13c11a428b4
3c8df3808b9e1f50dd61552ea986cf74f44ec5ad
fix ci
[ { "path": ".github/workflows/test.yaml", "patch": "@@ -3,6 +3,7 @@ name: test\n on:\n pull_request:\n paths:\n+ - '**/*'\n - '!docs/**'\n - '!examples/**'\n - '!README.md'", "additions": 1, "deletions": 0, "language": "YAML" } ]
2024-03-07T19:33:49
ggml-org/llama.cpp
7b43f5575399b18c7847efb665a3c1f16d5fcc2c
444f00b0ec814a071ce1b9dc0de5ea4b4850bd1b
ggml : improve error handling for search path existence checks (#17653) * Improve error handling for search path existence checks Refactor existence checks for search paths using std::error_code to handle potential errors. * Improve cache file existence check with error code Update fs::exists to use std::error_cod...
[ { "path": "ggml/src/ggml-backend-reg.cpp", "patch": "@@ -534,8 +534,12 @@ static ggml_backend_reg_t ggml_backend_load_best(const char * name, bool silent,\n fs::path best_path;\n \n for (const auto & search_path : search_paths) {\n- if (!fs::exists(search_path)) {\n- GGML_LOG_DEBUG...
2025-12-06T11:28:16
vuejs/vue
08a7fb539f9d3ab5b08a3c6cec9a6628929be3be
5665eaf985a56cfd183ce8ce93c4d813edbd2cf8
fix: v-on="object" listeners should fire after high-priority ones fix #6805
[ { "path": "src/core/instance/render-helpers/bind-object-listeners.js", "patch": "@@ -14,7 +14,7 @@ export function bindObjectListeners (data: any, value: any): VNodeData {\n for (const key in value) {\n const existing = on[key]\n const ours = value[key]\n- on[key] = existing ? [...
2017-10-13T13:59:23
huggingface/transformers
b6055550a15a8fab367cf983b743ff68cc58d81a
1ce12ac5de2d5d082825c7fab9d03ab5d110e400
[`Mistral Tokenizers`] Fix tokenizer detection (#42389) * fix * sanity check * style * comments * make it v5 explicit * make explicit fixes possible in local tokenizers * remove hub usage on local * fix * extend test for no config case * move mistral patch outside to separate fn * fix local path only * add ...
[ { "path": "src/transformers/tokenization_utils_base.py", "patch": "@@ -2099,12 +2099,13 @@ def from_pretrained(\n template = template.removesuffix(\".jinja\")\n vocab_files[f\"chat_template_{template}\"] = f\"{CHAT_TEMPLATE_DIR}/{template}.jinja\"\n \n...
2025-11-25T15:32:41
denoland/deno
fd9def342d39cb660de815f760745ea0aacda4e5
84f7a15a880d296faa7a71b108579159952197f8
fix(node): support ECDSA with secp256k1 in `node:crypto` (#32390) ## Summary - Adds secp256k1 ECDSA support to `node:crypto` key generation, signing, verification, and key import/export (DER, PEM, JWK) - Adds `Secp256k1` variant to `EcPrivateKey`/`EcPublicKey` enums using the existing `k256` crate - Enables `ecdh`, `...
[ { "path": "Cargo.toml", "patch": "@@ -391,7 +391,7 @@ ed25519-dalek = \"2.1.1\"\n ed448-goldilocks = \"0.8.3\"\n elliptic-curve = { version = \"0.13.4\", features = [\"alloc\", \"arithmetic\", \"ecdh\", \"std\", \"pem\", \"jwk\"] }\n hkdf = \"0.12.3\"\n-k256 = \"0.13.1\"\n+k256 = { version = \"0.13.1\", fea...
2026-03-02T13:21:03
ggml-org/llama.cpp
dbc15a79672e72e0b9c1832adddf3334f5c9229c
c6c5e859798163c2e41d848d1157438467a2a34a
convert: support Mistral 3 Large MoE (#17730) * convert: support Mistral 3 Large MoE * filter out vision tensors, add missing keys * handle vocab * add temperature_length * fix mscale_all_dim * clean up * Apply suggestions from code review Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com> * fix *...
[ { "path": "convert_hf_to_gguf.py", "patch": "@@ -1524,6 +1524,79 @@ def _set_vocab_interns1(self):\n special_vocab._set_special_token(\"bos\", 151643)\n special_vocab.add_to_gguf(self.gguf_writer)\n \n+ def _set_vocab_mistral(self):\n+ if not _mistral_common_installed:\n+ ...
2025-12-06T09:49:33
vuejs/vue
5665eaf985a56cfd183ce8ce93c4d813edbd2cf8
60b1af9e02b93d9223d2ed1f23e0a618537a4c96
fix: backwards compat with checkbox code generated in < 2.5 fix #6803
[ { "path": "src/platforms/web/compiler/directives/model.js", "patch": "@@ -9,6 +9,7 @@ let warn\n // in some cases, the event used has to be determined at runtime\n // so we used some reserved tokens during compile.\n export const RANGE_TOKEN = '__r'\n+export const CHECKBOX_RADIO_TOKEN = '__c'\n \n export de...
2017-10-13T13:41:52
ollama/ollama
6c5ccb11f993ccc88c4761b8c31e0fefcbc1900f
2e20110e50248080df0196a1b10095e5caf55f69
Revamp ROCm support This refines where we extract the LLM libraries to by adding a new OLLAMA_HOME env var, that defaults to `~/.ollama` The logic was already idempotenent, so this should speed up startups after the first time a new release is deployed. It also cleans up after itself. We now build only a single ROCm...
[ { "path": ".github/workflows/test.yaml", "patch": "@@ -7,12 +7,12 @@ jobs:\n generate:\n strategy:\n matrix:\n- os: [ubuntu-latest, macos-latest, windows-latest]\n+ os: [ubuntu-latest, macos-latest, windows-2019]\n arch: [amd64, arm64]\n exclude:\n - os: u...
2024-02-16T01:15:09
huggingface/transformers
f13b1009c2e2a89cac919584f7b67bfe55d4e226
45d273d52c2c9ecbebfd213a54ededd0b609b13c
Xcodec fix (#42095) * nit on dac! * fix * not for this pr * make style
[ { "path": "src/transformers/audio_utils.py", "patch": "@@ -219,6 +219,18 @@ def load_audio_as(\n raise ValueError(f\"Error loading audio: {e}\")\n \n \n+def conv1d_output_length(module: \"torch.nn.Conv1d\", input_length: int) -> int:\n+ \"\"\"\n+ Computes the output length of a 1D convolution ...
2025-11-25T14:39:01
denoland/deno
84f7a15a880d296faa7a71b108579159952197f8
4f22dcc1a0fbbd38474bd3774202ec3b7d47b3b2
fix(coverage): warn instead of erroring when source files are missing (#32398) When a source file is deleted after coverage was collected (e.g. a build artifact cleaned up before the report runs), `deno coverage` now prints a warning and skips the file instead of failing the entire report. Closes https://github.com/d...
[ { "path": "cli/tools/coverage/mod.rs", "patch": "@@ -620,7 +620,7 @@ pub fn cover_files(\n };\n let get_message = |specifier: &ModuleSpecifier| -> String {\n format!(\n- \"Failed to fetch \\\"{}\\\" from cache. Before generating coverage report, run `deno test --coverage` to ensure consistent s...
2026-03-02T13:20:28
vuejs/vue
60b1af9e02b93d9223d2ed1f23e0a618537a4c96
15031b85427df5409f0bc4c10589cc6259f8a5b2
fix(transition-group): work around rollup tree shaking (#6796) fix #6792
[ { "path": "src/platforms/web/runtime/components/transition-group.js", "patch": "@@ -101,8 +101,9 @@ export default {\n children.forEach(applyTranslation)\n \n // force reflow to put everything in position\n- const body: any = document.body\n- const f: number = body.offsetHeight // eslint-disab...
2017-10-13T13:05:53
ggml-org/llama.cpp
8ce774a102cd1b24f06cb74c056d05415cddb08e
67788f68468e1d8c0da40d7015e2e8b704c42f8a
metal : fix build(#17799) * metal : fix build * tests : fix context destruction
[ { "path": "examples/save-load-state/save-load-state.cpp", "patch": "@@ -241,6 +241,12 @@ int main(int argc, char ** argv) {\n \n llama_batch_free(batch);\n \n+ // this one is managed by common_init_result\n+ //llama_free(ctx);\n+\n+ llama_free(ctx2);\n+ llama_free(ctx3);\n+\n if (result0...
2025-12-06T07:33:59
ollama/ollama
23ebe8fe11995cfd99cafcf9871fce9a9a120fd1
2c017ca44170fb56141bd15d6929390b7be3830a
fix some typos (#2973) Signed-off-by: hishope <csqiye@126.com>
[ { "path": "docs/development.md", "patch": "@@ -69,7 +69,7 @@ go build .\n \n *Your operating system distribution may already have packages for AMD ROCm and CLBlast. Distro packages are often preferable, but instructions are distro-specific. Please consult distro-specific docs for dependencies if available!*...
2024-03-07T06:50:11
huggingface/transformers
59ed41e6a0c0b9ce7d7eaaa6fee513282f73bc86
5169c2364ba1ad2f555b83eac1b9065aa2955101
Fix tp (#42368) * up * oups you need renamed key for merge moduleoist * oups there was something I forgot * update
[ { "path": "src/transformers/core_model_loading.py", "patch": "@@ -745,7 +745,7 @@ def convert_and_load_state_dict_in_model(\n mapping.distributed_operation = tp_layer(\n device_mesh=device_mesh, rank=device_map[\"\"].index, empty_param=empty_param.clone()\...
2025-11-25T09:58:50
denoland/deno
4f22dcc1a0fbbd38474bd3774202ec3b7d47b3b2
b9f68895173ac6722006b38ab72268068cd4f284
fix(ext/node): use internal compileFunction binding in CJS wrap (#32373) Closes #32372 This PR replaces wrapping a CJS module into a function [here](https://github.com/denoland/deno/blob/0d2ef7c3b608350bda07279e845d8cabdedfac9e/ext/node/polyfills/01_require.js#L965-L968) with V8's [CompileFunction](https://v8.github....
[ { "path": "ext/node/lib.rs", "patch": "@@ -173,7 +173,6 @@ deno_core::extension!(deno_node,\n deps = [ deno_io, deno_fs ],\n parameters = [TInNpmPackageChecker: InNpmPackageChecker, TNpmPackageFolderResolver: NpmPackageFolderResolver, TSys: ExtNodeSys],\n ops = [\n- ops::assert::op_node_get_error_s...
2026-03-02T11:38:24
ggml-org/llama.cpp
d8c0a7b085302017f6583dbb309e4da83084087a
933414c0b6f21af269bdb4fa2fa1b257b9c0fc53
vulkan: Fix mismatch in TOPK_MOE unit test (#17541) * Fix shader to support 2D workgroup mapping to a single subgroup * Set required_subgroup_size topk_moe shader requires static WARP_SIZE and actual subgroup size to match
[ { "path": "ggml/src/ggml-vulkan/ggml-vulkan.cpp", "patch": "@@ -4174,9 +4174,9 @@ static void ggml_vk_load_shaders(vk_device& device) {\n ggml_vk_create_pipeline(device, device->pipeline_conv2d_dw_cwhn_f16_f32, \"conv2d_dw_cwhn_f16_f32\", conv2d_dw_cwhn_f16_f32_len, conv2d_dw_cwhn_f16_f32_data, \"main\"...
2025-12-06T05:23:30
vuejs/vue
15031b85427df5409f0bc4c10589cc6259f8a5b2
8645e5f04aff8b0eacd617bbcf2395615c25d68b
fix(v-model): allow arbitrary naems for type binding (#6802) Fix #6800
[ { "path": "src/platforms/web/compiler/modules/model.js", "patch": "@@ -36,7 +36,7 @@ function preTransformNode (el: ASTElement, options: CompilerOptions) {\n addRawAttr(branch0, 'type', 'checkbox')\n processElement(branch0, options)\n branch0.processed = true // prevent it from double-proc...
2017-10-13T12:55:02
ollama/ollama
be330174dd712353103887d16e9ec321f52422c0
ce9f7c467481e4d636de5c5befb0a09da06b3614
Allow setting max vram for workarounds Until we get all the memory calculations correct, this can provide and escape valve for users to workaround out of memory crashes.
[ { "path": "gpu/gpu.go", "patch": "@@ -242,6 +242,15 @@ func getCPUMem() (memInfo, error) {\n }\n \n func CheckVRAM() (int64, error) {\n+\tuserLimit := os.Getenv(\"OLLAMA_MAX_VRAM\")\n+\tif userLimit != \"\" {\n+\t\tavail, err := strconv.ParseInt(userLimit, 10, 64)\n+\t\tif err != nil {\n+\t\t\treturn 0, fmt...
2024-03-07T00:53:51
huggingface/transformers
5169c2364ba1ad2f555b83eac1b9065aa2955101
152f5b68feb4ddda2938a144d8514b44277037c5
[core] Fix torchao (#42289) * inital commit * up * update unexpected later on * fix * update * simplify our lives * isolate a bit more * fixup * small nits * style * nit * fix common cases * fix post merge * bnb needs missing keys * small fix * bettrer documentation * no veradict + base class * rake ...
[ { "path": "src/transformers/core_model_loading.py", "patch": "@@ -115,6 +115,8 @@ def convert(\n source_keys: list[str],\n target_keys: list[str],\n full_layer_name: str,\n+ model,\n+ missing_keys,\n config,\n **kwargs,\n ) -> dict[str, list[torch.Te...
2025-11-25T09:22:35
denoland/deno
759546d4bcd3c7c4ff298825eb01c6c79845422a
610143913437ee07c935821c365e4feaff875228
test: add regression test for catching error in require(esm) (#32352) Another test case for https://github.com/denoland/deno/issues/32279
[ { "path": "tests/specs/run/npm_pkg_requires_esm_js/__test__.jsonc", "patch": "@@ -1,5 +1,14 @@\n {\n- \"args\": \"run -A main.js\",\n- \"output\": \"output.out\",\n- \"exitCode\": 1\n+ \"tests\": {\n+ \"esm_file_loaded_via_require\": {\n+ \"args\": \"run -A main.js\",\n+ \"output\": \"outpu...
2026-03-02T11:22:14
vuejs/vue
5a15a8d2089bb833b892123c31a2ca04a511c4c8
1f84dd1c2488d12ef144d4b548b0e80647f9403c
fix(ssr): add semicolon before self-removal script (#6794)
[ { "path": "src/server/template-renderer/index.js", "patch": "@@ -192,7 +192,7 @@ export default class TemplateRenderer {\n windowKey = '__INITIAL_STATE__'\n } = options || {}\n const autoRemove = process.env.NODE_ENV === 'production'\n- ? '(function(){var s;(s=document.currentScript||docu...
2017-10-13T12:52:34