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
ollama/ollama
620d5c569e965ac93ac5c58bca5d3d8938cb98bc
cddc63381cffcde5ab50877e831b7595cc6a6d6e
fix parsing big endian gguf
[ { "path": "llm/ggml.go", "patch": "@@ -231,8 +231,7 @@ const (\n \t// Magic constant for `ggla` files (LoRA adapter).\n \tFILE_MAGIC_GGLA = 0x67676C61\n \t// Magic constant for `gguf` files (versioned, gguf)\n-\tFILE_MAGIC_GGUF_LE = 0x46554747\n-\tFILE_MAGIC_GGUF_BE = 0x47475546\n+\tFILE_MAGIC_GGUF = 0x4655...
2024-06-08T19:32:02
ggml-org/llama.cpp
15bff84bf56651d6f991f166a2bf0f362996f7f9
2524c2616458c7d6ee62fa4b4fa17e5091833544
ggml webgpu: initial flashattention implementation (#18610) * FlashAttention (#13) * Add inplace softmax * Move rms_norm to split row approach * Update debug for supports_op * clean up debug statements * neg f16xf32xip builds and runs, havent actually ran a model that uses neg kernel yet though * neg passes back...
[ { "path": ".github/workflows/build.yml", "patch": "@@ -152,13 +152,13 @@ jobs:\n DAWN_VERSION=\"v2.0.0\"\n DAWN_OWNER=\"reeselevine\"\n DAWN_REPO=\"dawn\"\n- DAWN_ASSET_NAME=\"Dawn-5e9a4865b1635796ccc77dd30057f2b4002a1355-macos-latest-Release.zip\"\n- echo \"F...
2026-01-08T16:23:39
vuejs/vue
1b4a8a0c1edaf9c7eb129ba61bca94ba607bbf56
448ba65d2b139b29f1e6891add9925ac22ffe10b
fix(compiler): fix codegen for v-for component inside template fix #9142
[ { "path": "src/compiler/codegen/index.js", "patch": "@@ -406,9 +406,7 @@ export function genChildren (\n el.tag !== 'template' &&\n el.tag !== 'slot'\n ) {\n- // because el may be a functional component and return an Array instead of a single root.\n- // In this case, just a simple...
2018-12-04T15:51:29
denoland/deno
38186cc4f5024d57a7eb9c4ba93b9062114dd8f0
49a9e1355f87028c8d648f9e58decd5fbd78c62c
fix(npm): create shims for all bin entries during global npm install (#32607) Fixes #26133 When running `deno install -g npm:pyright`, only the first bin entry (`pyright`) got a shim. The second entry (`pyright-langserver`) was missing. This affects any npm package with multiple `bin` entries in `package.json`. ### ...
[ { "path": "cli/tools/installer/bin_name_resolver.rs", "patch": "@@ -5,7 +5,6 @@ use std::path::PathBuf;\n use deno_core::error::AnyError;\n use deno_core::url::Url;\n use deno_npm::registry::NpmRegistryApi;\n-use deno_npm::resolution::NpmPackageVersionResolver;\n use deno_npm::resolution::NpmVersionResolver...
2026-03-21T08:51:15
huggingface/transformers
fc50bdc68529f05b5f571ef8aaa228698ea9b3f4
c7aec088a63183292518359880271e2e93a288cc
Remove null values from fast image processors dict (#42780) * remove null values from saved preporcessor file for fast image processor * preserve explicit None values != class default * Fix flava test * extend to video processor
[ { "path": "src/transformers/image_processing_utils_fast.py", "patch": "@@ -936,6 +936,18 @@ def _preprocess(\n \n def to_dict(self):\n encoder_dict = super().to_dict()\n- encoder_dict.pop(\"_valid_processor_keys\", None)\n- encoder_dict.pop(\"_valid_kwargs_names\", None)\n- ...
2025-12-15T16:01:33
ggml-org/llama.cpp
2524c2616458c7d6ee62fa4b4fa17e5091833544
cb14b069955f12992f6c47be98a061f23ea13cf5
vulkan: fix push constant size for quantize_q8_1 (#18687) I added an assert to catch further mismatches, and it found several. Fix those, too.
[ { "path": "ggml/src/ggml-vulkan/ggml-vulkan.cpp", "patch": "@@ -1505,6 +1505,11 @@ template <> void init_pushconst_fastdiv(vk_op_sum_rows_push_constants &p) {\n init_fastdiv_values(p.ne01, p.ne0_1mp, p.ne0_1L);\n }\n \n+struct vk_quantize_q8_1_push_constants {\n+ uint32_t ne;\n+ uint32_t n...
2026-01-08T14:40:58
vuejs/vue
448ba65d2b139b29f1e6891add9925ac22ffe10b
0d7fb739257cc3d3fc1ea6127eb1089a762b220c
fix(types): correct scopedSlot types (#9131) see #8946
[ { "path": "types/test/options-test.ts", "patch": "@@ -1,5 +1,5 @@\n import Vue, { VNode } from \"../index\";\n-import { AsyncComponent, ComponentOptions, FunctionalComponentOptions, Component } from \"../index\";\n+import { ComponentOptions, Component } from \"../index\";\n import { CreateElement } from \"....
2018-12-04T01:31:39
denoland/deno
49a9e1355f87028c8d648f9e58decd5fbd78c62c
301534307c61f21bf0185c2dcd981e27cfea8708
fix(ext/node): pass uid/gid to spawn and implement process.getgroups (#32772) Two fixes for child_process uid/gid support: 1. `uid` and `gid` options were destructured but not passed to `Deno.Command` in async `spawn()` (they were already passed in `spawnSync`). Also converts `PermissionDenied` errors to `EPERM` and ...
[ { "path": "ext/node/lib.rs", "patch": "@@ -272,6 +272,7 @@ deno_core::extension!(deno_node,\n ops::os::op_node_os_user_info,\n ops::os::op_geteuid,\n ops::os::op_getegid,\n+ ops::os::op_getgroups,\n ops::os::op_cpus,\n ops::os::op_homedir,\n op_node_build_os,", "additions": 1,...
2026-03-20T17:03:18
huggingface/transformers
c7aec088a63183292518359880271e2e93a288cc
f3d5f2558b75369416ffe255a87a34358d08e429
Enforce call to `post_init` and fix all of them (#42873) * fix all * style * simplify * improve msg * fix * oupsi * add colors and sort
[ { "path": "Makefile", "patch": "@@ -45,7 +45,7 @@ repo-consistency:\n \tpython utils/check_modular_conversion.py\n \tpython utils/check_dummies.py\n \tpython utils/check_repo.py\n-\tpython utils/check_init_weights_data.py\n+\tpython utils/check_modeling_structure.py\n \tpython utils/check_inits.py\n \tpytho...
2025-12-15T15:46:41
ollama/ollama
030e765e7667526e6ee6f3fba2074bee3092a009
ce0dc33cb809405fda18a8077da4058d1f7a5374
fix create model when template detection errors
[ { "path": "llm/gguf.go", "patch": "@@ -618,22 +618,8 @@ func (llm *gguf) Encode(ws io.WriteSeeker, kv KV, tensors []Tensor) error {\n \t\t}\n \t}\n \n-\toffset, err := ws.Seek(0, io.SeekCurrent)\n-\tif err != nil {\n-\t\treturn err\n-\t}\n-\n \tvar alignment int64 = 32\n-\tpadding := llm.padding(offset, ali...
2024-06-07T15:55:46
vuejs/vue
307835284a326569ea12c4a22c7dcb8f36d2d8ca
aca17b4064bf5017506d77e9f5bf286cfe926ae1
fix(ssr): resolve server directives the same as on client (#9129) fix #8961
[ { "path": "src/server/render.js", "patch": "@@ -3,6 +3,7 @@\n import { escape } from 'web/server/util'\n import { SSR_ATTR } from 'shared/constants'\n import { RenderContext } from './render-context'\n+import { resolveAsset } from 'core/util/options'\n import { generateComponentTrace } from 'core/util/debug...
2018-12-02T21:01:19
ggml-org/llama.cpp
cb14b069955f12992f6c47be98a061f23ea13cf5
55abc393552f3f2097f168cb6db4dc495a514d56
vulkan: optimize ssm_scan (#18630) * vulkan: optimize ssm_scan * fix warp vs subgroup naming
[ { "path": "ggml/src/ggml-vulkan/ggml-vulkan.cpp", "patch": "@@ -570,6 +570,7 @@ struct vk_device_struct {\n bool uma;\n bool prefer_host_memory;\n bool float_controls_rte_fp16;\n+ bool subgroup_basic;\n bool subgroup_arithmetic;\n bool subgroup_shuffle;\n bool subgroup_ballot;\n@@...
2026-01-08T14:16:54
denoland/deno
301534307c61f21bf0185c2dcd981e27cfea8708
63550f02b759b018f883a7ee7cb285c6d7cb64ed
fix(ext/node): sqlite prepare() options, defensive default, iterator invalidation (#32756) Fixes 3 SQLite Node.js compat tests from #32706: - **`prepare()` options**: Support `readBigInts`, `returnArrays`, `allowBareNamedParameters`, and `allowUnknownNamedParameters` as second argument to `DatabaseSync.prepare(sql, o...
[ { "path": "ext/node_sqlite/database.rs", "patch": "@@ -276,7 +276,7 @@ impl Default for DatabaseSyncOptions {\n return_arrays: false,\n allow_bare_named_params: true,\n allow_unknown_named_params: false,\n- is_defensive_mode: false,\n+ is_defensive_mode: true,\n timeout: 0,...
2026-03-20T16:47:08
huggingface/transformers
f3d5f2558b75369416ffe255a87a34358d08e429
298d08dc36fca2ebb8669e122036d9a3ce1f54b1
[CB] Easy optimizations for continuous batching (#42839) * Cb example more args * Remove useless sync * Better new tokens, and no more BS1 on outputs * Add dynamic to compile to avoid many graphs * Sort prefix to maximize cache hits * More robust ways to retrieve results in test * Style * Update src/transformer...
[ { "path": "examples/pytorch/continuous_batching.py", "patch": "@@ -182,13 +182,16 @@ def batch_generate(\n \n # Benchmark parameters\n parser.add_argument(\"--samples\", type=int, default=500, help=\"Number of samples to generate\")\n+ parser.add_argument(\n+ \"--input-length\", type=int, ...
2025-12-15T15:21:41
ollama/ollama
ce0dc33cb809405fda18a8077da4058d1f7a5374
78f81fc0e598fdb066e41608a8384d493308728e
llm: patch to fix qwen 2 temporarily on nvidia (#4897)
[ { "path": "llm/patches/06-qwen2.diff", "patch": "@@ -0,0 +1,13 @@\n+diff --git a/llama.cpp b/llama.cpp\n+index 40d2ec2c..f34eb79a 100644\n+--- a/llama.cpp\n++++ b/llama.cpp\n+@@ -6943,7 +6943,7 @@ static struct ggml_tensor * llm_build_kqv(\n+ struct ggml_tensor * kq = ggml_mul_mat(ctx, k, q);\n+ ...
2024-06-07T06:14:33
ggml-org/llama.cpp
9a5724dee2457d58e506268efcb1d2286498cf3d
9c142e3a2a8f1c7415511bd9d24f4790ce2dac88
ggml: add env var GGML_OP_OFFLOAD_MIN_BATCH (#18535) * ggml: add env var GGML_OP_OFFLOAD_MIN_BATCH * makes the min_batch_size for triggering op offload configurable via env var, defaulting to the prior hardcoded value of 32 * ggml: read GGML_OP_OFFLOAD_MIN_BATCH once and store to dev ctx * cann: forward declaration ...
[ { "path": "ggml/src/ggml-cann/ggml-cann.cpp", "patch": "@@ -2541,27 +2541,6 @@ static bool ggml_backend_buft_is_cann(ggml_backend_buffer_type_t buft) {\n return buft->iface.get_name == ggml_backend_cann_buffer_type_name;\n }\n \n-/**\n- * @brief Determines if a tensor operation should be offloaded to th...
2026-01-08T09:03:21
vuejs/vue
e4b1b57fd7117a19cdb376dcb156606e0cc32a94
19c33a7e4072b03069f803263ed0c49feb5f73a9
fix(ssr): adjust call stack size defer threshold fix #8545
[ { "path": "src/server/render-context.js", "patch": "@@ -66,55 +66,54 @@ export class RenderContext {\n }\n \n next () {\n- const lastState = this.renderStates[this.renderStates.length - 1]\n- if (isUndef(lastState)) {\n- return this.done()\n- }\n- /* eslint-disable no-case-declarations ...
2018-12-02T17:31:16
denoland/deno
4a0e228cbc524722d8944dfab63f56873aa10b50
b9589cb3e16765641f3705331afc3f74a9f0c4cd
fix(ext/node): don't fire upgrade event for h2c requests (#32866) HTTP requests to a `node:http` server hang when: 1. The server has an `"upgrade"` listener (e.g., Vite for WebSocket HMR) 2. A browser/client sends an HTTP/2 cleartext upgrade (`Upgrade: h2c`) The server was calling `upgradeHttpRaw()` and emitting the ...
[ { "path": "ext/node/polyfills/http.ts", "patch": "@@ -2342,7 +2342,14 @@ export class ServerImpl extends EventEmitter {\n request.headers.get(\"upgrade\");\n req[kRawHeaders] = request.headers;\n \n- if (req.upgrade && this.listenerCount(\"upgrade\") > 0) {\n+ // Don't fire the \"upg...
2026-03-20T15:45:57
huggingface/transformers
a187b857a72b10bdb37c91857a697324ab41f82d
64c12fdf5f6a35704f2932db533c76973b1e9a30
Remove tied weights from internal attribute if they are not tied (#42871) fix
[ { "path": "src/transformers/modeling_utils.py", "patch": "@@ -2398,13 +2398,15 @@ def tie_weights(self, missing_keys: Optional[set[str]] = None, recompute_mapping\n source_is_there = source_param_name not in missing_keys\n target_is_there = target_param_name not in missing_ke...
2025-12-15T13:03:10
ollama/ollama
60323e08057d36b617f11d3c4958d342a44d0342
d4a86102fd5f84cca50757af00296606ac191890
add embed model command and fix question invoke (#4766) * add embed model command and fix question invoke * Update docs/tutorials/langchainpy.md Co-authored-by: Kim Hallberg <hallberg.kim@gmail.com> * Update docs/tutorials/langchainpy.md --------- Co-authored-by: Kim Hallberg <hallberg.kim@gmail.com> C...
[ { "path": "docs/tutorials/langchainpy.md", "patch": "@@ -45,7 +45,7 @@ all_splits = text_splitter.split_documents(data)\n ```\n \n It's split up, but we have to find the relevant splits and then submit those to the model. We can do this by creating embeddings and storing them in a vector database. We can us...
2024-06-04T05:20:48
vuejs/vue
19c33a7e4072b03069f803263ed0c49feb5f73a9
530ca1b2db315fbd0e360807b2031d26665c5d3d
fix(v-on): correctly remove once listener (#8036) fix #8032
[ { "path": "src/core/instance/events.js", "patch": "@@ -21,25 +21,31 @@ export function initEvents (vm: Component) {\n \n let target: any\n \n-function add (event, fn, once) {\n- if (once) {\n- target.$once(event, fn)\n- } else {\n- target.$on(event, fn)\n- }\n+function add (event, fn) {\n+ target....
2018-12-02T16:32:59
ggml-org/llama.cpp
df7fb92170f1c6ed08bf0943d6d8bf1191543a95
2038101bd9b1dcf45b5410b969fbc5206e25d993
model-conversion : remove -st targets for converted model (#18689) This commit removes the '-st` make target for running the converted embedding model. The motivation for this is that the pooling type is now part of the .gguf metdata of the model and this is used by llama-debug when running the model. So there is no ...
[ { "path": "examples/model-conversion/Makefile", "patch": "@@ -138,16 +138,13 @@ embedding-run-original-model-st: embedding-run-original-model\n embedding-run-converted-model:\n \t@./scripts/embedding/run-converted-model.sh $(CONVERTED_EMBEDDING_MODEL) \\\n \t$(if $(PROMPTS_FILE),--prompts-file \"$(PROMPTS_F...
2026-01-08T08:29:15
denoland/deno
b9589cb3e16765641f3705331afc3f74a9f0c4cd
301f6d1a9b956c61d9f9fc2360b6e6ed86449645
fix(ext/node): update process.versions.napi to 9 (#32870) Closes #30091 `process.versions.napi` was hardcoded to `"8"` but the actual NAPI implementation (`ext/napi/js_native_api.rs`) defines `NAPI_VERSION: u32 = 9`. This mismatch caused `node-pre-gyp` and similar tools to reject NAPI v9 native modules (like `get-win...
[ { "path": "ext/node/polyfills/_process/process.ts", "patch": "@@ -191,7 +191,7 @@ export const versions = {\n ares: \"1.18.1\",\n modules: \"108\",\n nghttp2: \"1.47.0\",\n- napi: \"8\",\n+ napi: \"9\",\n llhttp: \"6.0.10\",\n openssl: \"3.0.7+quic\",\n cldr: \"41.0\",", "additions": 1, ...
2026-03-20T15:24:37
huggingface/transformers
64c12fdf5f6a35704f2932db533c76973b1e9a30
f0d9cd1ff60d5fe79191a3223ec752192b0e19b0
[docs] Improve contribution guidelines for Quantization (#42870) * update * fix * nit * nit
[ { "path": "docs/source/en/quantization/contribute.md", "patch": "@@ -16,7 +16,7 @@ rendered properly in your Markdown viewer.\n \n # Contribute\n \n-Transformers supports many quantization methods such as QLoRA, GPTQ, LLM.int8, and AWQ. However, there are still many more quantization approaches that haven't...
2025-12-15T11:08:40
ollama/ollama
476fb8e89242720a7cdd57400ba928de4dde9cc1
829ff87bd1a98eff727003d3b24748f0f7d8c3ac
Limit GPU lib search for now (#4777) * fix oneapi errors on windows 10
[ { "path": "gpu/gpu.go", "patch": "@@ -16,13 +16,12 @@ import (\n \t\"os\"\n \t\"path/filepath\"\n \t\"runtime\"\n-\t\"strconv\"\n \t\"strings\"\n \t\"sync\"\n \t\"unsafe\"\n \n-\t\"github.com/ollama/ollama/format\"\n \t\"github.com/ollama/ollama/envconfig\"\n+\t\"github.com/ollama/ollama/format\"\n )\n \n t...
2024-06-02T02:24:33
ggml-org/llama.cpp
2038101bd9b1dcf45b5410b969fbc5206e25d993
568371a7264c30ad4583f1859cb815dfc0bc14fa
llama : add `use_direct_io` flag for model loading (#18166) * Adding --direct-io flag for model loading * Fixing read_raw() calls * Fixing Windows read_raw_at * Changing type off_t to size_t for windows and Renaming functions * disable direct io when mmap is explicitly enabled * Use read_raw_unsafe when upload_ba...
[ { "path": "common/arg.cpp", "patch": "@@ -2088,11 +2088,22 @@ common_params_context common_params_parser_init(common_params & params, llama_ex\n add_opt(common_arg(\n {\"--mmap\"},\n {\"--no-mmap\"},\n- string_format(\"whether to memory-map model (if disabled, slower load but may ...
2026-01-08T06:35:30
vuejs/vue
530ca1b2db315fbd0e360807b2031d26665c5d3d
097f6229dffc34af452b106ad2a3b58845588807
fix(core): properly handle reused vnodes This also removes the restrictions on rendering the same slot multiple times. close #7913
[ { "path": "src/core/instance/render-helpers/render-slot.js", "patch": "@@ -26,19 +26,7 @@ export function renderSlot (\n }\n nodes = scopedSlotFn(props) || fallback\n } else {\n- const slotNodes = this.$slots[name]\n- // warn duplicate slot usage\n- if (slotNodes) {\n- if (process.en...
2018-12-02T02:07:18
denoland/deno
301f6d1a9b956c61d9f9fc2360b6e6ed86449645
5a92ed20c018c894e2e7eab81b80f8d2caec5d40
fix(ext/node): improve process.hrtime argument validation (#32837) - Adds proper `ERR_INVALID_ARG_TYPE` validation when `process.hrtime()` receives a non-array argument - Adds `ERR_OUT_OF_RANGE` validation when the array length is not 2 - Fixes nanosecond underflow bug by borrowing from seconds when `nano - prevNano` ...
[ { "path": "ext/node/polyfills/process.ts", "patch": "@@ -103,6 +103,7 @@ const lazyLoadFsUtils = core.createLazyLoader<typeof fsUtils>(\n );\n \n const {\n+ ArrayIsArray,\n NumberMAX_SAFE_INTEGER,\n ObjectDefineProperty,\n ObjectPrototypeIsPrototypeOf,\n@@ -372,8 +373,20 @@ export function hrtime(tim...
2026-03-20T14:45:38
huggingface/transformers
f0d9cd1ff60d5fe79191a3223ec752192b0e19b0
66623a1fd62d54159ad757b68c0aed8dc229d917
Fixes 2 failing tests from AMD CI (#42777) Fix failing tests
[ { "path": "tests/models/helium/test_modeling_helium.py", "patch": "@@ -56,7 +56,7 @@ def test_model_2b(self):\n model_id = \"kyutai/helium-1-preview\"\n expected_texts = Expectations(\n {\n- (\"rocm\", (9, 5)): [\"Hello, today is a great day to start a new project....
2025-12-15T11:02:38
ollama/ollama
763bb65dbb88004cd046c8acc0c8e889816e1828
7ca9605f54105d3209ad72397f96040b381b8ad2
use `int32_t` for call to tokenize (#4738) * use `int32_t` for call to tokenize * variable naming * cleanup * fix crash
[ { "path": "llm/llm.go", "patch": "@@ -63,12 +63,27 @@ func (llm *llamaModel) Tokenize(s string) []int {\n \tcs := C.CString(s)\n \tdefer C.free(unsafe.Pointer(cs))\n \n-\ttokens := make([]int, len(s)+2)\n-\tif n := C.llama_tokenize(llm.m, cs, C.int(len(s)), (*C.llama_token)(unsafe.Pointer(&tokens[0])), C.in...
2024-05-31T04:43:30
vuejs/vue
097f6229dffc34af452b106ad2a3b58845588807
ef8524ab7db8d64ac449ce74f5858aa9d91357ad
fix(core): avoid mutating original children when cloning vnode The on-demand clone strategy introduced in 956756b mutates the owner array of the cloned vnode. This causes the newly cloned vnode to be destroyed when the parent node is destroyed. This is fixed by cloning the children array when cloning a vnode. fix #79...
[ { "path": "src/core/vdom/vnode.js", "patch": "@@ -90,7 +90,10 @@ export function cloneVNode (vnode: VNode): VNode {\n const cloned = new VNode(\n vnode.tag,\n vnode.data,\n- vnode.children,\n+ // #7975\n+ // clone children array to avoid mutating original in case of cloning\n+ // a chi...
2018-12-01T22:05:36
ggml-org/llama.cpp
5b8844ae531d8ff09c1c00a2022293d5b674c787
7e16fef085e8727d534b21a148aa70ebe94f23a2
scripts : fix repos cloned with .git extension (#18669)
[ { "path": "scripts/pr2wt.sh", "patch": "@@ -34,6 +34,7 @@ url_origin=$(git config --get remote.origin.url) || {\n }\n \n org_repo=$(echo $url_origin | cut -d/ -f4-)\n+org_repo=${org_repo%.git}\n \n echo \"org/repo: $org_repo\"\n ", "additions": 1, "deletions": 0, "language": "Shell Script" } ]
2026-01-07T21:35:34
denoland/deno
5a92ed20c018c894e2e7eab81b80f8d2caec5d40
8a4988797cbb1f706ec6b3760080c15b78c092ec
fix(ext/node): implement process.threadCpuUsage() (#32829) - Implement `process.threadCpuUsage()` returning per-thread CPU usage (user/system in microseconds) - Add `op_current_thread_cpu_usage` Rust op reusing existing platform-specific thread CPU usage infrastructure (mach `thread_info` on macOS, `/proc/self/task` o...
[ { "path": "ext/node/polyfills/process.ts", "patch": "@@ -8,6 +8,7 @@ import { core, internals, primordials } from \"ext:core/mod.js\";\n import { initializeDebugEnv } from \"ext:deno_node/internal/util/debuglog.ts\";\n import { format } from \"ext:deno_node/internal/util/inspect.mjs\";\n import {\n+ op_cur...
2026-03-20T14:27:07
huggingface/transformers
66623a1fd62d54159ad757b68c0aed8dc229d917
e17b1b85e3a895c2b3d44f647270ecc8466f4095
Fix speccht5_tts pipeline (#42830) * Fix speccht5_tts pipeline Signed-off-by: jiqing-feng <jiqing.feng@intel.com> * Update src/transformers/pipelines/text_to_audio.py Co-authored-by: Anton Vlasjuk <73884904+vasqu@users.noreply.github.com> --------- Signed-off-by: jiqing-feng <jiqing.feng@intel.com> Co-authored-by...
[ { "path": "src/transformers/pipelines/text_to_audio.py", "patch": "@@ -117,8 +117,8 @@ def __init__(self, *args, vocoder=None, sampling_rate=None, **kwargs):\n else vocoder\n )\n \n- if self.model.config.model_type in [\"musicgen\"]:\n- # MusicGen expect to use ...
2025-12-15T09:51:03
vuejs/vue
ef8524ab7db8d64ac449ce74f5858aa9d91357ad
0b31647c41b0c916d40b962723b1de06672d0e01
fix(compiler): wrap scoped slots v-if conditions in parens (#9119) fix #9114
[ { "path": "src/compiler/codegen/index.js", "patch": "@@ -367,7 +367,7 @@ function genScopedSlot (\n const fn = `function(${String(el.slotScope)}){` +\n `return ${el.tag === 'template'\n ? el.if\n- ? `${el.if}?${genChildren(el, state) || 'undefined'}:undefined`\n+ ? `(${el.if})?${ge...
2018-12-01T05:33:59
ggml-org/llama.cpp
f5245b5e4eff4f0d5624dca39b0bd612da7111ff
ae9f8df77882716b1702df2bed8919499e64cc28
cuda : fix build on cuda 12.8 (#18672) compute121 requires 12.9 Signed-off-by: Oliver Walsh <owalsh@redhat.com>
[ { "path": "ggml/src/ggml-cuda/CMakeLists.txt", "patch": "@@ -47,7 +47,10 @@ if (CUDAToolkit_FOUND)\n # check Modules/Internal/CMakeCUDAArchitecturesValidate.cmake in the CMake git repository instead.\n # However, the architectures 120a-real and 121a-real should work with ...
2026-01-07T21:32:44
denoland/deno
8a4988797cbb1f706ec6b3760080c15b78c092ec
fe6dcf2e7dd7a12d1f92298dc5fa02db2401e9cd
fix(ext/node): freeze os.constants.signals to match Node.js (#32836) - Freezes `os.constants.signals` so that property assignment throws `TypeError` in strict mode, matching Node.js behavior - Enables `parallel/test-os-constants-signals.js` in node compat tests Co-authored-by: Claude Opus 4.6 (1M context) <noreply@an...
[ { "path": "ext/node/polyfills/internal_binding/constants.ts", "patch": "@@ -1,5 +1,7 @@\n // Copyright 2018-2026 the Deno authors. MIT license.\n \n+import { primordials } from \"ext:core/mod.js\";\n+const { ObjectFreeze } = primordials;\n import { op_node_build_os, op_node_fs_constants } from \"ext:core/op...
2026-03-20T10:28:51
huggingface/transformers
e17b1b85e3a895c2b3d44f647270ecc8466f4095
40dc11cd3eb4126652aa41ef8272525affd4a636
[Fix] Fix FA2 kernels ut (#42803) * Fixed FA2 kernels UT * Update * Refactor FA2 kernel map * Update * Update
[ { "path": "src/transformers/modeling_utils.py", "patch": "@@ -155,6 +155,12 @@\n _is_quantized = False\n _is_ds_init_called = False\n \n+# Mapping from flash attention implementations to their kernel fallback repositories\n+FLASH_ATTN_KERNEL_FALLBACK = {\n+ \"flash_attention_2\": \"kernels-community/flas...
2025-12-15T09:24:52
ollama/ollama
bd1d119ba9369804b71fb7e8564a16d98c8a26cd
a03be18189b5cc3b8eb07efde68d358a4293471e
fixed japanese characters deleted at end of line
[ { "path": "cmd/cmd.go", "patch": "@@ -755,7 +755,11 @@ func displayResponse(content string, wordWrap bool, state *displayResponseState)\n \t\t\t\t}\n \n \t\t\t\t// backtrack the length of the last word and clear to the end of the line\n-\t\t\t\tfmt.Printf(\"\\x1b[%dD\\x1b[K\\n\", runewidth.StringWidth(state...
2024-05-30T17:24:21
vuejs/vue
0b31647c41b0c916d40b962723b1de06672d0e01
0d62bb84ffa1af7a4826aecc11c429c7a020645c
chore: fix grammar in core/observer/index.js comments (#9116)
[ { "path": "src/core/observer/index.js", "patch": "@@ -37,7 +37,7 @@ export function toggleObserving (value: boolean) {\n export class Observer {\n value: any;\n dep: Dep;\n- vmCount: number; // number of vms that has this object as root $data\n+ vmCount: number; // number of vms that have this object ...
2018-12-01T05:29:37
ggml-org/llama.cpp
ae9f8df77882716b1702df2bed8919499e64cc28
56d2fed2b3970ae55eebd0e5426d402304b1358a
fix(docker): add missing libglvnd libraries to Vulkan image (#18664) Add libglvnd0, libgl1, libglx0, libegl1, libgles2 to the Vulkan Dockerfile base image. These libraries are required by mesa-vulkan-drivers to properly initialize the Vulkan ICD and detect GPU devices. Without these libraries, vkEnumeratePhysicalDevi...
[ { "path": ".devops/vulkan.Dockerfile", "patch": "@@ -33,6 +33,7 @@ FROM ubuntu:$UBUNTU_VERSION AS base\n \n RUN apt-get update \\\n && apt-get install -y libgomp1 curl libvulkan1 mesa-vulkan-drivers \\\n+ libglvnd0 libgl1 libglx0 libegl1 libgles2 \\\n && apt autoremove -y \\\n && apt clean -y...
2026-01-07T15:57:42
denoland/deno
fe6dcf2e7dd7a12d1f92298dc5fa02db2401e9cd
3b8478109b212be528d385da3a936ea0f110ce24
fix(ext/node): enable node compat test for v8.getHeapStatistics (#32833) Enables `parallel/test-v8-stats.js` in `tests/node_compat/config.jsonc` Ref https://github.com/denoland/deno/issues/32706 --------- Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
[ { "path": "ext/node/ops/v8.rs", "patch": "@@ -37,6 +37,7 @@ pub fn op_v8_get_heap_statistics(\n buffer[11] = stats.total_global_handles_size() as f64;\n buffer[12] = stats.used_global_handles_size() as f64;\n buffer[13] = stats.external_memory() as f64;\n+ buffer[14] = stats.total_allocated_bytes() a...
2026-03-20T10:28:14
huggingface/transformers
40dc11cd3eb4126652aa41ef8272525affd4a636
c247063001ce34d09395ee58a7c5ecf14a25f2c7
Fix Gemma (#42847) fix
[ { "path": "src/transformers/models/gemma/modeling_gemma.py", "patch": "@@ -410,16 +410,14 @@ def forward(\n if position_ids is None:\n position_ids = cache_position.unsqueeze(0)\n \n- # It may already have been prepared by e.g. `generate`\n- if not isinstance(causal_mask_ma...
2025-12-12T18:36:59
ollama/ollama
a03be18189b5cc3b8eb07efde68d358a4293471e
96bc232b43b072a0ffa8cb1d633c6d55f96c9cbb
Fix OLLAMA_LLM_LIBRARY with wrong map name and add more env vars to help message (#4663) * envconfig/config.go: Fix wrong description of OLLAMA_LLM_LIBRARY Signed-off-by: Lei Jitang <leijitang@outlook.com> * serve: Add more env to help message of ollama serve Add more enviroment variables to `ollama serve --...
[ { "path": "cmd/cmd.go", "patch": "@@ -1251,6 +1251,9 @@ func NewCLI() *cobra.Command {\n \t\t\t\tenvVars[\"OLLAMA_NOPRUNE\"],\n \t\t\t\tenvVars[\"OLLAMA_ORIGINS\"],\n \t\t\t\tenvVars[\"OLLAMA_TMPDIR\"],\n+\t\t\t\tenvVars[\"OLLAMA_FLASH_ATTENTION\"],\n+\t\t\t\tenvVars[\"OLLAMA_LLM_LIBRARY\"],\n+\t\t\t\tenvVa...
2024-05-30T16:36:51
ggml-org/llama.cpp
03023296cf63f4177f51db9126b16b06f0e0af98
8c77a04cc723909eab5d3bc3ae14c82f4db1afc7
vulkan: Warptile tuning for Intel Xe2/Xe3 (#18178) * modify warptile tuning for xe3 * intel vendor check w/ coopmat support * fix back formatting * fix formatting change 2 * move intel check to chip specific tuning part * Change to support both windows and linux * modify m_warptile to l_warptile for intel * mod...
[ { "path": "ggml/src/ggml-vulkan/ggml-vulkan.cpp", "patch": "@@ -2996,6 +2996,10 @@ static void ggml_vk_load_shaders(vk_device& device) {\n if ((device->architecture == AMD_GCN) && (device->driver_id != vk::DriverId::eAmdProprietary)) {\n m_warptile_mmq = m_warptile_mmq_int = { 256, 64, 6...
2026-01-07T10:59:47
vuejs/vue
0d62bb84ffa1af7a4826aecc11c429c7a020645c
b0ccb8618382b6ee143d68cf4498d6b3ec18aa81
fix: actually disable dep collection when invoking lifecycle hooks (#9095) fix #9046
[ { "path": "src/core/observer/dep.js", "patch": "@@ -55,11 +55,12 @@ export default class Dep {\n Dep.target = null\n const targetStack = []\n \n-export function pushTarget (_target: ?Watcher) {\n- if (Dep.target) targetStack.push(Dep.target)\n- Dep.target = _target\n+export function pushTarget (target: ?W...
2018-12-01T05:24:30
denoland/deno
3b8478109b212be528d385da3a936ea0f110ce24
8f200f195312c69ec6c53a34dc0de7245b28e1e2
fix: rewrite approve_scripts_no_lock as PTY tests, skip lockfile test on Windows (#32853) - Rewrite `approve_scripts_no_lock` spec tests as PTY-based integration tests — the spec test `approve_scripts_detects_packages` hangs because `deno approve-scripts` (no args) launches an interactive picker that blocks with...
[ { "path": "tests/integration/pm_tests.rs", "patch": "@@ -280,3 +280,59 @@ fn approve_scripts_deny_some() {\n .join(\"install.txt\")\n .assert_matches_text(\"Installed by @denotest/node-lifecycle-scripts!\");\n }\n+\n+#[test(flaky)]\n+fn approve_scripts_no_lock_explicit_package() {\n+ let context = ...
2026-03-20T10:23:47
ollama/ollama
45cbfc5aee6af8e53b99b38e6aa49152df10db81
6d423b383b340b4fd5e99ab03c1e22ad8a59c56d
fix wsl2 status check for nvidia cards (#4689)
[ { "path": "scripts/install.sh", "patch": "@@ -134,7 +134,7 @@ fi\n # WSL2 only supports GPUs via nvidia passthrough\n # so check for nvidia-smi to determine if GPU is available\n if [ \"$IS_WSL2\" = true ]; then\n- if available nvidia-smi && [ -z \"$(nvidia-smi | grep -o \"CUDA Version: [0-9]*\\.[0-9]*\"...
2024-05-28T21:49:46
huggingface/transformers
c247063001ce34d09395ee58a7c5ecf14a25f2c7
a61aba59fb182572790a4d79c85c247a6e319ca7
Reapply modular examples (#42846) * reapply * fix * fix
[ { "path": "examples/modular-transformers/configuration_duplicated_method.py", "patch": "@@ -8,7 +8,7 @@\n from typing import Optional\n \n from ...configuration_utils import PreTrainedConfig\n-from ...modeling_rope_utils import RopeParameters, rope_config_validation, standardize_rope_params\n+from ...modeli...
2025-12-12T18:16:42
vuejs/vue
b111de486b1bdc747fe0f5795fe22697d151bb8c
33e669b22f69a1f9c9147528360fe0bba85534f0
fix: make sure global state is restored in the case of an exception in macrotask callback (#9093)
[ { "path": "src/core/util/next-tick.js", "patch": "@@ -81,9 +81,11 @@ if (typeof Promise !== 'undefined' && isNative(Promise)) {\n export function withMacroTask (fn: Function): Function {\n return fn._withTask || (fn._withTask = function () {\n useMacroTask = true\n- const res = fn.apply(null, argum...
2018-12-01T05:07:46
denoland/deno
8f200f195312c69ec6c53a34dc0de7245b28e1e2
391a2c8e9430e01611886c609f67a45e86b2b007
fix(npm): `approve-scripts` detects packages when lockfile is disabled (#32842) ## Summary - Fix `deno approve-scripts` failing to detect packages with lifecycle scripts when `"lock": false` is set in `deno.json` - The issue was that `approve-scripts` queried the npm resolution snapshot without first resolving package...
[ { "path": "cli/tools/pm/approve_scripts.rs", "patch": "@@ -78,6 +78,10 @@ pub async fn approve_scripts(\n Vec::new(),\n )\n } else {\n+ let npm_installer = factory.npm_installer().await?;\n+ npm_installer\n+ .ensure_top_level_package_json_install()\n+ .await?;\n let npm_resol...
2026-03-20T06:19:22
ggml-org/llama.cpp
ffba4f29e6a9ed7165ea6b94150856c5b49925cb
3333951d86c0a67bac318cc6b0a924b673f6e4e5
examples : add debug utility/example (#18464) * examples : add debug utility/example This commit introduces a new example named llama-debug which is a utility that is intended to be used to assist with developing/debugging a converted model. The motivation for this utilitiy is to assist in model conversion work to v...
[ { "path": "common/arg.cpp", "patch": "@@ -1445,7 +1445,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex\n [](common_params & params, bool value) {\n params.warmup = value;\n }\n- ).set_examples({LLAMA_EXAMPLE_COMPLETION, LLAMA_EXAMPLE_CLI, ...
2026-01-07T09:42:19
ollama/ollama
ad897080a299bf86aee16b498edb5ddb250edd35
b7d316d98d6f6393c58eb8235c07d896d7a479ea
working on integration of multi-byte and multi-width runes (#4549) * integrated runewidth for display management - fixed cursor movement for mutli-width char * updated input and deletion of multi-byte chars * fixed line history with some exceptions * improved insert and add * fixed issues with moving acros...
[ { "path": "readline/buffer.go", "patch": "@@ -5,16 +5,20 @@ import (\n \t\"os\"\n \n \t\"github.com/emirpasic/gods/lists/arraylist\"\n+\t\"github.com/mattn/go-runewidth\"\n \t\"golang.org/x/term\"\n )\n \n type Buffer struct {\n-\tPos int\n-\tBuf *arraylist.List\n-\tPrompt *Prompt\n-\tLineWid...
2024-05-28T19:04:03
huggingface/transformers
a61aba59fb182572790a4d79c85c247a6e319ca7
5b710c75426639aff3348589fd9095b5f229cb87
Improve BatchFeature: stack list and lists of torch tensors (#42750) * stack lists of tensors in BatchFeature, improve error messages, add tests * remove unnecessary stack in fast image processors and video processors * make style * fix tests
[ { "path": "src/transformers/feature_extraction_utils.py", "patch": "@@ -67,11 +67,18 @@ class BatchFeature(UserDict):\n tensor_type (`Union[None, str, TensorType]`, *optional*):\n You can give a tensor_type here to convert the lists of integers in PyTorch/Numpy Tensors at\n i...
2025-12-12T15:24:14
vuejs/vue
33e669b22f69a1f9c9147528360fe0bba85534f0
f43ae6d151de3ab00a5fbd1bb5fd3fabf454de85
fix(ssr): computed properties should pass vm as first argument in ssr (#9090) fix #8977
[ { "path": "src/core/instance/state.js", "patch": "@@ -216,13 +216,13 @@ export function defineComputed (\n if (typeof userDef === 'function') {\n sharedPropertyDefinition.get = shouldCache\n ? createComputedGetter(key)\n- : userDef\n+ : createGetterInvoker(userDef)\n sharedProperty...
2018-12-01T05:05:13
denoland/deno
756d5d92e8028dbaff836e50f31e5875b757c7d4
8ab268c51ce041f60fe6b514030a818b28043e11
fix(ext/node): emit request "close" before socket "free" in keep-alive path (#32811) Fixes a keep-alive socket lifecycle ordering bug where `socket.emit("free")` fired before `request.on("close")` handlers had a chance to run. This caused libraries like `node-fetch` that attach per-request listeners to the socket (and...
[ { "path": "ext/node/polyfills/http.ts", "patch": "@@ -1305,13 +1305,24 @@ export class IncomingMessageForClient extends NodeReadable {\n req._socketErrorListener = null;\n }\n \n- socket.emit(\"free\");\n-\n- // Clear references so old request/response don't destroy the pooled socket...
2026-03-19T19:44:53
ggml-org/llama.cpp
3333951d86c0a67bac318cc6b0a924b673f6e4e5
193ee38a1bc2aed807823112cf030edcae186892
CANN: Fix rename for get_env (#18652) In #18624, get_env in ggml-cann was renamed to get_env_as_lowercase to accurately reflect the function’s behavior and reduce the chance of misuse. However, the update missed renaming call sites in other files. This commit fixes that oversight.
[ { "path": "ggml/src/ggml-cann/aclnn_ops.cpp", "patch": "@@ -1963,7 +1963,7 @@ static void ggml_cann_mat_mul_fp(ggml_backend_cann_context & ctx, ggml_tensor *\n acl_tensor_ptr acl_weight_tensor;\n \n // Only check env once.\n- static bool weight_to_nz = parse_bool(get_env(\"GGML_CANN_WEIGHT_NZ\")....
2026-01-07T08:11:31
ollama/ollama
b7d316d98d6f6393c58eb8235c07d896d7a479ea
d7339fad520df940d4d1ff6aa8553c7be421ccd2
fix nvidia detection in install script (#4683)
[ { "path": "scripts/install.sh", "patch": "@@ -153,6 +153,11 @@ check_gpu() {\n esac\n }\n \n+if check_gpu nvidia-smi; then\n+ status \"NVIDIA GPU installed.\"\n+ exit 0\n+fi\n+\n if ! check_gpu lspci nvidia && ! check_gpu lshw nvidia && ! check_gpu lspci amdgpu && ! check_gpu lshw amdgpu; then\n ...
2024-05-28T16:59:36
huggingface/transformers
5b710c75426639aff3348589fd9095b5f229cb87
33c948e494ecbfe5237000e9c8c1cc4b1fc848e6
Do not rely on config for inferring model dtype (#42838) fix
[ { "path": "src/transformers/modeling_utils.py", "patch": "@@ -883,11 +883,7 @@ def dtype(self) -> torch.dtype:\n \"\"\"\n `torch.dtype`: The dtype of the module (assuming that all the module parameters have the same dtype).\n \"\"\"\n- # Use config dtype, if it's a proper `tor...
2025-12-12T14:46:06
vuejs/vue
f43ae6d151de3ab00a5fbd1bb5fd3fabf454de85
05e8bcfe5d308f280f3640df96bd170fbcf1a9b5
chore(typo): fix typo (#9085)
[ { "path": "types/options.d.ts", "patch": "@@ -68,7 +68,7 @@ export interface ComponentOptions<\n \n el?: Element | string;\n template?: string;\n- // hack is for funcitonal component type inference, should not used in user code\n+ // hack is for functional component type inference, should not be used ...
2018-12-01T04:58:58
ggml-org/llama.cpp
95ea9e0861b28adca740dbc09494f72105c9b92b
ccbc84a5374bab7a01f68b129411772ddd8e7c79
Hexagon add support for f16/f32 flash attention, scale, set-rows and improve f16/32 matmul (#18611) * hexagon: improve fp16 matmul and add fp32/fp16 flash-attention * hexagon: add support for set-rows fp32 -> fp16 with i32/i64 row-idx * hexagon: add support for SCALE fp32 * hexagon: replace scalar fp32 -> fp16 copy...
[ { "path": "ggml/src/ggml-hexagon/ggml-hexagon.cpp", "patch": "@@ -1773,6 +1773,37 @@ static bool hex_supported_dims2(const struct ggml_tensor * x, const struct ggml_\n return true;\n }\n \n+static bool ggml_hexagon_supported_flash_attn_ext(const struct ggml_hexagon_session * sess, const struct ggml_tens...
2026-01-07T01:38:29
denoland/deno
8a5a704ce83c562c7ed710172c5b3ae67b2d933a
af16b8aebf7b87336bec9dc6b74bae6df22ef71c
fix: make flamegraph SVG fill full browser viewport (#32727) Fixes #32725 The flamegraph SVG had a fixed pixel `height` attribute, which combined with the `viewBox` aspect ratio, caused the browser to constrain the rendered width — leaving the right side of the viewport empty.
[ { "path": "runtime/cpu_profiler/flamegraph.js", "patch": "@@ -28,7 +28,7 @@ function init(evt) {\n toggle_invert();\n });\n }\n- orig_height = parseFloat(svg.attributes.height.value);\n+ orig_height = parseFloat(frames.getAttribute(\"fg:content_height\"));\n // Fluid: fill viewport width and...
2026-03-19T15:08:48
huggingface/transformers
33c948e494ecbfe5237000e9c8c1cc4b1fc848e6
e6b9d061472afc4ee376573789cd44b0180827f3
[`T5Gemma2`] Fix bidirectional mask for encoder (#42820) * fix * add test and fix copies * comments * fixup * fix docstrings
[ { "path": "src/transformers/models/t5gemma2/configuration_t5gemma2.py", "patch": "@@ -32,9 +32,9 @@\n \n class T5Gemma2TextConfig(PreTrainedConfig):\n r\"\"\"\n- This is the configuration class to store the configuration of a [`T5Gemma2TextModel`]. It is used to instantiate an T5Gemma2Text\n- mode...
2025-12-12T14:34:39
ollama/ollama
7487229c343e5177356a40ddc5061a9fc6dfaded
8a8e7afa968f9c241a6bf85e2e9f711e8be41c7c
llm/server.go: Fix 2 minor typos (#4661) Signed-off-by: Lei Jitang <leijitang@outlook.com>
[ { "path": "llm/server.go", "patch": "@@ -24,9 +24,9 @@ import (\n \t\"golang.org/x/sync/semaphore\"\n \n \t\"github.com/ollama/ollama/api\"\n+\t\"github.com/ollama/ollama/envconfig\"\n \t\"github.com/ollama/ollama/format\"\n \t\"github.com/ollama/ollama/gpu\"\n-\t\"github.com/ollama/ollama/envconfig\"\n )\n...
2024-05-28T00:21:10
vuejs/vue
05e8bcfe5d308f280f3640df96bd170fbcf1a9b5
1129d184482720bb52fa61f0a309645b7caaddd9
fix: fix server env detection in wechat mini program (#9075)
[ { "path": "src/core/util/env.js", "patch": "@@ -41,7 +41,7 @@ export const isServerRendering = () => {\n if (!inBrowser && !inWeex && typeof global !== 'undefined') {\n // detect presence of vue-server-renderer and avoid\n // Webpack shimming the process\n- _isServer = global['process']...
2018-12-01T04:58:29
ggml-org/llama.cpp
68b4d516c305325d31e698c4673b691d2a9d879f
24af22fc365ea6ef8e37875108a83658aa16fc8a
llama-params-fit: fix last devices with low VRAM (#18494)
[ { "path": "src/llama.cpp", "patch": "@@ -359,6 +359,11 @@ static void llama_params_fit_impl(\n \n // for the first partial layer varying parts can overflow, all further layers use LAYER_FRACTION_MOE:\n layer_fraction_t overflow_type = LAYER_FRACTION_MOE;\n+\n+ uint32_t n_full() const ...
2026-01-06T19:02:30
denoland/deno
f45e443d834fab85dd87b7d6253006fcc2cf50ff
c436b5fba79032e3fa0e80f795eb25c3bec55243
refactor(core): move timer processing from Rust to JavaScript (#32543) ## Summary - Moves user timer management (setTimeout/setInterval) from Rust's `BTreeSet` to a JavaScript-side implementation matching Node.js's timer architecture: linked lists bucketed by duration, binary min-heap priority queue, and a single nat...
[ { "path": "ext/node/polyfills/internal/crypto/_randomBytes.ts", "patch": "@@ -77,7 +77,7 @@ export default function randomBytes(\n const resource = {};\n emitInit(asyncId, \"RANDOMBYTESREQUEST\", triggerAsyncId, resource);\n \n- setTimeout(() => {\n+ process.nextTick(() => {\n emitBefore...
2026-03-19T14:01:05
huggingface/transformers
e6b9d061472afc4ee376573789cd44b0180827f3
65dc261512cbdb1ee72b88ae5b222f2605aad8e5
[saving] Simplify general logic (#42766) * parallelize and cleanup * simplify offloading * fix * oupsi * add env variable to deactivate * revert threading -> safetensors does not release the GIL * comment * create helper * move it to accelerate integration
[ { "path": "src/transformers/integrations/accelerate.py", "patch": "@@ -554,6 +554,32 @@ def offload_weight(weight: torch.Tensor, weight_name: str, offload_folder: str |\n return offload_index\n \n \n+def load_offloaded_parameter(model: \"PreTrainedModel\", param_name: str) -> torch.Tensor:\n+ \"\"\"L...
2025-12-12T14:18:55
ollama/ollama
8a8e7afa968f9c241a6bf85e2e9f711e8be41c7c
c79f8c9c3934ee8c9507ab89d4ec31c16bbd0fd8
small fix on examples/python-simplechat/client.py to actually get a streamed response and get tokens printed as we receive it (#4671)
[ { "path": "examples/python-simplechat/client.py", "patch": "@@ -9,6 +9,7 @@ def chat(messages):\n r = requests.post(\n \"http://0.0.0.0:11434/api/chat\",\n json={\"model\": model, \"messages\": messages, \"stream\": True},\n+\tstream=True\n )\n r.raise_for_status()\n output =...
2024-05-28T00:19:20
ggml-org/llama.cpp
3d26a09dc7b1a7c13da57fdd26d1cf22efa81229
bd2a93d4753c4f00443f561ee039220283016ee8
server : add thinking content blocks to Anthropic Messages API (#18551) * server : add thinking content blocks to Anthropic Messages API Add support for returning reasoning/thinking content in Anthropic API responses when using models with --reasoning-format deepseek and the thinking parameter enabled. - Non-streami...
[ { "path": "tools/server/server-task.cpp", "patch": "@@ -814,6 +814,15 @@ json server_task_result_cmpl_final::to_json_anthropic() {\n msg.content = content;\n }\n \n+ // thinking block comes first (Anthropic extended thinking format)\n+ if (!msg.reasoning_content.empty()) {\n+ conten...
2026-01-06T15:17:13
vuejs/vue
97086f365808a040f6cf1ddb12e2b3f63d7769bf
76443803e8878dc13a31ebed119abfd31573e7f1
fix(types): fix vm.$once argument type (#8995) vm.$once should accept an event or array of events as first parameter. fix #8983
[ { "path": "types/vue.d.ts", "patch": "@@ -53,7 +53,7 @@ export interface Vue {\n options?: WatchOptions\n ): (() => void);\n $on(event: string | string[], callback: Function): this;\n- $once(event: string, callback: Function): this;\n+ $once(event: string | string[], callback: Function): this;\n ...
2018-12-01T04:41:27
denoland/deno
c436b5fba79032e3fa0e80f795eb25c3bec55243
b3e93cf7bd7e795ffe09d9212f72d8ae65f5986e
fix(ext/node): improve readline and EventEmitter compat (#32826) ## Summary - **readline**: `pause()`/`resume()` now throw `ERR_USE_AFTER_CLOSE` when called after `close()`, matching Node.js behavior - **events**: `listenerCount()` now supports `EventTarget` objects (delegates to `getEventListeners()`) - **events**: ...
[ { "path": "ext/node/polyfills/_events.mjs", "patch": "@@ -50,6 +50,7 @@ const {\n } = primordials;\n \n const kRejection = SymbolFor(\"nodejs.rejection\");\n+const kWatermarkData = SymbolFor(\"nodejs.watermarkData\");\n export const kEvents = Symbol(\"kEvents\");\n \n import { inspect } from \"ext:deno_node...
2026-03-19T13:58:51
huggingface/transformers
64a7cc82a681803d6bdc9e96a1a212cebc5655d7
37426b27bf51fb4cc9f7f7645a61d1b5dd28f20f
Simplify dtype instantiation (#42825) * fix * simplify * align tests * fix * oupsi - super bad traceback * fix * fix * fix * simplify * fix * comment
[ { "path": "src/transformers/modeling_utils.py", "patch": "@@ -233,23 +233,28 @@ def set_zero3_state():\n _is_ds_init_called = False\n \n \n-def restore_default_dtype(func):\n+@contextmanager\n+def local_torch_dtype(dtype: torch.dtype, model_class_name: str | None = None):\n \"\"\"\n- Decorato...
2025-12-12T13:46:02
ggml-org/llama.cpp
e75ee11024befa163cbc0398f9e697e4b32c5f2c
da9b8d330052916fece07f668750d07be87978c5
ggml : fix avx512bf16 build (#18623) - include `immintrin.h` when required - remove unused m512bh Signed-off-by: Adrien Gallouët <angt@huggingface.co>
[ { "path": "ggml/src/ggml.c", "patch": "@@ -53,13 +53,15 @@\n \n #define UNUSED GGML_UNUSED\n \n+// Needed for ggml_fp32_to_bf16_row()\n+#if defined(__AVX512BF16__)\n #if defined(_MSC_VER)\n-#define m512bh(p) p\n #define m512i(p) p\n #else\n-#define m512bh(p) (__m512bh)(p)\n+#include <immintrin.h>\n #define ...
2026-01-06T06:54:10
vuejs/vue
76443803e8878dc13a31ebed119abfd31573e7f1
3cd4af4af0a8a67f5887d5fc967147d433c8612c
Typings: Improve $slots and $scopedSlots type to prevent unchecked access to undefined (#8946) * fix(types): Declare $scopedSlots as potentially undefined to enable stricter TS checks * fix(types): Fix tests * fix(types): declare $slots option as potentially undefined declare $slots option as potentially unde...
[ { "path": "types/test/options-test.ts", "patch": "@@ -304,8 +304,8 @@ Vue.component('component-with-scoped-slot', {\n child: {\n render (this: Vue, h: CreateElement) {\n return h('div', [\n- this.$scopedSlots['default']({ msg: 'hi' }),\n- this.$scopedSlots['item']({ msg: ...
2018-12-01T04:31:45
denoland/deno
42eb6ab0d626addee4cc8aee76a4f6d69a1609c2
a2a795569abadeb64d04416e28a13bc38614fbe0
fix(ext/node): add key/IV length validation for aes-128-cbc and ecb cipher modes (#32824) Validate key and IV lengths for aes-128-cbc, aes-128-ecb, aes-192-ecb, and aes-256-ecb before constructing ciphers, matching the pattern already used by aes-256-cbc, ctr, and gcm modes. Co-authored-by: Claude Opus 4.6 (1M contex...
[ { "path": "ext/node_crypto/cipher.rs", "patch": "@@ -221,11 +221,32 @@ impl Cipher {\n use Cipher::*;\n Ok(match algorithm_name {\n \"aes128\" | \"aes-128-cbc\" => {\n+ if key.len() != 16 {\n+ return Err(CipherError::InvalidKeyLength);\n+ }\n+ if iv.len() != 16 {\...
2026-03-19T07:02:28
huggingface/transformers
37426b27bf51fb4cc9f7f7645a61d1b5dd28f20f
aa495f62dec75a7c6593e978e5fed475d9d1efc0
Fix a typo in MoE models (#42835) fix
[ { "path": "src/transformers/models/deepseek_v3/modeling_deepseek_v3.py", "patch": "@@ -157,7 +157,7 @@ def __init__(self, config):\n super().__init__()\n self.num_experts = config.num_local_experts\n self.hidden_dim = config.hidden_size\n- self.intermediate_dim = config.interm...
2025-12-12T12:33:36
vuejs/vue
3cd4af4af0a8a67f5887d5fc967147d433c8612c
0ba79e2588309ba386f570ed84d372611c4dd165
fix(types): support typing $el as SVGElement (#8809)
[ { "path": "types/test/vue-test.ts", "patch": "@@ -19,6 +19,7 @@ class Test extends Vue {\n }\n \n // test property reification\n+ $el!: HTMLElement | SVGElement;\n $refs!: {\n vue: Vue,\n element: HTMLInputElement,", "additions": 1, "deletions": 0, "language": "Unknown" }, { ...
2018-12-01T04:30:31
ggml-org/llama.cpp
df17a4c94f09c0e978e83102fcdbdf6020599460
1871f0ba56e57826c1c630c5f57274624d68788e
CUDA: fix FA FP16 accumulator overflow for Granite (#18614)
[ { "path": "ggml/src/ggml-cuda/fattn-common.cuh", "patch": "@@ -11,10 +11,12 @@\n #define SOFTMAX_FTZ_THRESHOLD -20.0f // Softmax exp. of values smaller than this are flushed to zero to avoid NaNs.\n \n // log(2) = 0.6931, by adding this to the KQ maximum used for the softmax the numerical ...
2026-01-05T18:51:13
denoland/deno
a2a795569abadeb64d04416e28a13bc38614fbe0
a087e7ccf983ab2ef304a1c3baf1c47892207d78
fix(ext/node): improve node:dns module compatibility (#32704) ## Summary - Support `--dns-result-order` flag via `NODE_OPTIONS` env var by parsing it in the node_options binding and making `dnsOrder` initialization lazy (evaluated at runtime instead of snapshot time) - Replace `node:punycode` import with `internal/idn...
[ { "path": "ext/node/polyfills/internal/dns/promises.ts", "patch": "@@ -64,7 +64,7 @@ import cares, {\n GetNameInfoReqWrap,\n QueryReqWrap,\n } from \"ext:deno_node/internal_binding/cares_wrap.ts\";\n-import { toASCII } from \"node:punycode\";\n+import { domainToASCII } from \"ext:deno_node/internal/idna...
2026-03-19T06:51:35
huggingface/transformers
aa495f62dec75a7c6593e978e5fed475d9d1efc0
c24b51dd78b947517bd23b70ce0717f3a6d0107f
Fixes for the failures of AMD CI (#42718) * Bump timm requirement * Add sync to avoid params not transfered * Multi-device BLT error * Multi-device EfficientNet error * Multi-device Ernie error * Multi-device Mimi error * Makes * Stop using .to and use no_split_modules instead * Restore no_split to Ernie
[ { "path": "setup.py", "patch": "@@ -140,7 +140,7 @@\n \"tensorboard\",\n \"timeout-decorator\",\n \"tiktoken\",\n- \"timm<=1.0.19,!=1.0.18\",\n+ \"timm>=1.0.20\",\n \"tokenizers>=0.22.0,<=0.23.0\",\n \"torch>=2.2\",\n \"torchaudio\",", "additions": 1, "deletions": 1, ...
2025-12-12T11:34:10
ggml-org/llama.cpp
da143b99403fd526e61f080dcc27aed88b97a914
f1768d8f03fe514794349790de0785eafded6c0b
server : fix router child env in containerized environments (#18562)
[ { "path": "tools/server/server-models.cpp", "patch": "@@ -21,11 +21,13 @@\n \n #ifdef _WIN32\n #include <winsock2.h>\n+#include <windows.h>\n #else\n #include <sys/socket.h>\n #include <netinet/in.h>\n #include <arpa/inet.h>\n #include <unistd.h>\n+extern char **environ;\n #endif\n \n #if defined(__APPLE__)...
2026-01-05T13:12:05
vuejs/vue
0ba79e2588309ba386f570ed84d372611c4dd165
ac217d2472bb92ce901ef1f46595b44a1b5d1a18
fix(types): accept `number` type as key on Vue.set/delete (#8707) (#8709)
[ { "path": "types/test/vue-test.ts", "patch": "@@ -87,8 +87,10 @@ class Test extends Vue {\n this.nextTick(() => {});\n this.nextTick().then(() => {});\n this.set({}, \"\", \"\");\n+ this.set({}, 1, \"\");\n this.set([true, false, true], 1, true);\n this.delete({}, \"\");\n+ this.de...
2018-12-01T04:28:52
huggingface/transformers
c24b51dd78b947517bd23b70ce0717f3a6d0107f
b19844eef694bea15ba5b1a765997292a26ef6d2
Fix xpu output check for Ministral3 tests (#42761) * fix xpu output check for ministral3 Signed-off-by: jiqing-feng <jiqing.feng@intel.com> * fix format Signed-off-by: jiqing-feng <jiqing.feng@intel.com> * fix xpu check Signed-off-by: jiqing-feng <jiqing.feng@intel.com> * use Expectations Signed-off-by: jiqing-...
[ { "path": "tests/models/ministral3/test_modeling_ministral3.py", "patch": "@@ -21,8 +21,10 @@\n \n from transformers import AutoTokenizer, Mistral3ForConditionalGeneration, is_torch_available\n from transformers.testing_utils import (\n+ Expectations,\n backend_empty_cache,\n cleanup,\n+ requi...
2025-12-12T11:26:14
denoland/deno
a087e7ccf983ab2ef304a1c3baf1c47892207d78
e1ef181d95d0864615b6ccc2b36ff8599ef33f71
fix(watch): support graceful shutdown via SIGTERM dispatch (#32564) ## Summary - When watch mode restarts (file change) or exits (Ctrl+C), synthetically dispatch SIGTERM to any registered signal handlers before cancelling the running operation - If SIGTERM handlers are registered, wait up to 10 seconds for the operat...
[ { "path": "cli/util/file_watcher.rs", "patch": "@@ -39,6 +39,7 @@ use crate::util::fs::canonicalize_path;\n \n const CLEAR_SCREEN: &str = \"\\x1B[H\\x1B[2J\\x1B[3J\";\n const DEBOUNCE_INTERVAL: Duration = Duration::from_millis(200);\n+const GRACEFUL_SHUTDOWN_TIMEOUT: Duration = Duration::from_secs(5);\n \n ...
2026-03-18T21:40:16
ggml-org/llama.cpp
f1768d8f03fe514794349790de0785eafded6c0b
2da64a2f8a47894d70bc87a160dcb57a3df63b6e
vulkan: fix topk_moe_sigmoid_norm_bias failures in GLM-4.6 (#18582)
[ { "path": "ggml/src/ggml-vulkan/vulkan-shaders/topk_moe.comp", "patch": "@@ -101,6 +101,10 @@ void main() {\n const uint lane = gl_SubgroupInvocationID;\n \n float probs[experts_per_thread];\n+ [[unroll]]\n+ for (int i = 0; i < experts_per_thread; i++) {\n+ probs[i] = -INFINITY;\n+ }...
2026-01-05T10:51:39
huggingface/transformers
b19844eef694bea15ba5b1a765997292a26ef6d2
780cc65907949a48d6d4908ec851c9dabe6ee455
Compatible with GPTQModel FORAMT.LLM_AWQ (#42833) * Compatible with llm-awq Signed-off-by: ZX-ModelCloud <zx@modelcloud.ai> * Apply style fixes --------- Signed-off-by: ZX-ModelCloud <zx@modelcloud.ai> Co-authored-by: Qubitium-ModelCloud <qubitium@modelcloud.ai> Co-authored-by: github-actions[bot] <github-actions[...
[ { "path": "src/transformers/utils/quantization_config.py", "patch": "@@ -69,6 +69,7 @@ class AwqFormat(str, Enum):\n GEMM = \"gemm\"\n GEMV = \"gemv\"\n GEMV_FAST = \"gemv_fast\"\n+ LLM_AWQ = \"llm-awq\"\n \n \n class AwqBackend(str, Enum):\n@@ -838,14 +839,13 @@ def post_init(self):\n ...
2025-12-12T11:16:53
ollama/ollama
8f440d579aad22faf191ef2f7e9b38b4f614e070
4cc3be30358efcbd463ec30c30998dacdb0cfb5c
fix q5_0, q5_1
[ { "path": "llm/ggml.go", "patch": "@@ -127,7 +127,7 @@ func (t Tensor) blockSize() uint64 {\n \tswitch t.Kind {\n \tcase 0, 1, 24, 25, 26, 27, 28, 31: // F32, F16, I8, I16, I32, I64, F64, BF16\n \t\treturn 1\n-\tcase 2, 3, 8, 9, 20: // Q4_0, Q4_1, Q8_0, Q8_1, IQ4_NL\n+\tcase 2, 3, 4, 5, 6, 7, 8, 9, 20: // Q...
2024-05-24T23:01:37
denoland/deno
e1ef181d95d0864615b6ccc2b36ff8599ef33f71
e03b7909beafddefa84872a74136c1d7b61d6a5f
fix(ext/node): set kLastWriteWasAsync in JS write path to prevent double callback (#32814) ## Summary - Set `streamBaseState[kLastWriteWasAsync] = 1` in the JS `writeBuffer` and `writev` fast path before returning, so `afterWriteDispatched` correctly identifies the write as async and defers the callback - Without thi...
[ { "path": "ext/node/polyfills/internal_binding/stream_wrap.ts", "patch": "@@ -213,6 +213,13 @@ export class LibuvStreamWrap extends HandleWrap {\n );\n }\n \n+ // Mark as async so afterWriteDispatched does not fire the callback\n+ // synchronously. The Rust write_buffer path sets this on the...
2026-03-18T21:39:30
ggml-org/llama.cpp
2da64a2f8a47894d70bc87a160dcb57a3df63b6e
b37124d2d2f0e4f6468746f2cbda286944627a75
models : fix backend assignment for Granite/Nemotron graphs (#18599) * models : fix backend assignment for Granite/Nemotron graphs * cont : add ref * cont : move call to build_inp_embd()
[ { "path": "src/llama-graph.cpp", "patch": "@@ -1326,6 +1326,10 @@ ggml_tensor * llm_graph_context::build_inp_embd(ggml_tensor * tok_embd) const {\n \n res->add_input(std::move(inp));\n \n+ // make sure the produced embeddings are immediately materialized in the ggml graph\n+ // ref: https://github...
2026-01-05T10:34:23
huggingface/transformers
780cc65907949a48d6d4908ec851c9dabe6ee455
3fbd59b6f101375efef801430053aa22f4211317
Fix deepspeed sp loss due to missing labels (#42812) fix-trainer
[ { "path": "src/transformers/trainer.py", "patch": "@@ -3943,6 +3943,9 @@ def _deepspeed_sp_compute_loss(self, model, inputs, return_outputs, pc):\n # Both standard transformer models and Liger-patched models handle shift_labels correctly,\n # so we can directly use the computed loss from the...
2025-12-12T11:08:48
ollama/ollama
db2ffa79f10ebcb6cd702bfceb3533a97e892409
afd2b058b4ee36230ab2a06927bdc0ff41b1e7ae
Fix download retry issue
[ { "path": "server/download.go", "patch": "@@ -221,7 +221,7 @@ func (b *blobDownload) downloadChunk(ctx context.Context, requestURL *url.URL, w\n \t\t}\n \t\tdefer resp.Body.Close()\n \n-\t\tn, err := io.CopyN(w, io.TeeReader(resp.Body, part), part.Size)\n+\t\tn, err := io.CopyN(w, io.TeeReader(resp.Body, pa...
2024-05-24T18:30:42
vuejs/vue
ac217d2472bb92ce901ef1f46595b44a1b5d1a18
c711ec189aaf46399756e34d933ba5e0b6576c36
fix(types): fix `renderError`arguments type (#8636) fix #8635
[ { "path": "types/options.d.ts", "patch": "@@ -70,7 +70,7 @@ export interface ComponentOptions<\n template?: string;\n // hack is for funcitonal component type inference, should not used in user code\n render?(createElement: CreateElement, hack: RenderContext<Props>): VNode;\n- renderError?: (h: () =>...
2018-12-01T04:28:29
denoland/deno
e03b7909beafddefa84872a74136c1d7b61d6a5f
852d4994cb559385db0e6dba3b88da00b602f93d
fix(core): implement select fallback on macOS (#32812) On macos, certain files are incompatible with kqueue. Chief among them being some of the `/dev/` files. In libuv, there's code that checks if a file works with kqueue, and if it doesn't then it has a fallback which uses a background thread that polls via select an...
[ { "path": "libs/core/uv_compat.rs", "patch": "@@ -550,10 +550,18 @@ impl UvLoopInner {\n tty.internal_write_queue.clear();\n tty.internal_shutdown = None;\n \n- // Drop the AsyncFd to deregister from the reactor, then close the fd.\n+ // Drop the reactor (AsyncFd or select fallback) to...
2026-03-18T21:12:38
ggml-org/llama.cpp
b37124d2d2f0e4f6468746f2cbda286944627a75
eadc4184caee5b5f68f31f19a2f65c6961748e46
vulkan: handle quantize_q8_1 overflowing the max workgroup count (#18515) * vulkan: handle quantize_q8_1 overflowing the max workgroup count * vulkan: Fix small tile size matmul on lavapipe * fix mul_mat_id failures
[ { "path": "ggml/src/ggml-vulkan/ggml-vulkan.cpp", "patch": "@@ -2898,39 +2898,41 @@ static void ggml_vk_load_shaders(vk_device& device) {\n const uint32_t tk_m = device->coopmat_support ? device->coopmat_k : 1;\n const uint32_t tk_s = device->coopmat_support ? device->coopmat_k : 1;\n \n- ...
2026-01-05T10:30:14
huggingface/transformers
f80b0485feccccb8102ee788b7609e3ec98a1567
6d00f6b0a5679c36510f203e4226e36f517c3032
[XPU] Fix UT errors in the sam3 and lfm series model. (#42798) * Make sam3 tests pass on XPU * Update flm2 tests GT for XPU * Remove the skip tests of local mask for XPU * Pass position_ids to varlen FA2 * Change modular also * Skip FA2 bwd tests * Make style * Increase rtol * Adapt to the main branch * fix c...
[ { "path": "tests/models/lfm2_moe/test_modeling_lfm2_moe.py", "patch": "@@ -176,8 +176,8 @@ def test_model_1a8b_logits(self):\n # Expected mean on dim = -1\n EXPECTED_MEANS = Expectations(\n {\n- (\"cuda\", None): torch.tensor([[-1.3855, -0.5123, -1.3143, -1.2144, -...
2025-12-12T08:41:23
vuejs/vue
c711ec189aaf46399756e34d933ba5e0b6576c36
613cb52bf3e5fd6caee0a94fab953ac9fdd37924
fix(types): support chain call for Vue.use and Vue.mixin (#8595)
[ { "path": "flow/global-api.js", "patch": "@@ -8,8 +8,8 @@ declare interface GlobalAPI {\n set: <T>(target: Object | Array<T>, key: string | number, value: T) => T;\n delete: <T>(target: Object| Array<T>, key: string | number) => void;\n nextTick: (fn: Function, context?: Object) => void | Promise<*>;\...
2018-12-01T04:27:24
denoland/deno
852d4994cb559385db0e6dba3b88da00b602f93d
10128048bf30b6325d419d992af616cd2e8b9303
fix(ext/node): use constant-time comparison for GCM auth tag verification (#32817) ## Summary - Replace direct `==` comparison with `subtle::ConstantTimeEq` for AES-GCM authentication tag verification in both AES-128-GCM and AES-256-GCM decrypt paths - The standard `==` operator short-circuits on the first differing ...
[ { "path": "Cargo.lock", "patch": "@@ -2828,6 +2828,7 @@ dependencies = [\n \"signature 2.2.0\",\n \"sm3\",\n \"spki 0.7.3\",\n+ \"subtle\",\n \"thiserror 2.0.12\",\n \"tokio\",\n \"x25519-dalek\",", "additions": 1, "deletions": 0, "language": "Unknown" }, { "path": "Cargo.toml", ...
2026-03-18T20:10:18
huggingface/transformers
6217adc6c8f0be7b5374e6a46129ad2214e4c6ed
8a2a83d574fd461697a29410a36737ed112f8ba7
Default auto (#42805) * default to `"auto"` dtype * the actual change? * up? * style * up? * only sam models were broken with this * fix sams * update * fix sam2 now * up * this? * proper fix * lol * fix * fixes * nit * fix * fix copies * fixes * fix bigbird * revert one bit
[ { "path": "src/transformers/core_model_loading.py", "patch": "@@ -885,7 +885,7 @@ def convert_and_load_state_dict_in_model(\n elif dtype_plan != {} and dtype_policy_alt.search(renamed_key):\n matched_dtype_pattern = dtype_policy_alt.search(renamed_key)\n if matche...
2025-12-11T17:03:32
vuejs/vue
a7658e03a16dc507f0abeba41aee705f773727d0
05001e695ebd0b0504d664197a4771463a0f5328
fix(data): skip recursive call if values are identical (#8967)
[ { "path": "src/core/util/options.js", "patch": "@@ -55,7 +55,11 @@ function mergeData (to: Object, from: ?Object): Object {\n fromVal = from[key]\n if (!hasOwn(to, key)) {\n set(to, key, fromVal)\n- } else if (isPlainObject(toVal) && isPlainObject(fromVal)) {\n+ } else if (\n+ toVal...
2018-11-30T23:04:05