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
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
nodejs/node
70896c7e3b97cac03b645a5c450a617131e5b4eb
0ce6be07dadc3e8b3d761d1e3128c68a5f70ed0a
test: fixtures in repl persistent history test PR-URL: https://github.com/nodejs/node/pull/15999 Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de> Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Lance Ball <lball@redhat.com>
[ { "path": "test/parallel/test-repl-persistent-history.js", "patch": "@@ -3,6 +3,7 @@\n // Flags: --expose-internals\n \n const common = require('../common');\n+const fixtures = require('../common/fixtures');\n const stream = require('stream');\n const REPL = require('internal/repl');\n const assert = requir...
2017-10-06T18:28:03
huggingface/transformers
6c7335e053ae882979c4fac01ad6866e564ac21f
f3b3533e19728d5e356d8ff9a09a3672cb26735f
Jamba: fix left-padding test (#30389) fix test
[ { "path": "tests/models/jamba/test_modeling_jamba.py", "patch": "@@ -483,7 +483,7 @@ def _prepare_model_kwargs(input_ids, attention_mask, signature):\n return model_kwargs\n \n for model_class in decoder_only_classes:\n- config, input_ids, attention_mask, _ = self._get_input_i...
2024-04-22T16:02:55
golang/go
b96d32bd92087470f85cfab99e289e609a593d03
2556eb76c8e752907600274d323a52e97418d14f
cmd/go/internal/modload: track which packages are in 'all' during loading If the user requests the 'all' pattern in addition to explicit roots outside of 'all', we should not load the transitive dependencies of those explicit roots as if they were *in* 'all'. Without the '-test' flag, we should not load test dependenc...
[ { "path": "src/cmd/go/internal/modload/load.go", "patch": "@@ -4,6 +4,95 @@\n \n package modload\n \n+// This file contains the module-mode package loader, as well as some accessory\n+// functions pertaining to the package import graph.\n+//\n+// There are several exported entry points into package loading ...
2020-07-01T04:45:34
huggingface/transformers
f3b3533e19728d5e356d8ff9a09a3672cb26735f
0d84901cb7e797c90653e2c8ca2ce2a6b3498208
Fix layerwise GaLore optimizer hard to converge with warmup scheduler (#30372) Update optimization.py
[ { "path": "src/transformers/optimization.py", "patch": "@@ -444,9 +444,8 @@ def get_scheduler(\n \n def scheduler_hook(param):\n # Since the optimizer hook has been already attached we only need to\n- # attach the scheduler hook\n- if param.grad is not None:\n- ...
2024-04-22T16:00:26
nodejs/node
006fdb2fe5d9d1681f25c44009242304740ce154
6bc82daef0550ca820c2c30b40670f3ee810e2c4
test: replace fixturesDir with common.fixtures PR-URL: https://github.com/nodejs/node/pull/15810 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Luigi Pinca <luigipinca@gmail.com> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: Gireesh Punathil <gpunathi@in.ibm.com>
[ { "path": "test/parallel/test-tls-key-mismatch.js", "patch": "@@ -25,15 +25,16 @@ const common = require('../common');\n if (!common.hasCrypto)\n common.skip('missing crypto');\n \n+const fixtures = require('../common/fixtures');\n+\n const assert = require('assert');\n const tls = require('tls');\n-const...
2017-10-06T16:40:55
huggingface/transformers
0d84901cb7e797c90653e2c8ca2ce2a6b3498208
0e9d44d7a17c99467ea27f76c958bad8a4b78485
Terminator strings for generate() (#28932) * stash commit (will discard all of this) * stash commit * First commit - needs a lot of testing! * Add a test * Fix imports and make the tests actually test something * Tests pass! * Rearrange test * Add comments (but it's still a bit confusing) * Sto...
[ { "path": "src/transformers/generation/__init__.py", "patch": "@@ -86,6 +86,7 @@\n \"StoppingCriteria\",\n \"StoppingCriteriaList\",\n \"validate_stopping_criteria\",\n+ \"StopStringCriteria\",\n ]\n _import_structure[\"utils\"] = [\n \"GenerationMixin\",\n@@ -...
2024-04-22T13:13:04
golang/go
2556eb76c8e752907600274d323a52e97418d14f
34835df04891a1d54394888b763af88f9476101d
runtime: ignore SIGPROF if profiling disable for thread This avoids a deadlock on prof.signalLock between setcpuprofilerate and cpuprof.add if a SIGPROF is delivered to the thread between the call to setThreadCPUProfiler and acquiring prof.signalLock. Fixes #41014 Change-Id: Ie825e8594f93a19fb1a6320ed640f4e631553596...
[ { "path": "src/runtime/proc.go", "patch": "@@ -3928,6 +3928,13 @@ func sigprof(pc, sp, lr uintptr, gp *g, mp *m) {\n \t\treturn\n \t}\n \n+\t// If mp.profilehz is 0, then profiling is not enabled for this thread.\n+\t// We must check this to avoid a deadlock between setcpuprofilerate\n+\t// and the call to ...
2020-09-09T17:40:11
nodejs/node
3ddc88b5c2c310a127cc63463b068c6bbf4ce526
abbdcaa6e833864905ded87c3c6b26eefaf7deda
crypto: migrate Certificate to internal/errors Move argument type checking to js, use internal/errors PR-URL: https://github.com/nodejs/node/pull/15756 Reviewed-By: Luigi Pinca <luigipinca@gmail.com> Reviewed-By: Joyee Cheung <joyeec9h3@gmail.com> Reviewed-By: Anna Henningsen <anna@addaleax.net>
[ { "path": "doc/api/crypto.md", "patch": "@@ -64,11 +64,12 @@ console.log(challenge.toString('utf8'));\n // Prints: the challenge as a UTF8 string\n ```\n \n-### Certificate.exportPublicKey(spkac)\n+### Certificate.exportPublicKey(spkac[, encoding])\n <!-- YAML\n added: REPLACEME\n -->\n - `spkac` {string | ...
2017-10-03T13:40:56
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
rust-lang/rust
dae664d09a0f4c72702d96877aeace83cc100f41
444ce09d53eb7bf966f155a322407924fc6ce8cd
Cofigurate out ohos target to avoid compilation crashes
[ { "path": "src/tools/rust-analyzer/crates/profile/Cargo.toml", "patch": "@@ -17,7 +17,7 @@ cfg-if = \"1.0.0\"\n libc.workspace = true\n jemalloc-ctl = { version = \"0.5.0\", package = \"tikv-jemalloc-ctl\", optional = true }\n \n-[target.'cfg(target_os = \"linux\")'.dependencies]\n+[target.'cfg(all(target_o...
2025-02-27T03:43:03
huggingface/transformers
0e9d44d7a17c99467ea27f76c958bad8a4b78485
2d92db8458f7143f64f9b13cbcfee5eb8d0cab90
Update docstrings for text generation pipeline (#30343) * Update docstrings for text generation pipeline * Fix docstring arg * Update docstring to explain chat mode * Fix doctests * Fix doctests
[ { "path": "src/transformers/pipelines/text_generation.py", "patch": "@@ -37,10 +37,11 @@ def __init__(self, messages: Dict):\n class TextGenerationPipeline(Pipeline):\n \"\"\"\n Language generation pipeline using any `ModelWithLMHead`. This pipeline predicts the words that will follow a\n- specif...
2024-04-22T13:01:30
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
golang/go
34835df04891a1d54394888b763af88f9476101d
9ef3ee339634d1a349e7b9bb4cae32aacc326f61
runtime: fix ReadMemStatsSlow's and CheckScavengedBits' chunk iteration Both ReadMemStatsSlow and CheckScavengedBits iterate over the page allocator's chunks but don't actually check if they exist. During the development process the chunks index became sparse, so now this was a possibility. If the runtime tests' heap ...
[ { "path": "src/runtime/export_test.go", "patch": "@@ -358,7 +358,11 @@ func ReadMemStatsSlow() (base, slow MemStats) {\n \t\t}\n \n \t\tfor i := mheap_.pages.start; i < mheap_.pages.end; i++ {\n-\t\t\tpg := mheap_.pages.chunkOf(i).scavenged.popcntRange(0, pallocChunkPages)\n+\t\t\tchunk := mheap_.pages.tryC...
2020-09-09T16:52:18
nodejs/node
bb9ed92f0a4ada8c0ed297c4b8042c853481fb3c
3cc725d9c6b8ed3834abd9c6a7ef44a89d02d012
test: replaced fs.readSync with fixtures.readSync NINA 2017 Code and Learn session for first time contributors. This touched test/parallel/test-https-socket-options.js to replace common.fixturesDir PR-URL: https://github.com/nodejs/node/pull/15882 Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: James M Sne...
[ { "path": "test/parallel/test-https-socket-options.js", "patch": "@@ -25,13 +25,13 @@ const common = require('../common');\n if (!common.hasCrypto)\n common.skip('missing crypto');\n \n+const fixtures = require('../common/fixtures');\n const https = require('https');\n-const fs = require('fs');\n const ht...
2017-10-06T17:04:46
ollama/ollama
b7e137323a190845219206473db3c729b95638a5
ad5ee20c7b5f9113dde3122848442a8cf9c6f08e
Fix a typo (#818)
[ { "path": "docs/api.md", "patch": "@@ -355,7 +355,7 @@ curl -X POST http://localhost:11434/api/embeddings -d '{\n \n ```json\n {\n- \"embeddings\": [\n+ \"embedding\": [\n 0.5670403838157654, 0.009260174818336964, 0.23178744316101074, -0.2916173040866852, -0.8924556970596313,\n 0.8785552978515625,...
2023-10-17T13:00:15
huggingface/transformers
2d92db8458f7143f64f9b13cbcfee5eb8d0cab90
f16caf44bb1606652ac6c7c4ad4bf44973d4e545
`Llama` family, fix `use_cache=False` generation (#30380) * nit to make sure cache positions are not sliced * fix other models * nit * style
[ { "path": "src/transformers/models/cohere/modeling_cohere.py", "patch": "@@ -1175,7 +1175,14 @@ def forward(\n )\n \n def prepare_inputs_for_generation(\n- self, input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, cache_position=None, **kwargs\n+ self,\n+ ...
2024-04-22T12:42:57
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
golang/go
bdad4285709d1c5e04458268880775087be63027
518369601ca2499cea68af86451f17d2856895f8
cmd/go: make 'go get' preserve sums for content of new requirements This preserves zip sums when 'go get' is run on a module that does not have a package in the root directory. The zip must be fetched to determine whether the package should be loaded, so we already load and verify the sum. Note that 'go mod tidy' may...
[ { "path": "src/cmd/go/internal/modload/init.go", "patch": "@@ -862,14 +862,11 @@ func WriteGoMod() {\n \t\t}\n \t}\n \n-\t// Always update go.sum, even if we didn't change go.mod: we may have\n-\t// downloaded modules that we didn't have before.\n-\tmodfetch.WriteGoSum(keepSums())\n-\n \tif !dirty && cfg.Cm...
2020-08-28T19:05:44
nodejs/node
d43e7f39929846f02dd74a827e4b2269f3e977d9
018375cd4e8381da0133bd2e1010e25386c072d2
test: improve coverage for process.umask This ensures that process.umask() throws with the correct error when invalid inputs are supplied. PR-URL: https://github.com/nodejs/node/pull/16188 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: Tobias Nießen <tniessen@...
[ { "path": "test/parallel/test-umask.js", "patch": "@@ -40,3 +40,11 @@ assert.strictEqual(parseInt(mask, 8), process.umask(old));\n assert.strictEqual(old, process.umask());\n // 2. If the test fails, process.umask() will return 0\n assert.strictEqual(old, process.umask());\n+\n+assert.throws(() => {\n+ pro...
2017-10-13T18:15:06
rust-lang/rust
c79a7ed0687c771b5baeb781ea81917d35780533
8f001eaf62ca469887db7c9b47b6feb94cef597c
Introduce `AssocOp::Binary`. It mirrors `ExprKind::Binary`, and contains a `BinOpKind`. This makes `AssocOp` more like `ExprKind`. Note that the variants removed from `AssocOp` are all named differently to `BinOpToken`, e.g. `Multiply` instead of `Mul`, so that's an inconsistency removed. The commit adds `precedence`...
[ { "path": "clippy_lints/src/operators/float_equality_without_abs.rs", "patch": "@@ -50,7 +50,7 @@ pub(crate) fn check<'tcx>(\n // format the suggestion\n let suggestion = format!(\n \"{}.abs()\",\n- sugg::make_assoc(AssocOp::Subtract, &sug_l, &sug_r).maybe_par()\n+ ...
2024-12-19T07:24:07
ollama/ollama
fddb303f23e68ef80a028257554979b22addb438
9ef2fce33a4625da88d4201cdb5e3a074e268def
fix: format string wrong type
[ { "path": "llm/llama.go", "patch": "@@ -70,7 +70,7 @@ func chooseRunners(workDir, runnerType string) []ModelRunner {\n \t\tfiles, err := fs.Glob(llamaCppEmbed, path.Join(path.Dir(r.Path), \"*\"))\n \t\tif err != nil {\n \t\t\t// this is expected, ollama may be compiled without all runners packed in\n-\t\t\t...
2023-10-16T23:14:12
huggingface/transformers
f16caf44bb1606652ac6c7c4ad4bf44973d4e545
9138935784583203fb5f61e8f581cdfdcd887e0f
Add FSDP config for CPU RAM efficient loading through accelerate (#30002) * Add FSDP config for CPU RAM efficient loading * Style fix * Update src/transformers/training_args.py Co-authored-by: Zach Mueller <muellerzr@gmail.com> * Update src/transformers/training_args.py Co-authored-by: amyeroberts <2261...
[ { "path": "src/transformers/training_args.py", "patch": "@@ -513,6 +513,11 @@ class TrainingArguments:\n - sync_module_states (`bool`, *optional*, defaults to `True`)\n If `\"True\"`, each individually wrapped FSDP unit will broadcast module parameters from rank 0 to\n ...
2024-04-22T12:15:28
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
golang/go
bdb480fd623e58d0d1d0689a3755367379ea57bc
1e6ad65b43ee392676a69f769b1942edd8af0e86
cmd/compile: fix mishandling of unsafe-uintptr arguments in go/defer Currently, the statement: go g(uintptr(f())) gets rewritten into: tmp := f() newproc(8, g, uintptr(tmp)) runtime.KeepAlive(tmp) which doesn't guarantee that tmp is still alive by time the g call is scheduled to run. This CL fixes the issue, ...
[ { "path": "src/cmd/compile/internal/gc/order.go", "patch": "@@ -502,6 +502,7 @@ func (o *Order) call(n *Node) {\n \t\t\tx := o.copyExpr(arg.Left, arg.Left.Type, false)\n \t\t\tx.Name.SetKeepalive(true)\n \t\t\targ.Left = x\n+\t\t\tn.SetNeedsWrapper(true)\n \t\t}\n \t}\n ", "additions": 1, "deletions...
2020-09-08T08:28:43
nodejs/node
42479130a93b83f9422071fd4c1ad6c678ad6588
cc443cb44706d5751378d6512ab619eed9976150
doc: fix paragraph line-height issue Fix a line-height issue introduced in #15660 where paragraphs containing <code> blocks would have unequal line heights. Fixes: https://github.com/nodejs/nodejs.org/issues/1399 PR-URL: https://github.com/nodejs/node/pull/16200 Reviewed-By: Gireesh Punathil <gpunathi@in.ibm.com> Rev...
[ { "path": "doc/api_assets/style.css", "patch": "@@ -211,6 +211,7 @@ abbr {\n p {\n text-rendering: optimizeLegibility;\n margin: 0 0 1.125rem 0;\n+ line-height: 1.5;\n }\n \n #apicontent > *:last-child {", "additions": 1, "deletions": 0, "language": "CSS" } ]
2017-10-14T06:18:54
rust-lang/rust
ceafbad81fcf71423d4dc7a90fe962bc5895b108
a8364f3b2ad63c19037bdf060e49ad9ca07e16b6
Introduce `AssocOp::Binary`. It mirrors `ExprKind::Binary`, and contains a `BinOpKind`. This makes `AssocOp` more like `ExprKind`. Note that the variants removed from `AssocOp` are all named differently to `BinOpToken`, e.g. `Multiply` instead of `Mul`, so that's an inconsistency removed. The commit adds `precedence`...
[ { "path": "compiler/rustc_ast/src/ast.rs", "patch": "@@ -39,7 +39,7 @@ pub use crate::format::*;\n use crate::ptr::P;\n use crate::token::{self, CommentKind, Delimiter};\n use crate::tokenstream::{DelimSpan, LazyAttrTokenStream, TokenStream};\n-use crate::util::parser::{AssocOp, ExprPrecedence};\n+use crate...
2024-12-19T07:24:07
huggingface/transformers
d2cec09baa8e88402ac3da86dcee478ae1446c83
21c912e79c8ee62034177bd43c9c628be9b46e2a
Add TF swiftformer (#23342) * Duplicate swiftformer * Convert SwiftFormerPatchEmbedding * Convert SwiftFormerEmbeddings * Convert TFSwiftFormerMlp * Convert TFSwiftFormerConvEncoder * Convert TFSwiftFormerLocalRepresentation * convert TFSwiftFormerEncoderBlock * Convert SwiftFormerStage * Conve...
[ { "path": "docs/source/en/index.md", "patch": "@@ -275,7 +275,7 @@ Flax), PyTorch, and/or TensorFlow.\n | [StableLm](model_doc/stablelm) | ✅ | ❌ | ❌ |\n | [Starcoder2](model_doc/starcoder2) ...
2024-04-19T17:31:43
ollama/ollama
cb4a80b6934c1b62c98a01aab4ae156ba340acc5
9ef2fce33a4625da88d4201cdb5e3a074e268def
fix: regression unsupported metal types omitting `--n-gpu-layers` means use metal on macos which isn't correct since ollama uses `num_gpu=0` to explicitly disable gpu for file types that are not implemented in metal
[ { "path": "llm/llama.go", "patch": "@@ -292,13 +292,10 @@ func newLlama(model string, adapters []string, runners []ModelRunner, numLayers\n \t\t\"--rope-freq-base\", fmt.Sprintf(\"%f\", opts.RopeFrequencyBase),\n \t\t\"--rope-freq-scale\", fmt.Sprintf(\"%f\", opts.RopeFrequencyScale),\n \t\t\"--batch-size\"...
2023-10-16T21:37:17
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
golang/go
1e6ad65b43ee392676a69f769b1942edd8af0e86
a52a5d8a43ac23c34231e92da3aba61cd8cf7d97
cmd/link: enable DWARF combining on macOS ARM64 It appears the machoCalcStart function is meant to align the segment, but it doesn't. Replace it with an actual alignment calculation. Also, use the alignment from the configuration, instead of hardcode. With this fix we could enable DWARF combining on macOS ARM64. Cha...
[ { "path": "src/cmd/link/internal/arm64/obj.go", "patch": "@@ -105,7 +105,7 @@ func archinit(ctxt *ld.Link) {\n \t\t\t*ld.FlagTextAddr = 4096 + int64(ld.HEADR)\n \t\t}\n \t\tif *ld.FlagRound == -1 {\n-\t\t\t*ld.FlagRound = 4096\n+\t\t\t*ld.FlagRound = 16384 // 16K page alignment\n \t\t}\n \t}\n }", "addi...
2020-09-04T18:35:57
nodejs/node
cc443cb44706d5751378d6512ab619eed9976150
411695e1d26565c7498992be19731bbea6b6aa58
build: set disable_glibcxx_debug flag This breaks a few tests when Node is built in debug mode, because V8 is built with `-D_GLIBCXX_DEBUG=1` and Node is built without, which makes e.g. `std::vector` ABI-incompatible between the two. PR-URL: https://github.com/nodejs/node/pull/16159 Reviewed-By: James Snell <jasnell@...
[ { "path": "common.gypi", "patch": "@@ -31,6 +31,10 @@\n # Don't bake anything extra into the snapshot.\n 'v8_use_external_startup_data%': 0,\n \n+ # Some STL containers (e.g. std::vector) do not preserve ABI compatibility\n+ # between debug and non-debug mode.\n+ 'disable_glibcxx_debug': 1,...
2017-10-12T07:12:12
huggingface/transformers
21c912e79c8ee62034177bd43c9c628be9b46e2a
b1cd48740ea52535926631e9e42beee4ba8d8740
Fix config + attn_implementation in AutoModelForCausalLM.from_pretrained (#30299) * Update modeling_utils.py * Update test_modeling_utils.py * Update test_modeling_utils.py * Update test_modeling_utils.py
[ { "path": "src/transformers/modeling_utils.py", "patch": "@@ -3146,7 +3146,7 @@ def from_pretrained(\n config = copy.deepcopy(config)\n \n kwarg_attn_imp = kwargs.pop(\"attn_implementation\", None)\n- if kwarg_attn_imp is not None and config._attn_implementation != kwarg_a...
2024-04-19T16:45:53
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
golang/go
e8f918535e6c9fc7b85a9dadc9e8035a2e1d1fa4
565ad134c994b7a831e1bea336fdc64f0aeefc85
cmd/link: strip STAB (symbolic debugging) symbols on darwin On darwin, with external linking, the system linker produces STAB (symbolic debugging) symbols in the binary's symbol table. These include paths of the intermediate object files, like <tmpdir>/go.o, which changes from run to run, making the build non-reproduc...
[ { "path": "src/cmd/link/internal/ld/lib.go", "patch": "@@ -1238,6 +1238,10 @@ func (ctxt *Link) hostlink() {\n \t\t}\n \t}\n \n+\t// On darwin, whether to combine DWARF into executable.\n+\t// Only macOS supports unmapped segments such as our __DWARF segment.\n+\tcombineDwarf := ctxt.IsDarwin() && !*FlagS &...
2020-08-26T22:19:03
nodejs/node
41573427445022b172f3386fa576e51e8d30c782
7f55349079705a9e1d43024ed45c1351907f3d8e
build: use local node-gyp for benchmark addon Move the logic for building the benchmark/misc/function_call to the top-level Makefile and use our local copy of node-gyp. PR-URL: https://github.com/nodejs/node/pull/16160 Fixes: https://github.com/nodejs/node/issues/16154 Reviewed-By: Joyee Cheung <joyeec9h3@gmail.com> ...
[ { "path": "Makefile", "patch": "@@ -227,6 +227,14 @@ test-valgrind: all\n test-check-deopts: all\n \t$(PYTHON) tools/test.py --mode=release --check-deopts parallel sequential -J\n \n+benchmark/misc/function_call/build/Release/binding.node: all \\\n+\t\tbenchmark/misc/function_call/binding.cc \\\n+\t\tbenchm...
2017-10-12T10:04:02
huggingface/transformers
b4fd49b6c54ac34d45cc656f2872b5f392029590
fbd8c51ffcc23611f99f5a75fe232f1b010eeb72
Update unwrap from accelerate (#29933) * Use unwrap with the one in accelerate * oups * update unwrap * fix * wording * raise error instead * comment * doc * Update src/transformers/modeling_utils.py Co-authored-by: Zach Mueller <muellerzr@gmail.com> * style * put else --------- ...
[ { "path": "src/transformers/modeling_utils.py", "patch": "@@ -109,6 +109,7 @@\n from accelerate.hooks import add_hook_to_module\n from accelerate.utils import (\n check_tied_parameters_on_same_device,\n+ extract_model_from_parallel,\n find_tied_parameters,\n get_balanc...
2024-04-19T16:05:34
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
ollama/ollama
92189a58553ac88edf19ea5a8834478bd5fd6d86
d790bf99166d5a97d9d01c78b0658706c28580b5
fix memory check
[ { "path": "api/client.go", "patch": "@@ -14,6 +14,7 @@ import (\n \t\"runtime\"\n \t\"strings\"\n \n+\t\"github.com/jmorganca/ollama/format\"\n \t\"github.com/jmorganca/ollama/version\"\n )\n \n@@ -127,7 +128,7 @@ func (c *Client) do(ctx context.Context, method, path string, reqData, respData\n \treturn nil...
2023-10-12T16:34:16
huggingface/transformers
fbd8c51ffcc23611f99f5a75fe232f1b010eeb72
0927bfd002f2691059125b7fb8f6e0fc081de695
Restore casting of masked_spec_embed (#30336) * fix Parameter dtype in audio models * restore casting of masked_spec_embed * restore casting of masked_spec_embed
[ { "path": "src/transformers/models/data2vec/modeling_data2vec_audio.py", "patch": "@@ -858,7 +858,7 @@ def _mask_hidden_states(\n \n if mask_time_indices is not None:\n # apply SpecAugment along time axis with given mask_time_indices\n- hidden_states[mask_time_indices] = self....
2024-04-19T15:18:36
nodejs/node
7f55349079705a9e1d43024ed45c1351907f3d8e
2b76b5d8dd14b57eaad15b4082cc007296b4d630
net: convert to using internal/errors Covert lib/net.js over to using lib/internal/errors.js - Replace thrown errors in lib/net.js with errors from lib/internal/errors. The ERR_INVALID_OPT_VALUE error have been used in the Server.prototype.listen() method - Update tests according to the above modifications PR-...
[ { "path": "doc/api/errors.md", "patch": "@@ -940,6 +940,11 @@ Used when `hostname` can not be parsed from a provided URL.\n \n Used when a file descriptor ('fd') is not valid (e.g. it has a negative value).\n \n+<a id=\"ERR_INVALID_FD_TYPE\"></a>\n+### ERR_INVALID_FD_TYPE\n+\n+Used when a file descriptor ('...
2017-08-11T21:02:15
golang/go
dc025c0f9b75edc5140f1895083055a559f6bd79
0e19aaabc052f858beed7d174a55d4110c8d66a0
cmd/compile: handle ODDD in exprformat Fixes #41247 Change-Id: Iaa9502cc610e2cc64be5dfd91ba3187f86f87cbd Reviewed-on: https://go-review.googlesource.com/c/go/+/252942 Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rob Pike <r@golang.org>
[ { "path": "src/cmd/compile/internal/gc/fmt.go", "patch": "@@ -1616,7 +1616,8 @@ func (n *Node) exprfmt(s fmt.State, prec int, mode fmtMode) {\n \t\t\t}\n \t\t\tn1.exprfmt(s, nprec, mode)\n \t\t}\n-\n+\tcase ODDD:\n+\t\tmode.Fprintf(s, \"...\")\n \tdefault:\n \t\tmode.Fprintf(s, \"<node %v>\", n.Op)\n \t}", ...
2020-09-07T05:14:30
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
nodejs/node
2b76b5d8dd14b57eaad15b4082cc007296b4d630
520cf1d6de3373a0949203aac7539cbc96761814
test: use fixtures module Replace common.fixturesDir with fixtures module usage in test/parallel/test-tls-fast-writing.js. PR-URL: https://github.com/nodejs/node/pull/15808 Reviewed-By: Ryan Graham <r.m.graham@gmail.com> Reviewed-By: Benjamin Gruenbaum <benjamingr@gmail.com> Reviewed-By: Luigi Pinca <luigipinca@gmail...
[ { "path": "test/parallel/test-tls-fast-writing.js", "patch": "@@ -24,14 +24,13 @@ const common = require('../common');\n if (!common.hasCrypto)\n common.skip('missing crypto');\n \n+const fixtures = require('../common/fixtures');\n const assert = require('assert');\n const tls = require('tls');\n-const fs...
2017-10-06T16:32:08
golang/go
0e19aaabc052f858beed7d174a55d4110c8d66a0
1a119edd8b4364a8c8342194648166ca5851f061
cmd/asm: fix the error of checking the post-index offset of VLD[1-4]R instructions of arm64 The post-index offset of VLD[1-4]R instructions is decided by the "size" field not "Q" field, the current assembler uses "Q" fileld to check the correctness of post-index offset which is not correct. This patch fixes it. Fixes...
[ { "path": "src/cmd/asm/internal/asm/testdata/arm64.s", "patch": "@@ -359,18 +359,22 @@ TEXT\tfoo(SB), DUPOK|NOSPLIT, $-8\n \tVLD4\t(R15), [V10.H4, V11.H4, V12.H4, V13.H4] // ea05400c\n \tVLD4.P\t32(R24), [V31.B8, V0.B8, V1.B8, V2.B8] // 1f03df0c\n \tVLD4.P\t(R13)(R9), [V14.S2, V15.S2, V16.S...
2020-08-20T10:39:46
rust-lang/rust
8f001eaf62ca469887db7c9b47b6feb94cef597c
17bda0c0febb8c51fb14a6856d365eab86fc6162
In `AssocOp::AssignOp`, use `BinOpKind` instead of `BinOpToken` `AssocOp::AssignOp` contains a `BinOpToken`. `ExprKind::AssignOp` contains a `BinOpKind`. Given that `AssocOp` is basically a cut-down version of `ExprKind`, it makes sense to make `AssocOp` more like `ExprKind`. Especially given that `AssocOp` and `BinOp...
[ { "path": "clippy_utils/src/lib.rs", "patch": "@@ -31,7 +31,6 @@\n // (Currently there is no way to opt into sysroot crates without `extern crate`.)\n extern crate rustc_abi;\n extern crate rustc_ast;\n-extern crate rustc_ast_pretty;\n extern crate rustc_attr_parsing;\n extern crate rustc_const_eval;\n exte...
2024-12-19T05:24:06
ggml-org/llama.cpp
d3dce4e0a5f2fd497a312a3152f2eab939f0d519
4974bf53cf14073c7b66e1151348156aabd42cb8
sampling : add support for backend sampling (#17004) * sampling : add support for backend sampling This commit adds support for performing sampling operations on the backend (e.g. GPU) as part of the model computation graph. The motivation for this feature is to enable sampling to be performed directly on the backen...
[ { "path": ".github/workflows/build.yml", "patch": "@@ -1098,6 +1098,7 @@ jobs:\n save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}\n \n - name: Build with CMake\n+ # TODO: Remove GGML_CUDA_CUB_3DOT2 flag once CCCL 3.2 is bundled within CTK and that CT...
2026-01-04T20:22:16
ollama/ollama
6fe178134dd27c6aaf64374f4163abf6c8d624dd
d890890f665b5220ec1f35829cb7352abdb19c0a
improve api error handling (#781) - remove new lines from llama.cpp error messages relayed to client - check api option types and return error on wrong type - change num layers from 95% VRAM to 92% VRAM
[ { "path": "api/types.go", "patch": "@@ -3,7 +3,6 @@ package api\n import (\n \t\"encoding/json\"\n \t\"fmt\"\n-\t\"log\"\n \t\"math\"\n \t\"os\"\n \t\"reflect\"\n@@ -238,44 +237,39 @@ func (opts *Options) FromMap(m map[string]interface{}) error {\n \t\t\t\t\t\t// when JSON unmarshals numbers, it uses float6...
2023-10-13T20:57:10
huggingface/transformers
91472cf5fc8186a2ca10dd47eb48ab077f26598b
cd09a8dfbce5f4696ba5ec0c711567df214571c0
Pipeline: fix `pad_token_id` again (#30338) fix again
[ { "path": "src/transformers/pipelines/base.py", "patch": "@@ -896,7 +896,7 @@ def __init__(\n and self.tokenizer.pad_token_id is not None\n and self.model.generation_config.pad_token_id is None\n ):\n- kwargs[\"pad_token_id\"] = self.tokenizer.pad_token_id\n+ ...
2024-04-19T11:04:11
nodejs/node
e8a2438cb6a796209464f16262f2ee3961cab1b8
e30876121aa8db0317d22c368e4de08df0ba87d5
test: refactor test-cluster-setup-master - use mustCall instead of counters - include totalWorkers and settings in the error messages PR-URL: https://github.com/nodejs/node/pull/16065 Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de> Reviewed-By: Joyee Cheung <joyeec9h3@gmail.com> Reviewed-By: Yuta Hiroto <hello@...
[ { "path": "test/parallel/test-cluster-setup-master.js", "patch": "@@ -20,7 +20,7 @@\n // USE OR OTHER DEALINGS IN THE SOFTWARE.\n \n 'use strict';\n-require('../common');\n+const common = require('../common');\n const assert = require('assert');\n const cluster = require('cluster');\n \n@@ -38,7 +38,7 @@ if...
2017-10-06T19:32:25
rust-lang/rust
a8364f3b2ad63c19037bdf060e49ad9ca07e16b6
ac91805f3179fc2225c60e8ccf5a1daa09d43f3d
In `AssocOp::AssignOp`, use `BinOpKind` instead of `BinOpToken` `AssocOp::AssignOp` contains a `BinOpToken`. `ExprKind::AssignOp` contains a `BinOpKind`. Given that `AssocOp` is basically a cut-down version of `ExprKind`, it makes sense to make `AssocOp` more like `ExprKind`. Especially given that `AssocOp` and `BinOp...
[ { "path": "compiler/rustc_ast/src/util/parser.rs", "patch": "@@ -46,8 +46,8 @@ pub enum AssocOp {\n GreaterEqual,\n /// `=`\n Assign,\n- /// `?=` where ? is one of the BinOpToken\n- AssignOp(BinOpToken),\n+ /// `?=` where ? is one of the assignable BinOps\n+ AssignOp(BinOpKind),\n ...
2024-12-19T05:24:06
golang/go
1a119edd8b4364a8c8342194648166ca5851f061
5cc030aa1996762e48ce446001078ce6447f105e
runtime: rotate map key seed on clearing up maps Same thing as CL 253020 did for map clear idiom. name old time/op new time/op delta MapDelete/Int32/100-12 30.0ns ± 1% 30.7ns ± 3% ~ (p=0.400 n=3+3) MapDelete/Int32/1000-12 26.6ns ± 2% 28.1ns ± 3% ~ (p=0.100 n=3+3) MapDel...
[ { "path": "src/runtime/map.go", "patch": "@@ -780,6 +780,11 @@ search:\n \t\t\t}\n \t\tnotLast:\n \t\t\th.count--\n+\t\t\t// Reset the hash seed to make it more difficult for attackers to\n+\t\t\t// repeatedly trigger hash collisions. See issue 25237.\n+\t\t\tif h.count == 0 {\n+\t\t\t\th.hash0 = fastrand()...
2020-09-04T10:47:44
ggml-org/llama.cpp
c69c7ebc901acfce2e0a4cacaf3f10085741c5d4
e57f52334b2e8436a94f7e332462dfc63a08f995
graph : fix graph reuse logic when `n_pos_per_embd > 1` (#18566)
[ { "path": "src/llama-graph.cpp", "patch": "@@ -32,7 +32,7 @@ bool llm_graph_input_embd::can_reuse(const llm_graph_params & params) {\n bool res = true;\n \n res &= (!tokens && !params.ubatch.token) || (tokens && tokens->ne[0] == params.ubatch.n_tokens);\n- res &= (!embd && !params.ubatch.embd) ...
2026-01-03T21:59:06
huggingface/transformers
cd09a8dfbce5f4696ba5ec0c711567df214571c0
4ab7a28216211571fdddba414d4edd8426ab6489
[Feature Extractors] Fix kwargs to pre-trained (#30260) fixes
[ { "path": "src/transformers/feature_extraction_utils.py", "patch": "@@ -566,17 +566,17 @@ def from_dict(cls, feature_extractor_dict: Dict[str, Any], **kwargs) -> PreTrain\n \"\"\"\n return_unused_kwargs = kwargs.pop(\"return_unused_kwargs\", False)\n \n- feature_extractor = cls(**feat...
2024-04-19T10:16:08
nodejs/node
e30876121aa8db0317d22c368e4de08df0ba87d5
5aaa23eedd9bf40813fba37a6bc307eda33113fb
test: add tests for ERR_HTTP2_FRAME_ERROR PR-URL: https://github.com/nodejs/node/pull/16107 Reviewed-By: Gibson Fahnestock <gibfahn@gmail.com> Reviewed-By: Daijiro Wachi <daijiro.wachi@gmail.com> Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Yuta Hiroto <hello@about-hiroppy.com>
[ { "path": "test/parallel/test-internal-errors.js", "patch": "@@ -297,6 +297,14 @@ assert.strictEqual(\n errors.message('ERR_HTTP2_HEADER_REQUIRED', ['test']),\n 'The test header is required');\n \n+// Test ERR_HTTP2_FRAME_ERROR\n+assert.strictEqual(\n+ errors.message('ERR_HTTP2_FRAME_ERROR', ['foo', 'b...
2017-10-09T17:39:53
golang/go
5cc030aa1996762e48ce446001078ce6447f105e
617f2c3e35cdc8483b950aa3ef18d92965d63197
go/ast: note that in BasicLit CHARs and STRINGs are quoted The Value field of ast.BasicLit is a string field holding the literal string. For CHARs and STRINGs, the BasicLit.Value literal includes quotes, so to use the value in practise one will often need to Unquote it. Since this is a common gotcha (I've been bitten...
[ { "path": "src/go/ast/ast.go", "patch": "@@ -285,6 +285,12 @@ type (\n \t}\n \n \t// A BasicLit node represents a literal of basic type.\n+\t//\n+\t// Note that for the CHAR and STRING kinds, the literal is stored\n+\t// with its quotes. For example, for a double-quoted STRING, the\n+\t// first and the last...
2020-07-27T12:42:06
ggml-org/llama.cpp
e57f52334b2e8436a94f7e332462dfc63a08f995
a554a1ecc75f6125c30f65fa18327ea68ef52ae6
ggml-cuda: fixes for concurrent streams (#18496)
[ { "path": "ggml/src/ggml-cuda/common.cuh", "patch": "@@ -1063,6 +1063,7 @@ struct ggml_cuda_graph {\n bool disable_due_to_too_many_updates = false;\n bool disable_due_to_failed_graph_capture = false;\n int number_consecutive_updates = 0;\n+ bool cuda_graphs_enabled = false;\n std::vector<...
2026-01-03T15:15:01
huggingface/transformers
4ab7a28216211571fdddba414d4edd8426ab6489
30b453206d224ee5f747afa33ff216671558e6a0
feat: Upgrade Weights & Biases callback (#30135) * feat: upgrade wandb callback with new features * fix: ci issues with imports and run fixup
[ { "path": "src/transformers/integrations/integration_utils.py", "patch": "@@ -31,8 +31,17 @@\n import numpy as np\n import packaging.version\n \n+from .. import PreTrainedModel, TFPreTrainedModel\n from .. import __version__ as version\n-from ..utils import flatten_dict, is_datasets_available, is_pandas_ava...
2024-04-19T10:03:32
ollama/ollama
257ffeb99747a2bb9caed54b1d51020aca95b0ae
92578798bb1abcedd6bc99479d804f32d9ee2f6c
fix download
[ { "path": "server/download.go", "patch": "@@ -144,7 +144,7 @@ func (b *blobDownload) run(ctx context.Context, requestURL *url.URL, opts *Regis\n \n \tfile.Truncate(b.Total)\n \n-\tg, _ := errgroup.WithContext(ctx)\n+\tg, inner := errgroup.WithContext(ctx)\n \tg.SetLimit(numDownloadParts)\n \tfor i := range ...
2023-10-12T19:52:35
nodejs/node
f88178906bd0e9678f70e8f95522192adc5b3872
971aad1b132a5ec37216c0d755d3ea59afc63812
doc: fix some links PR-URL: https://github.com/nodejs/node/pull/16202 Reviewed-By: Daijiro Wachi <daijiro.wachi@gmail.com> Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
[ { "path": "doc/api/assert.md", "patch": "@@ -47,10 +47,10 @@ Primitive values are compared with the [Abstract Equality Comparison][]\n Only [enumerable \"own\" properties][] are considered. The\n [`assert.deepEqual()`][] implementation does not test the\n [`[[Prototype]]`][prototype-spec] of objects or enum...
2017-10-14T14:08:22
golang/go
617f2c3e35cdc8483b950aa3ef18d92965d63197
b60ec4cc4b230f4d0787acf82057947b8bf80cea
net/http: mark http/2 connections active On Server.Shutdown, all idle connections are closed. A caveat for new connections is that they are marked idle after 5 seconds. Previously new HTTP/2 connections were marked New, and after 5 seconds, they would then become idle. With this change, we now mark HTTP/2 connections ...
[ { "path": "src/net/http/export_test.go", "patch": "@@ -274,6 +274,17 @@ func (s *Server) ExportAllConnsIdle() bool {\n \treturn true\n }\n \n+func (s *Server) ExportAllConnsByState() map[ConnState]int {\n+\tstates := map[ConnState]int{}\n+\ts.mu.Lock()\n+\tdefer s.mu.Unlock()\n+\tfor c := range s.activeConn...
2020-06-27T19:31:34
huggingface/transformers
30b453206d224ee5f747afa33ff216671558e6a0
ecfe9be7054e81f8841b8e97e6599e1a2d35ed7e
Enable multi-device for some models (#30207) * feat: multidevice for resnet * feat: yes! resnet * fix: compare all elements in tuple * feat: support for regnet * feat: support for convnextv2 * feat: support for bit * feat: support for cvt * feat: add support for focalnet * feat: support for yol...
[ { "path": "src/transformers/models/bit/modeling_bit.py", "patch": "@@ -658,6 +658,7 @@ class BitPreTrainedModel(PreTrainedModel):\n config_class = BitConfig\n base_model_prefix = \"bit\"\n main_input_name = \"pixel_values\"\n+ _no_split_modules = [\"BitEmbeddings\"]\n \n def _init_weights...
2024-04-19T08:24:44
ggml-org/llama.cpp
a554a1ecc75f6125c30f65fa18327ea68ef52ae6
0f2e42ca1d1d025e6c4cb4bffb78da8972dec17f
context : fix reserve token padding to n_seqs (#18536)
[ { "path": "src/llama-context.cpp", "patch": "@@ -1458,7 +1458,7 @@ ggml_cgraph * llama_context::graph_reserve(\n \n if (n_tokens % n_seqs != 0) {\n n_tokens = ((n_tokens + (n_seqs - 1)) / n_seqs) * n_seqs; // round to next multiple of n_seqs\n- n_outputs = std::min(n_outputs, n_tokens);\n...
2026-01-03T13:45:34
ollama/ollama
56497663c8bc7de73f4ff8d53235664a044c9d95
92578798bb1abcedd6bc99479d804f32d9ee2f6c
relay model runner error message to client (#720) * give direction to user when runner fails * also relay errors from timeout * increase timeout to 3 minutes
[ { "path": "cmd/cmd.go", "patch": "@@ -479,18 +479,7 @@ func generate(cmd *cobra.Command, model, prompt string, wordWrap bool) error {\n \t}\n \n \tif err := client.Generate(cancelCtx, &request, fn); err != nil {\n-\t\tif strings.Contains(err.Error(), \"failed to load model\") {\n-\t\t\t// tell the user to c...
2023-10-12T15:16:37
huggingface/transformers
d9850abd407821fb9f4ad5aabb206adad3610f75
01ae3b87c0c9ae5267a1f15edcc93715a78e1975
Fix `AssertionError` in clip conversion script (#30321) * fix * fix * fix * update comments --------- Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
[ { "path": "src/transformers/models/clip/convert_clip_original_pytorch_to_hf.py", "patch": "@@ -124,7 +124,15 @@ def convert_clip_checkpoint(checkpoint_path, pytorch_dump_folder_path, config_pa\n copy_vison_model_and_projection(hf_model, pt_model)\n hf_model.logit_scale = pt_model.logit_scale\n \n- ...
2024-04-18T18:18:02
ggml-org/llama.cpp
9dba9f5352308894bfb8786fcfe7c284168ff8f5
bcfc8c3cec68f8c3154e7e5d5a2efd807b241767
(Bugfix, ggml-cuda) Pool alloc count fix + small size computation type adjustment (#18559) * CUDA: Fixed obj byte size instead of obj count being passed to pool alloc (fattn-common, dst_tmp_meta) * CUDA: Explicitly casted some of the int alloc counts before multiplication in argsort --------- Co-authored-by: pl752 ...
[ { "path": "ggml/src/ggml-cuda/argsort.cu", "patch": "@@ -29,8 +29,8 @@ static void argsort_f32_i32_cuda_cub(ggml_cuda_pool & pool,\n const int nrows,\n ggml_sort_order order,\n cudaStream_t...
2026-01-03T10:13:40
golang/go
62fe10bf4e62c97af3bb8eb2ef72d9224a8752ba
bf833ead6250290dce039ffeee88f20a086b5dbe
src/go.mod, net/http: update bundled and latest golang.org/x/net Updates x/net/http2 to git rev 62affa334b73ec65ed44a326519ac12c421905e3 x/net/http2: reject HTTP/2 Content-Length headers containing a sign https://go-review.googlesource.com/c/net/+/236098/ (fixes #39017) also updates the vendored version of g...
[ { "path": "src/go.mod", "patch": "@@ -4,7 +4,7 @@ go 1.15\n \n require (\n \tgolang.org/x/crypto v0.0.0-20200622213623-75b288015ac9\n-\tgolang.org/x/net v0.0.0-20200822124328-c89045814202\n+\tgolang.org/x/net v0.0.0-20200904194848-62affa334b73\n \tgolang.org/x/sys v0.0.0-20200501145240-bc7a7d42d5c3 // indir...
2020-09-05T12:01:19
huggingface/transformers
01ae3b87c0c9ae5267a1f15edcc93715a78e1975
60d5f8f9f04026cb801d0dc5158bf4531e250072
Avoid `jnp` import in `utils/generic.py` (#30322) fix Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
[ { "path": "src/transformers/utils/generic.py", "patch": "@@ -38,10 +38,6 @@\n )\n \n \n-if is_flax_available():\n- import jax.numpy as jnp\n-\n-\n class cached_property(property):\n \"\"\"\n Descriptor that mimics @property but caches output in member variable.\n@@ -624,6 +620,8 @@ def transpose(...
2024-04-18T17:46:46
ggml-org/llama.cpp
bcfc8c3cec68f8c3154e7e5d5a2efd807b241767
18ddaea2aecf7fbfe7acab77465808f3cf6200d3
ggml-hexagon: optimize activation function (#18393) * refactor: refactor silu * refactor: optimize swiglu * refactor: remove unncessary if in swiglu * refactor: refactor swiglu_oai * chore: fix formatting issue
[ { "path": "ggml/src/ggml-hexagon/htp/act-ops.c", "patch": "@@ -85,13 +85,16 @@ static void glu_swiglu_fp32_per_thread(const struct htp_tensor * src0,\n struct htp_spad * dst_spad,\n uint32_t nth,\n ...
2026-01-03T05:24:24
nodejs/node
971aad1b132a5ec37216c0d755d3ea59afc63812
4826ac537f408304a8a208a2365d3bb415065c14
test: use relative path in pipePrefix Modified pipePrefix to use relative path on windows, previously tests failed when the full path was 120+ characters PR-URL: https://github.com/nodejs/node/pull/15988 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Luigi Pinca <luigipinca@gmail.com>
[ { "path": "test/common/index.js", "patch": "@@ -258,7 +258,8 @@ Object.defineProperty(exports, 'hasFipsCrypto', {\n });\n \n {\n- const pipePrefix = exports.isWindows ? '\\\\\\\\.\\\\pipe\\\\' : `${exports.tmpDir}/`;\n+ const localRelative = path.relative(process.cwd(), `${exports.tmpDir}/`);\n+ const pi...
2017-10-06T17:51:12
golang/go
721819e7bad078ac350bdebebb28793e13081e82
93810ac1f4574e1e2a79ea156781bafaf8b8ebe0
testing: make TempDir idempotent for both Cleanup and Benchmark Ensures that calling TempDir() in either of Cleanup or Benchmark doesn't cause test failures which were previously caused by the created directory having been deleted after the first run, yet we prevented the recreation of the directory due to our selecti...
[ { "path": "src/testing/testing.go", "patch": "@@ -413,10 +413,10 @@ type common struct {\n \tsignal chan bool // To signal a test is done.\n \tsub []*T // Queue of subtests to be run in parallel.\n \n-\ttempDirOnce sync.Once\n-\ttempDir string\n-\ttempDirErr error\n-\ttempDirSeq int32\n+\t...
2020-08-28T10:13:37
ollama/ollama
92578798bb1abcedd6bc99479d804f32d9ee2f6c
788637918ad76d2639114c5736e9d70b94758c4e
fix relative links in `README.md`
[ { "path": "README.md", "patch": "@@ -13,7 +13,7 @@ Get up and running with large language models locally.\n \n ### macOS\n \n-[Download](https://ollama.ai/download/Ollama-darwin.zip) \n+[Download](https://ollama.ai/download/Ollama-darwin.zip)\n \n ### Linux & WSL2\n \n@@ -23,7 +23,7 @@ curl https://ollama.a...
2023-10-11T23:24:06
huggingface/transformers
60d5f8f9f04026cb801d0dc5158bf4531e250072
c86d020ead7cabfc992fb766be1eead9768f7417
🚨🚨🚨Deprecate `evaluation_strategy` to `eval_strategy`🚨🚨🚨 (#30190) * Alias * Note alias * Tests and src * Rest * Clean * Change typing? * Fix tests * Deprecation versions
[ { "path": "docs/source/de/training.md", "patch": "@@ -128,12 +128,12 @@ Rufen Sie [`~evaluate.compute`] auf `metric` auf, um die Genauigkeit Ihrer Vorhe\n ... return metric.compute(predictions=predictions, references=labels)\n ```\n \n-Wenn Sie Ihre Bewertungsmetriken während der Feinabstimmung überwach...
2024-04-18T16:49:43
ggml-org/llama.cpp
5755e52d15d234579b84ace4ccd1b1cdd4596a9b
f38de163419d9381464b9185bca55c54207b3e43
model : Maincoder-1B support (#18534) * Add Maincoder model support * Removed SPM model vocabulary setting and MOE related GGUF parameters Removed trailing spaces from maincoder.cpp * removed set_vocab * added new line * Fix formatting * Add a new line for PEP8
[ { "path": "convert_hf_to_gguf.py", "patch": "@@ -6415,6 +6415,17 @@ def set_gguf_parameters(self):\n self.gguf_writer.add_head_count(0)\n \n \n+@ModelBase.register(\"MaincoderForCausalLM\")\n+class MaincoderModel(TextModel):\n+ model_arch = gguf.MODEL_ARCH.MAINCODER\n+\n+ def set_gguf_paramete...
2026-01-02T19:11:59
nodejs/node
4826ac537f408304a8a208a2365d3bb415065c14
24d72944da402dd17542f3c958d0f033c9e20a3a
test: replace fixtureDir with fixtures methods PR-URL: https://github.com/nodejs/node/pull/16114 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Gireesh Punathil <gpunathi@in.ibm.com>
[ { "path": "test/parallel/test-fs-read.js", "patch": "@@ -21,10 +21,10 @@\n \n 'use strict';\n const common = require('../common');\n+const fixtures = require('../common/fixtures');\n const assert = require('assert');\n-const path = require('path');\n const fs = require('fs');\n-const filepath = path.join(co...
2017-10-10T01:11:26
golang/go
93810ac1f4574e1e2a79ea156781bafaf8b8ebe0
612b1194475a23760ec502b48a93fea7237f3ae6
runtime: opportunistically rotate map key seed When clearing a map, reinitialize the hash seed with random data. This makes it more difficult for attackers to trigger pathological performance via repeated hash collisions. The extra reinitialization causes no statistically significant slowdown: name ...
[ { "path": "src/runtime/map.go", "patch": "@@ -993,6 +993,10 @@ func mapclear(t *maptype, h *hmap) {\n \th.noverflow = 0\n \th.count = 0\n \n+\t// Reset the hash seed to make it more difficult for attackers to\n+\t// repeatedly trigger hash collisions. See issue 25237.\n+\th.hash0 = fastrand()\n+\n \t// Keep...
2020-09-03T20:36:38
ollama/ollama
c413a5509313b6c9800d60ecf926a368f141d42b
630bb75d2acd2ae6bc5182e2f53dc8b27b7877b3
download: handle inner errors
[ { "path": "server/download.go", "patch": "@@ -31,11 +31,13 @@ type blobDownload struct {\n \n \tTotal int64\n \tCompleted atomic.Int64\n-\tdone bool\n \n \tParts []*blobDownloadPart\n \n \tcontext.CancelFunc\n+\n+\tdone bool\n+\terr error\n \treferences atomic.Int32\n }\n \n@@ -125,7 +...
2023-10-11T20:49:01
huggingface/transformers
c86d020ead7cabfc992fb766be1eead9768f7417
57b92bbfe5c703159e11836ac78bde4f5e32d782
Fix test transposing image with EXIF Orientation tag (#30319) * Fix test with exif_transpose image * Replace datasets with PIL to load image in tests
[ { "path": "tests/utils/test_image_utils.py", "patch": "@@ -16,11 +16,13 @@\n import os\n import tempfile\n import unittest\n+from io import BytesIO\n+from typing import Optional\n \n-import datasets\n import numpy as np\n import pytest\n-from huggingface_hub.file_download import http_get\n+import requests\n...
2024-04-18T16:41:20
ggml-org/llama.cpp
d84a6a98be9e952060cb58b1a824f82339b3391f
c6f0e832da05bd5342e39b0fb1026fcb3b6b8bf2
vocab : reduce debug logs about non-EOG control tokens (#18541) * vocab : reduce debug logs about non-EOG control tokens * cont : add comment
[ { "path": "src/llama-vocab.cpp", "patch": "@@ -2203,6 +2203,8 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {\n // for now, we apply this workaround to find the tokens based on their text\n \n for (const auto & t : token_to_id) {\n+ auto & attr ...
2026-01-02T14:17:33
golang/go
612b1194475a23760ec502b48a93fea7237f3ae6
e61d17d3b9ccf4c3e8ac87add9d74da7afa76488
cmd/link: pass darwin/amd64-specific flags only on AMD64 The linker assumed macOS is AMD64 (and 386 in the past). It passes darwin/amd64-specific flags to the external linker when building for macOS. They don't work for ARM64-based macOS. So only pass them on AMD64. Disable DWARF combining for macOS ARM64 for now. Th...
[ { "path": "src/cmd/link/internal/ld/lib.go", "patch": "@@ -1240,7 +1240,8 @@ func (ctxt *Link) hostlink() {\n \n \tswitch ctxt.HeadType {\n \tcase objabi.Hdarwin:\n-\t\tif machoPlatform == PLATFORM_MACOS {\n+\t\tif machoPlatform == PLATFORM_MACOS && ctxt.IsAMD64() {\n+\t\t\t// Leave room for DWARF combining...
2020-09-03T01:04:12
nodejs/node
24d72944da402dd17542f3c958d0f033c9e20a3a
4e835b334bd96d99436b86a867559ae050d66e97
test: remove error messages in crypto-binary test PR-URL: https://github.com/nodejs/node/pull/15981 Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Luigi Pinca <luigipinca@gmail.com>
[ { "path": "test/parallel/test-crypto-binary-default.js", "patch": "@@ -372,15 +372,13 @@ if (!common.hasFipsCrypto) {\n const a0 = crypto.createHash('md5').update('Test123').digest('latin1');\n assert.strictEqual(\n a0,\n- 'h\\u00ea\\u00cb\\u0097\\u00d8o\\fF!\\u00fa+\\u000e\\u0017\\u00ca\\u00bd\\...
2017-10-06T17:52:40
ollama/ollama
ab0668293cbfc2188b736d4c2b7dc0b7b997f5bf
af4cf55884ac54b9e637cd71dadfe9b7a5685877
llm: fix build on `amd64`
[ { "path": "llm/llama.cpp/generate_darwin_amd64.go", "patch": "@@ -15,4 +15,4 @@ package llm\n //go:generate git -C gguf apply ../patches/0001-remove-warm-up-logging.patch\n //go:generate cmake -S gguf -B gguf/build/cpu -DLLAMA_ACCELERATE=on -DLLAMA_K_QUANTS=on -DCMAKE_SYSTEM_PROCESSOR=x86_64 -DCMAKE_OSX_ARC...
2023-10-06T21:39:54
huggingface/transformers
68be1d3c1698eede18a4dc62e5c81814546a766b
791321451d771e43253673dbef772b5df3b24a78
fix Parameter dtype in audio models (#30310)
[ { "path": "src/transformers/models/data2vec/modeling_data2vec_audio.py", "patch": "@@ -822,7 +822,7 @@ def __init__(self, config: Data2VecAudioConfig):\n \n # model only needs masking vector if mask prob is > 0.0\n if config.mask_time_prob > 0.0 or config.mask_feature_prob > 0.0:\n- ...
2024-04-18T15:18:01
ggml-org/llama.cpp
e86f3c22211d9b5c3842e2961a022aac9cdbacad
169ee68ffb7281e7af632419513f55b6f14a8892
cuda : fix copy of large tensors (ggml_nbytes <= INT_MAX assertion) (#18433) * ggml-cuda: fixed assertion in ggml_cuda_cpy (#18140) * ggml-cuda: changes in data types to int64_t * ggml-cuda: added asserts for CUDA block numbers * ggml-cuda: changed the condition for y and z dimension
[ { "path": "ggml/src/ggml-cuda/cpy.cu", "patch": "@@ -12,11 +12,11 @@ const int CUDA_CPY_BLOCK_NM = 8; // block size of 3rd dimension if available\n const int CUDA_CPY_BLOCK_ROWS = 8; // block dimension for marching through rows\n \n template <cpy_kernel_t cpy_1>\n-static __global__ void cpy_scalar(con...
2026-01-01T23:24:20
golang/go
5402d40d5b041399392b29e4543f5fc4506197bd
0ef562592fe05b50b0ae8fce495ee7e2eec791f0
[dev.link] cmd/link: fix memory growth on dev.link CL 247399 caused memory growth in the linker. Fix this by adjusting how we preallocate the number of symbols we'll need. cmd/compile (Darwin), alloc/op: Loadlib_GC 33.5MB ± 0% 27.3MB ± 0% Change-Id: I34997329ea4412716114df97fc9dad6ad0c171ee Revi...
[ { "path": "src/cmd/link/internal/ld/lib.go", "patch": "@@ -543,7 +543,7 @@ func (ctxt *Link) loadlib() {\n \t}\n \n \t// Add non-package symbols and references of externally defined symbols.\n-\tctxt.loader.LoadNonpkgSyms(ctxt.Arch)\n+\tctxt.loader.LoadSyms(ctxt.Arch)\n \n \t// Load symbols from shared libr...
2020-08-18T20:35:26
nodejs/node
4e835b334bd96d99436b86a867559ae050d66e97
b9ad14d69d23c53bc65efb178c461f523b244ae1
test: use fixtures module over fixturesDir clean up code by using fixtures helper module instead of fixturesDir directly. PR-URL: https://github.com/nodejs/node/pull/15847 Reviewed-By: Ryan Graham <r.m.graham@gmail.com> Reviewed-By: Benjamin Gruenbaum <benjamingr@gmail.com> Reviewed-By: James M Snell <jasnell@gmail.c...
[ { "path": "test/sequential/test-deprecation-flags.js", "patch": "@@ -20,20 +20,21 @@\n // USE OR OTHER DEALINGS IN THE SOFTWARE.\n \n 'use strict';\n-const common = require('../common');\n+require('../common');\n+const fixtures = require('../common/fixtures');\n const assert = require('assert');\n const exe...
2017-10-06T16:53:41
huggingface/transformers
791321451d771e43253673dbef772b5df3b24a78
df96438484b62516689d67c00d4d9188f42e29ca
Fix: remove `pad token id` in pipeline forward arguments (#30285)
[ { "path": "src/transformers/pipelines/base.py", "patch": "@@ -888,11 +888,6 @@ def __init__(\n if self.model.can_generate():\n self.model.generation_config.update(**task_specific_params.get(task))\n \n- self.call_count = 0\n- self._batch_size = kwargs.pop(\"batch_si...
2024-04-18T14:31:32
ollama/ollama
af4cf55884ac54b9e637cd71dadfe9b7a5685877
d6786f294519f290b3a5a41d69ea669177491f46
not found error before pulling model (#718)
[ { "path": "server/routes.go", "patch": "@@ -3,8 +3,10 @@ package server\n import (\n \t\"context\"\n \t\"encoding/json\"\n+\t\"errors\"\n \t\"fmt\"\n \t\"io\"\n+\t\"io/fs\"\n \t\"log\"\n \t\"net\"\n \t\"net/http\"\n@@ -170,6 +172,11 @@ func GenerateHandler(c *gin.Context) {\n \n \tmodel, err := GetModel(req...
2023-10-06T20:06:20
nodejs/node
b9ad14d69d23c53bc65efb178c461f523b244ae1
371d1db60cad72f629992df20e347abab56fea5b
test: use common.fixtures module PR-URL: https://github.com/nodejs/node/pull/15992 Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de> Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Gireesh Punathil <gpunathi@in.ibm.com> Reviewed-By: Luigi Pinca <luigipinca@gmail.com>
[ { "path": "test/parallel/test-eval-require.js", "patch": "@@ -21,11 +21,12 @@\n \n 'use strict';\n const common = require('../common');\n+const { fixturesDir } = require('../common/fixtures');\n const assert = require('assert');\n const spawn = require('child_process').spawn;\n \n const options = {\n- cwd:...
2017-10-06T18:17:02
ggml-org/llama.cpp
ced765be44ce173c374f295b3c6f4175f8fd109b
3ccccc83f79c187e7fd3e19bdc5bf4efe8928697
model: support youtu-vl model (#18479) * Support Youtu-VL Model * merge code * fix bug * revert qwen2 code & support rsplit in minja.hpp * update warm info * fix annotation * u * revert minja.hpp * fix * Do not write routed_scaling_factor to gguf when routed_scaling_factor is None * fix expert_weights_scale ...
[ { "path": "convert_hf_to_gguf.py", "patch": "@@ -1233,6 +1233,9 @@ def get_vocab_base_pre(self, tokenizer) -> str:\n if chkhsh == \"4a2e2abae11ca2b86d570fc5b44be4d5eb5e72cc8f22dd136a94b37da83ab665\":\n # ref: https://huggingface.co/KORMo-Team/KORMo-tokenizer\n res = \"kormo\"...
2026-01-01T18:25:54
golang/go
2b8b06ebbf0198d3c7a9b4d839bc05d9b13ecbe7
6c76edeb1b67a5751dff215aaa712572d87a4ce8
cmd/internal/objabi: add regabi GOEXPERIMENT This is the "feature flag" for the register calling convention work (though since this work is expected to extend over a few releases, it's not version-prefixed). This will let us develop the register calling convention on the main branch while maintaining an easy toggle be...
[ { "path": "src/cmd/internal/objabi/util.go", "patch": "@@ -131,6 +131,11 @@ func init() {\n \t\t\taddexp(f)\n \t\t}\n \t}\n+\n+\t// regabi is only supported on amd64.\n+\tif GOARCH != \"amd64\" {\n+\t\tRegabi_enabled = 0\n+\t}\n }\n \n // Note: must agree with runtime.framepointer_enabled.\n@@ -161,6 +166,7...
2020-09-01T15:16:33
huggingface/transformers
df96438484b62516689d67c00d4d9188f42e29ca
ce8e64fbe22118d5b75ccf2da3e524a893ed2b99
Fix missing `prev_ci_results` (#30313) fix Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
[ { "path": ".github/workflows/slack-report.yml", "patch": "@@ -57,6 +57,15 @@ jobs:\n pip install slack_sdk\n pip show slack_sdk\n python utils/notification_service.py \"${{ inputs.folder_slices }}\"\n+\n+ # Upload complete failure tables, as they might be big and only trun...
2024-04-18T14:10:25
ggml-org/llama.cpp
d0a6a314702c8f84dd30a54a98f6b8d9810b02e6
2b2afade9fb38b8d699ed561d20a259561c00fc3
model : add support for JinaBertModel with non-gated ffn (#18475) * WIP: Initial commit for fixing JinaBert original FF type support * convert: add jina-v2-de tokenizer variant for German_Semantic_V3 * convert: fix token collision in BERT phantom vocab conversion * convert: add feed_forward_type metadata * model: ...
[ { "path": "convert_hf_to_gguf.py", "patch": "@@ -1062,6 +1062,9 @@ def get_vocab_base_pre(self, tokenizer) -> str:\n if chkhsh == \"66b8d4e19ab16c3bfd89bce5d785fb7e0155e8648708a1f42077cb9fe002c273\":\n # ref: https://huggingface.co/alvarobartt/grok-2-tokenizer\n res = \"grok-...
2026-01-01T17:38:51
ollama/ollama
288814d3e4b60a73cdfff80df9df2f572952aca3
04733438daa9ce274b3dd1d1cad66b51b43ef6e1
fix ref counts
[ { "path": "server/download.go", "patch": "@@ -36,7 +36,7 @@ type blobDownload struct {\n \n \tdone chan struct{}\n \tcontext.CancelFunc\n-\trefCount atomic.Int32\n+\treferences atomic.Int32\n }\n \n type blobDownloadPart struct {\n@@ -241,8 +241,19 @@ func (b *blobDownload) Write(p []byte) (n int, err error...
2023-10-03T23:44:35
golang/go
ef20f76b8bc4e082d5f81fd818890d707751475b
be9ed03f1aa5f348aa836c4ffe1904d8e37a629a
net/http: reject negative suffix-length Range:bytes=--N with 416 status code Fixes the file server to reject requests of the form: "Range": "bytes=--N" where "-N" is a negative suffix-length as designated by the grammar in RFC 7233 Section 2.1, "Byte-Ranges", which specifies that suffix-length MUST be of the form ...
[ { "path": "doc/go1.16.html", "patch": "@@ -167,3 +167,9 @@ <h3 id=\"minor_library_changes\">Minor changes to the library</h3>\n handler serves a 404 instead of its previous behavior of invoking the\n underlying handler with a mismatched <code>Path</code>/<code>RawPath</code> pair.\n </p>\n+\n+<p>\n+ The...
2020-09-02T08:08:02
huggingface/transformers
5728b5ad0071be8fa062f8b72c1345343d9b1a48
005b957fb851607115800f50e594b77662a771ba
FIX: Fixes unexpected behaviour for Llava / LLama & AWQ Fused modules + revert #30070 at the same time (#30317) * Update awq.py * style * revert felix PR * fix * add felix comments
[ { "path": "src/transformers/integrations/awq.py", "patch": "@@ -229,6 +229,8 @@ def fuse_awq_modules(model, quantization_config):\n else:\n raise ValueError(\"Fusing is only supported for the AutoAWQ backend\")\n \n+ fused_attention_modules = []\n+\n for name, module in model.named_module...
2024-04-18T13:51:17
rust-lang/rust
96264d2fc6bd8a54c9845ad0b0401a9681e0523b
328f453c2c0d431581b02bf2261f3fd037e48cef
docs: pull latest version from tags instead of releases in docs page (#6487) Fixes 6485 The releases only go back to `v1.6.0` since the team isn't releasing standalone binaries on GitHub releases.
[ { "path": "docs/index.html", "patch": "@@ -108,7 +108,6 @@\n </div>\n <script>\n const RustfmtTagsUrl = 'https://api.github.com/repos/rust-lang/rustfmt/tags';\n- const RustfmtLatestUrl = 'https://api.github.com/repos/rust-lang/rustfmt/releases/latest';\n co...
2025-02-26T16:16:23
ggml-org/llama.cpp
2b2afade9fb38b8d699ed561d20a259561c00fc3
f4f501925418ae38bb1e2d8c5e054c436686a782
convert : fix encoding of WPM vocab for BERT models (#18500) * convert: avoid token collision when stripping ## prefix * convert: use token types for BERT special tokens check * Update convert_hf_to_gguf.py Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com> --------- Co-authored-by: Sigbjørn Skjæret <...
[ { "path": "convert_hf_to_gguf.py", "patch": "@@ -5287,13 +5287,14 @@ def set_vocab(self):\n self.gguf_writer.add_token_type_count(self.hparams.get(\"type_vocab_size\", 1))\n \n # convert to phantom space vocab\n- def phantom(tok):\n- if tok.startswith(\"[\") and tok.endswit...
2026-01-01T17:27:07
ollama/ollama
711e891f0f6c2f36b5178c1a65177859eb0855b8
090d08422b361bcbef82a04d0e6e160caaad8f89
fix resumable downloads glob returns files in lexical order which is not appropriate when rebuilding the parts list
[ { "path": "server/download.go", "patch": "@@ -12,6 +12,7 @@ import (\n \t\"os\"\n \t\"path/filepath\"\n \t\"strconv\"\n+\t\"strings\"\n \t\"sync\"\n \t\"sync/atomic\"\n \t\"time\"\n@@ -39,9 +40,18 @@ type blobDownload struct {\n }\n \n type blobDownloadPart struct {\n+\tN int\n \tOffset int64\n \...
2023-10-02T22:26:27
golang/go
51c0bdc6d15dcd7f753c25896039ab41ac787ebb
e6583dc95375c4e266bffab6f8888e8e557b6355
testing: flush test summaries to stdout atomically when streaming output While debugging #40771, I realized that the chatty printer should only ever print to a single io.Writer (normally os.Stdout). The other Writer implementations in the chain write to local buffers, but if we wrote a test's output to a local buffer,...
[ { "path": "src/testing/benchmark.go", "patch": "@@ -242,7 +242,7 @@ func (b *B) run1() bool {\n \t\tif b.skipped {\n \t\t\ttag = \"SKIP\"\n \t\t}\n-\t\tif b.chatty && (len(b.output) > 0 || b.finished) {\n+\t\tif b.chatty != nil && (len(b.output) > 0 || b.finished) {\n \t\t\tb.trimOutput()\n \t\t\tfmt.Fprint...
2020-08-19T02:47:12