Update all non-major dependencies #23
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
![renovate[bot]](https://avatars.githubusercontent.com/in/2740?s=60&v=4){kind=small}
renovate
bot
force-pushed
the
renovate/all-minor-patch
branch
from
8a4afd8 to
f7d4a53
Compare
renovate
bot
changed the title
renovate
bot
force-pushed
the
renovate/all-minor-patch
branch
2 times, most recently
from
47ad09b to
82bf05b
Compare
renovate
bot
changed the title
renovate
bot
force-pushed
the
renovate/all-minor-patch
branch
5 times, most recently
from
ad528b0 to
89108f5
Compare
renovate
bot
force-pushed
the
renovate/all-minor-patch
branch
2 times, most recently
from
c86e3aa to
89b9e47
Compare
renovate
bot
force-pushed
the
renovate/all-minor-patch
branch
from
89b9e47 to
5dab9ed
Compare
abdolence
approved these changes
Sep 1, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR contains the following updates:
1.38.0->1.40.01.44.0->1.47.00.8.0->0.8.10.12.5->0.12.70.10->0.130.10->0.131.0.208->1.0.2091.0.122->1.0.1271.39.2->1.40.02.1.6->2.2.0Release Notes
robertknight/ocrs (ocrs)
v0.8.1Compare Source
Added ability to customize the alphabet used by the recognition model
https://github.com/robertknight/ocrs/pull/100/100). Thanks @Phaired.
Updated rten to v0.13.1. This enables running custom models in the V2
rten model format
seanmonstar/reqwest (reqwest)
v0.12.7Compare Source
impl Service<http::Request<_>>forClient.v0.12.6Compare Source
danger_accept_invalid_hostnamesforrustls.impl Service<http::Request<Body>>forClientand&'_ Client.!Syncbodies inBody::wrap_stream().hickory-dnsis used.Proxyso thatHTTP(S)_PROXYvalues take precendence overALL_PROXY.blocking::RequestBuilder::header()from unsettingsensitiveon passed header values.robertknight/rten (rten)
v0.13.1Compare Source
rten
New features
Added speech detection example using Silero VAD
https://github.com/robertknight/rten/pull/338/338)
Support int tensors in ArgMin and ArgMax ops
https://github.com/robertknight/rten/pull/329/329)
Support "reflect" padding mode (https://github.com/robertknight/rten/pull/326)
Bug fixes
Fixed panic with certain combinations of input, kernel size and padding in
depthwise convolutihttps://github.com/robertknight/rten/pull/336/336)
Fixed attempted out-of-bounds slice in depthwise convolution when input tensor
has a row stride that exceeds the row lenghttps://github.com/robertknight/rten/pull/335pull/335)
Fixed conversion of
auto_padattribute for Conv operatorhttps://github.com/robertknight/rten/pull/333/333)
Round timings to microseconds in verbose log
https://github.com/robertknight/rten/pull/331/331)
Fixed panic when slicing empty tensors
https://github.com/robertknight/rten/pull/325/325)
Fixed 1D convolution failing with non-contiguous inputs
https://github.com/robertknight/rten/pull/324/324)
Fixed conversion of shape information for scalar tensors
https://github.com/robertknight/rten/pull/323/323)
Fixed panic in softmax if the size of the normalized axis is zero
https://github.com/robertknight/rten/pull/322/322)
rten-cli
--mmapflag to load model using memory mapping instead of readingwhole file into a buffhttps://github.com/robertknight/rten/pull/330/330)
v0.13.0Compare Source
This release adds the infrastructure to support subgraphs, which are used in
control flow operators like
If, plus an implementation of theIfoperatorand a TrOCR example which uses it.
rten
Added
Model::load_static_sliceAPI which can be used to load models embedded in the binary with
include_bytes!. Thanks @hsfzxjy.Added TrOCR example (https://github.com/robertknight/rten/pull/304)
Support
Ifoperator (https://github.com/robertknight/rten/pull/306)Added full support for
Einsumoperator (https://github.com/robertknight/rten/pull/297,https://github.com/robertknight/rten/pull/299/2https://github.com/robertknight/rten/pull/300/3https://github.com/robertknight/rten/pull/302puhttps://github.com/robertknight/rten/pull/303pull/303)
rten-cli
Added
--quietflag (https://github.com/robertknight/rten/pull/313)Inputs named
use_cache_branchnow get a default value of0(ddf4109)rten-generate
Support models with cross-attention KV caches that are computed on the first
run of the decodhttps://github.com/robertknight/rten/pull/318/318). This
is used by Hugging Face models for encoder-decoder systems.
Support models without a KV cache (https://github.com/robertknight/rten/pull/305)
rten-tensor
Tensor::remove_axis(b823d46)Tensor::from_storage_and_layout(54d2941)rten-text
vocabulary which are never generated by merges and are not added special
tokens (
18e9b2a)v0.12.0Compare Source
rten
Breaking changes
The
rten-converttool now generates models in the V2 format by defaulthttps://github.com/robertknight/rten/pull/272/272).
These models can only be loaded by RTen version 0.11.0 or later. The V1
format can be generated by specifying the
--v1flag. Thertencrate canload both V1 and V2 format models.
See the
.rtenfile format documentationfor more details.
The
reduce_{max, min, sum}tensor methods have moved from theFloatOperatorstrait to theOperatorstrait (https://github.com/robertknight/rten/pull/274).Examples and documentation
Added Segment Anything example (https://github.com/robertknight/rten/pull/295).
This supports the original SAM models plus several derivatives with
lighter-weight image encoders.
Added chatbot example using Qwen2 (https://github.com/robertknight/rten/pull/282).
This also works with SmolLM.
Model::load_mmapdocs now have a better explanation of the memory andperformance impact (
ce0b717)New features
Einsumoperator (https://github.com/robertknight/rten/pull/295).Performance improvements
Avoid allocations in most cases when broadcasting tensor shapes (
c4b5f26).Strides of size-1 dimensions are ignored when determining whether a tensor is
contiguohttps://github.com/robertknight/rten/pull/292/292). This allows more
operations to use fast paths for contiguous tensors.
Optimized
LayerNormalizationandReduceMean(https://github.com/robertknight/rten/pull/291)Added fast-path for
Resizeoperator when input scale is 1 (https://github.com/robertknight/rten/pull/290)Return input buffer to pool in
Castoperator if input needs to be copiedhttps://github.com/robertknight/rten/pull/289/289).
Implemented
LayerNormalizationfusion (https://github.com/robertknight/rten/pull/280)Implemented
GELUfusion (https://github.com/robertknight/rten/pull/277)rten-cli
*_idsnow use zero as theauto-generated input value (
78cd621)rten-generate
TopKSamplernow supports specifying a temperature (65b837b)Added
Generator::append_promptto append to prompt after initial generation.This is useful for chat-like applications (
5ef3cb2)Fixed an issue where
attention_maskinput had the wrong size (cae6134)rten-tensor
Breaking changes
tensorandndtensormacros have been deprecated in favor ofTensor::fromandNdTensor::from(https://github.com/robertknight/rten/pull/286).Other changes
Tensor::fromnow supports creating tensors from scalar values (d2ca876)Tensor::lanesiterator performance was improved by making them exact-sizedand fused (
9e31556)rten-text
Token IDs are now represented as
u32rather thanusize, for consistencywith rten-generahttps://github.com/robertknight/rten/pull/288/288).
The
vocabmapping intokenizer.jsonfiles is now used to determine tokenIDs when decodihttps://github.com/robertknight/rten/pull/287/287).
v0.11.1Compare Source
rten
Instant::nowhttps://github.com/robertknight/rten/pull/283/283).
v0.11.0Compare Source
rten
Breaking changes
The
inputsargument toModel::runnow accepts aVec<(NodeId, InputOrOutput)>instead of&[(NodeId, Input)], whereInputOrOutputis anenum that is either an owned
Tensoror aTensorView. This enables passingownership of an input to
Model::run, which is in turn enables efficientin-place updates to cache-like inputs.
The
InputOrOutputtype implementsFromfor tensors and tensor views, socode such as:
Becomes:
New features
Add a new version of the
.rtenfile format which supports models over 2GBin size. The
rten-converttool still generates V1 models by default butwill generate the V2 format if the
--v2flag is providedhttps://github.com/robertknight/rten/pull/260/260).
Support
Geluoperator (https://github.com/robertknight/rten/pull/248)Bug fixes
Prevent
Model::partial_runfrom propagating values through randomizedoperatohttps://github.com/robertknight/rten/pull/240/240).
Improved accuracy of timing metrics and eliminated unaccounted for
("[Other]") https://github.com/robertknight/rten/pull/254l/254.
Performance improvements
This release adds a new graph optimization step as part of loading models. This
performs fusions and other optimizations to speed up inference. These
optimizations are enabled by default, but can be disabled via options in
ModelOptions.Improved parallelism in the
Softmaxoperator (https://github.com/robertknight/rten/pull/258)Made
Tensor::inner_iterfaster (https://github.com/robertknight/rten/pull/259)Made
Gather,ConcatandUnsqueezeoperators faster for small inputs.These operations are common in subgraphs that operator on tensor shaphttps://github.com/robertknight/rten/pull/255puhttps://github.com/robertknight/rten/pull/256tehttps://github.com/robertknight/rten/pull/257ht/rten/pull/257.
Optimized vector-matrix multiplication (https://github.com/robertknight/rten/pull/250,
https://github.com/robertknight/rten/pull/253/253). This benefits transformer
decoder inference when the batch size is 1.
Fuse
Mul(X, Sigmoid(X))subgraphs into aSiluoperation. This speeds upYOLOv8 by 8%. https://github.com/robertknight/rten/pull/246/246.
Further reduce small allocations during graph execution
https://github.com/robertknight/rten/pull/243/2https://github.com/robertknight/rten/pull/245pull/245).
Fuse
MatMul(Transpose(X), Y)subgraphs to avoid materializing the transposedmatrhttps://github.com/robertknight/rten/pull/242/242).
Perform constant propagation when loading models
https://github.com/robertknight/rten/pull/241/241).
Enabled
Concatoperator to run in-place if the caller has specificallyreserved space in the first input's buffhttps://github.com/robertknight/rten/pull/239pull/239).
Cache the last-used execution plan. This avoids recomputing the sequence of
execution steps when a model is run in a lohttps://github.com/robertknight/rten/pull/234pull/234).
Improved performance of unary operators for non-contiguous inputs
https://github.com/robertknight/rten/pull/223/223)
Optimized
Whereoperator for non-contiguous inputshttps://github.com/robertknight/rten/pull/213/213)
Optimized variadic operators (https://github.com/robertknight/rten/pull/212)
Optimized
Powoperator (https://github.com/robertknight/rten/pull/219)rten-examples
rten-generate
This is a new crate which provides a convenient
Iterator-based interface forrunning auto-regressive decoder models. See the
gpt2anddistilvitexamplesin the
rten-examplescrate for code samples.rten-tensor
NdTensor::fromhttps://github.com/robertknight/rten/pull/226/226).
rten-text
serde-rs/serde (serde)
v1.0.209Compare Source
serde-rs/json (serde_json)
v1.0.127Compare Source
v1.0.126Compare Source
v1.0.125Compare Source
v1.0.124Compare Source
v1.0.123Compare Source
tokio-rs/tokio (tokio)
v1.40.0: Tokio v1.40.0Compare Source
1.40.0 (August 30th, 2024)
Added
util::SimplexStream(#6589)Command::process_group(#6731){TrySendError,SendTimeoutError}::into_inner(#6755)JoinSet::join_all(#6784)Added (unstable)
Builder::{on_task_spawn, on_task_terminate}(#6742)Changed
write_all_bufwhen possible (#6724)UnwindSafe(#6783)SleepandBatchSemaphoreinstrumentation explicit roots (#6727)NonZeroU64fortask::Id(#6733)JoinError(#6753)#[must_use]toJoinHandle::abort_handle(#6762)Documented
[build]section doesn't go in Cargo.toml (#6728)select!(#6774)v1.39.3: Tokio v1.39.3Compare Source
1.39.3 (August 17th, 2024)
This release fixes a regression where the unix socket api stopped accepting the abstract socket namespace. (#6772)
zip-rs/zip2 (zip)
v2.2.0Compare Source
🚀 Features
ZipArchive::central_directory_start(#232)Configuration
📅 Schedule: Branch creation - At any time (no schedule defined), Automerge - At any time (no schedule defined).
🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.
♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.
👻 Immortal: This PR will be recreated if closed unmerged. Get config help if that's undesired.
This PR was generated by Mend Renovate. View the repository job log.