profile
viewpoint
If you are wondering where the data of this site comes from, please visit https://api.github.com/users/lgeiger/events. GitMemory does not store any data, but only uses NGINX to cache data for a period of time. The idea behind GitMemory is simply to give users a better reading experience.
Lukas Geiger lgeiger @plumerai @larq London www.linkedin.com/in/lgeiger deep learning scientist | astroparticle physicist | software engineer

atom-community/atom-script 718

:runner: Run ( scripts | selections | source ) in Atom

larq/larq 444

An Open-Source Library for Training Binarized Neural Networks

atom-community/ide-python 225

Python language support for Atom-IDE :atom: :snake:

lgeiger/black-action 167

A GitHub action that runs black code formatter for Python

lgeiger/addons 0

Useful extra functionality for TensorFlow maintained by SIG-addons

lgeiger/altair 0

Declarative statistical visualization library for Python

lgeiger/angular 0

One framework. Mobile & desktop.

lgeiger/anser 0

:tv: A low level parser for ANSI sequences.

lgeiger/aperture.js 0

A library for screen recording

lgeiger/apm 0

Atom Package Manager

fork rgbkrk/documentation

The source for Datadog's documentation site.

http://docs.datadoghq.com

fork in 3 hours

PR opened nteract/ansi-to-react

fix: add decoration to style-attribute

The createStyle function adds not only Colors to the attribute, but also Decoration. Examples:

bold

decoration === "bold"

=>

<span style="font-weight: bold"></span>

dim

decoration === "dim"

=>

<span style="opacity: 0.5"></span>

italic

decoration === "italic"

=>

<span style="font-style: italic"></span>

hidden

decoration === "hidden"

=>

<span style="visibility: hidden"></span>

strikethrough

decoration === "strikethrough"

=>

<span style="text-decoration: line-through"></span>

underline

decoration === "underline"

=>

<span style="text-decoration: underline"></span>

blink

decoration === "blink"

=>

<span style="text-decoration: blink"></span>
+35 -2

0 comment

1 changed file

pr created time in 8 hours

issue commentrupert/pyls-black

How to add to vim/nvim LanguageClient (or lsp-nvim)?

Thank you, it does work indeed. However I had to use vim.lsp.buf.formatting_sync(), else due to the async nature of formating, the modifications appear after the file has been written to disk, hence I had to save the file twice.

RustemB

comment created time in 9 hours

issue commentatom-community/ide-python

Uncaught Error: Header must provide a Content-Length property.

Could you give more information about the reproduction of the bug?

rugby8724

comment created time in a day

issue openedatom-community/ide-python

Uncaught Error: Header must provide a Content-Length property.

[Enter steps to reproduce:]

  1. ...
  2. ...

Atom: 1.57.0 x64 Electron: 9.4.4 OS: Mac OS X 10.15.7 Thrown From: ide-python package 1.9.3

Stack Trace

Uncaught Error: Header must provide a Content-Length property.

At /Users/jeremiahwise/.atom/packages/ide-python/node_modules/atom-languageclient/node_modules/vscode-jsonrpc/lib/common/messageReader.js:138

Error: Header must provide a Content-Length property.
    at StreamMessageReader.onData (/packages/ide-python/node_modules/atom-languageclient/node_modules/vscode-jsonrpc/lib/common/messageReader.js:138:27)
    at /packages/ide-python/node_modules/atom-languageclient/node_modules/vscode-jsonrpc/lib/common/messageReader.js:122:18)
    at Socket.emit (events.js:223:5)
    at addChunk (_stream_readable.js:309:12)
    at readableAddChunk (_stream_readable.js:290:11)
    at Socket.Readable.push (_stream_readable.js:224:10)
    at Pipe.onStreamRead (internal/stream_base_commons.js:181:23)

Commands

     -1:43.5.0 intentions:highlight (input.hidden-input)
     -1:43.3.0 core:save (input.hidden-input)
     -1:42.6.0 intentions:highlight (input.hidden-input)
     -1:42.6.0 core:save (input.hidden-input)
     -1:41.9.0 intentions:highlight (input.hidden-input)
     -1:41.8.0 core:save (input.hidden-input)
     -1:41.1.0 intentions:highlight (input.hidden-input)
     -1:41.1.0 core:save (input.hidden-input)
     -1:40.3.0 intentions:highlight (input.hidden-input)
     -1:40.1.0 core:save (input.hidden-input)
     -1:38.9.0 intentions:highlight (input.hidden-input)
     -1:38.8.0 core:save (input.hidden-input)
     -1:38.1.0 intentions:highlight (input.hidden-input)
     -1:38 core:save (input.hidden-input)
     -1:37.2.0 intentions:highlight (input.hidden-input)
     -1:37.1.0 core:save (input.hidden-input)

Non-Core Packages

atom-beautify 0.33.4 
atom-css-comb 3.5.1 
atom-html-preview 0.2.6 
atom-ide-ui 0.13.0 
atom-path-intellisense 1.2.2 
atom-ternjs 0.20.0 
autoclose-html 0.23.0 
autocomplete-html-entities 0.2.0 
busy-signal 2.0.1 
color-picker 2.3.0 
csslint 1.2.0 
django-atom 0.2.4 
emmet 2.4.3 
file-icons 2.1.47 
hyperclick 0.0.0 
ide-json 0.2.1 
ide-python 1.9.3 
intentions 2.1.1 
javascript-snippets 1.2.1 
language-ejs 0.4.0 
linter 3.4.0 
linter-eslint 8.6.6 
linter-ui-default 3.3.1 
pigments 0.40.6 
react 0.18.0 
Sublime-Style-Column-Selection 1.7.5 

created time in a day

push eventtensorflow/build

Austin Anderson

commit sha 04f31223dfe1a010d0ba191502a5cf3490cacbe5

Fix busted rename cleanup

view details

push time in a day

push eventtensorflow/build

Austin Anderson

commit sha acc5083b2bff9e3b530105f25b9538e58ed47b22

Remove defunct OpenLab builds OpenLab seems to have shut down. I can't get into contact with any of the original contacts, and the builds are out-of-date, so I'm removing them for now.

view details

push time in a day

push eventtensorflow/runtime

Jing Dong

commit sha 94a9cc13caf5aa8e2ba54937ed837279c11c78e4

Make DType an enum class and remove DType::Kind. PiperOrigin-RevId: 380285253

view details

push time in a day

push eventtensorflow/build

Austin Anderson

commit sha 92915cd024477687b7a75a56cbda6509004d6792

Update CONTRIBUTING.md

view details

push time in a day

push eventtensorflow/build

Marcin Juszkiewicz

commit sha 164b859d25d6c48d9a9dbd80cdb9d5573cd656e4

Linaro builds switched to manylinux2014 (#33) After several requests for package for Python newer than 3.6/3.7 I rewrote CI jobs to use manylinux2014 container image. We are now building 1.x, 2.4.x, 2.5.x and git HEAD versions. Released ones are built when needed and only for selected versions so you need to check separate directories for wheel files. Git HEAD is built daily.

view details

push time in a day

PR merged tensorflow/build

Reviewers
Linaro builds switched to manylinux2014

After several requests for package for Python newer than 3.6/3.7 I rewrote CI jobs to use manylinux2014 container image.

We are now building 1.x, 2.4.x, 2.5.x and git HEAD versions.

Released ones are built when needed and only for selected versions so you need to check separate directories for wheel files. Git HEAD is built daily.

+2 -2

3 comments

1 changed file

hrw

pr closed time in a day

push eventtensorflow/runtime

Jing Dong

commit sha db9b478d4cf4dcbc396797594e214df3089f33db

Move IsValid(), IsInvalid(), and IsUnsupported() of DType as free functions. This is in preparation to remove the redundant DType and DType::Kind. PiperOrigin-RevId: 380263902

view details

push time in a day

push eventtensorflow/runtime

Jing Dong

commit sha 89f53cff21bbc3e3d7d528c8f497ece049b2a42a

Move GetHostSize() and GetAlignment() to be global functions. This is in preparation to remove the redundant DType and DType::Kind. PiperOrigin-RevId: 380256224

view details

push time in a day

release vercel/next.js

v11.0.1-canary.5

released time in a day

push eventtensorflow/runtime

Jing Dong

commit sha aa0ffad13fd49a6f17d5b3252d810dd75e69c4d5

Add missing header in dtype.cc. The missing header was causing failure in android_arm64 build. PiperOrigin-RevId: 380243576

view details

push time in a day

Pull request review commenttensorflow/runtime

Changes required to build TFRT on Windows

 filegroup(     ], ) +PYBIND11_COPTS = select({+    "@tf_runtime//:windows": [

Yes, this is one place where we could use TFRT_OPTS. I was thinking to keep that factoring out in a follow up PR, since there may be other changes (including potential differences in flags).

tlemo

comment created time in a day

Pull request review commenttensorflow/runtime

Changes required to build TFRT on Windows

 win32_cmake_vars = { # TODO(phawkins): use a better method to select the right host triple, rather

I tried to compare & sync the two versions. I think the TF has everything, and I only ported over the required bits (TF has a number of extra configs which are not defined in TFRT)

tlemo

comment created time in a day

Pull request review commenttensorflow/runtime

Changes required to build TFRT on Windows

 namespace tfrt {  void* AlignedAlloc(size_t alignment, size_t size) {-  if (alignment <= alignof(void*)) return std::malloc(size);   size = (size + alignment - 1) / alignment * alignment;--#if defined(__ANDROID__) || defined(OS_ANDROID)+#ifdef _WIN32+  return _aligned_malloc(size, alignment);+#elif defined(__ANDROID__) || defined(OS_ANDROID)

MSVC doesn't support aligned_alloc()

https://docs.microsoft.com/en-us/cpp/standard-library/cstdlib?view=msvc-160 (search for aligned_alloc)

tlemo

comment created time in a day

Pull request review commenttensorflow/runtime

Changes required to build TFRT on Windows

 struct AlignedAllocator {    using value_type = T; +  constexpr AlignedAllocator() noexcept = default;++  constexpr AlignedAllocator(const AlignedAllocator&) noexcept = default;++  template <class Other>+  constexpr AlignedAllocator(const AlignedAllocator<Other>&) noexcept {}+   T* allocate(size_t n) {     auto* ptr = AlignedAlloc(Align, n * sizeof(T));     return static_cast<T*>(ptr);   }-  void deallocate(T* p, size_t) noexcept { std::free(p); }+  void deallocate(T* p, size_t) noexcept { AlignedFree(p); }

Good questions!

The noexcept was there in the original code and I left it alone, although it's not really required for conformance to the allocator interface. Since you mentioned, it's better to remove it (to avoid suboptimal code generation in this case)

Regarding the need for AlignedFree() vs. just calling std::free(): On may unix implementation, the aligned allocations can be freed with free(). This is not always the case, and in particular with the MSVC runtime, _aligned_malloc must be paired with _aligned_free() (trying to free aligned blocks with free() results in nasty runtime issues)

Finally, the additional constructors are a required part of the allocator interface. MSVC debug runtime needs these constructors for implementing some of the debug checks.

tlemo

comment created time in a day

Pull request review commenttensorflow/runtime

Changes required to build TFRT on Windows

 namespace tfrt { class ExecutionContext;  // This is a simple representation of a decoded diagnostic.-struct DecodedDiagnostic {+class DecodedDiagnostic {

Yes. MSVC requires consistent struct/class usage across declarations and definitions.

This is technically a MSVC bug, but with a long history and unlikely to get fixed (the struct/class result in different name name mangling and may lead to painful to debug linking issues)

tlemo

comment created time in a day

Pull request review commenttensorflow/runtime

Changes required to build TFRT on Windows

 class AsyncValue {   const bool is_refcounted_ : 1;    // Unused padding bits.-  unsigned unused_ : 4;+  // unsigned unused_ : 4;

Yes, although I should really remove the line instead of commenting it out.

Long version: the original code is not portable - bit fields packing is implementation defined, and some compilers (notably MSVC) will not pack together bitfields with different base types.

tlemo

comment created time in a day

Pull request review commenttensorflow/runtime

Changes required to build TFRT on Windows

 class BEFKernel {   // num_used_bys as the number of results is not fixed.) in BEF. kernel_code,   // kernel_location, num_arguments, num_attributes, num_functions, and   // num_results in BEF can be directly mapped using this struct.-  struct __attribute__((packed)) BEFKernelHeader {+  PACKED_STRUCT(BEFKernelHeader {

I belive you're right. Is the suggestion to use that explicitly and do without the PACKED_STRUCT macro?

tlemo

comment created time in a day

Pull request review commenttensorflow/runtime

Changes required to build TFRT on Windows

 def clean_dep(dep):     return str(Label(dep)) +def if_oss(oss_value, google_value = []):+    """Returns one of the arguments based on the non-configurable build env.++    Specifically, it does not return a `select`, and can be used to e.g.+    compute elements of list attributes.+    """+    return oss_value  # copybara:comment_replace return google_value++def if_google(google_value, oss_value = []):+    """Returns one of the arguments based on the non-configurable build env.++    Specifically, it does not return a `select`, and can be used to e.g.+    compute elements of list attributes.+    """+    return oss_value  # copybara:comment_replace return google_value+ TFRT_COPTS = select({

I'd be happy to, but:

  1. TFRT_COPTS (and other TFRT_xxx variables) preceded my changes, would you like to change them all?
  2. I'm considering reusing TFRT_COPTS to factor out common flags and use it in other build files in a follow up iteration

What do you think?

tlemo

comment created time in a day

Pull request review commenttensorflow/runtime

Changes required to build TFRT on Windows

 tfrt_cc_library(         "lib/io/buffered_input_stream.cc",         "lib/io/file_input_stream.cc",         "lib/io/file_system.cc",-        "lib/io/posix_file_system.cc",-        "lib/io/posix_file_system.h",-    ],+    ] + select({+        "@tf_runtime//:windows": [

would that work ok when using TFRT in a "parent" workspace? (ex. TF)

I see most place use the fully qualified @tf_runtime and I tried to do the same. If I aimed for the wrong consistency please let me know and I'll make the changes.

tlemo

comment created time in a day

Pull request review commenttensorflow/runtime

Changes required to build TFRT on Windows

 config_setting(     visibility = ["//visibility:public"], ) +config_setting(+    name = "windows",+    # Internal builds query the target OS.+    flag_values = if_google(

I tried to keep the formatting reasonable, but ultimately I imagine that the round-trip through google3 would autoformat all the files, is that not the case?

tlemo

comment created time in a day

Pull request review commenttensorflow/runtime

Changes required to build TFRT on Windows

 build:nvcc --@rules_cuda//cuda:compiler=nvcc build:nvcc --repo_env=BAZEL_CXXOPTS=  # Prevent warning: 'std' redefinition. build:nvcc --@rules_cuda//cuda:copts=--expt-relaxed-constexpr -# Disable warnings we don't care about.-build --copt=-Wno-unused-local-typedef- # C++14 standard version is required.-build --cxxopt=-std=c++14 --host_cxxopt=-std=c++14--# Necessary linker options.-build --linkopt=-pthread-build --linkopt=-ldl-build --linkopt=-lm-build --host_linkopt=-pthread-build --host_linkopt=-ldl-build --host_linkopt=-lm+build:android --cxxopt=-std=c++14

I agree, per-toolchain is better than per-platform.

As a side note - today Windows is assumed to mean MSVC, but it may be worth considering a Windows build with Clang.

tlemo

comment created time in a day

PR opened tensorflow/community

Remove cpp style guide

TF now (as of #60fcc4cb35748b9b201e091e0c4edd9709a17f0a) supports StatusOr, in line with the Google C++ style guide. The exception no longer needs to be documented.

+0 -64

0 comment

1 changed file

pr created time in a day

push eventtensorflow/build

Austin Anderson

commit sha fe2642d35a0bbf84a77cb634a9e151a2fdadeb4a

Make wheel verification support multiple wheels

view details

push time in a day

fork teonbrooks/nimbus-sdk

Cross Platform Rapid Experiments "Nimbus" SDK

fork in a day

created tagnteract/ipypandex

tag0.2.1

A package for automatically turning on Data Explorer in Pandas for an IPython Jupyter kernel.

created time in a day