* Add spacy.TextCatParametricAttention.v1
This layer provides is a simplification of the ensemble classifier that
only uses paramteric attention. We have found empirically that with a
sufficient amount of training data, using the ensemble classifier with
BoW does not provide significant improvement in classifier accuracy.
However, plugging in a BoW classifier does reduce GPU training and
inference performance substantially, since it uses a GPU-only kernel.
* Fix merge fallout
* Add TextCatReduce.v1
This is a textcat classifier that pools the vectors generated by a
tok2vec implementation and then applies a classifier to the pooled
representation. Three reductions are supported for pooling: first, max,
and mean. When multiple reductions are enabled, the reductions are
concatenated before providing them to the classification layer.
This model is a generalization of the TextCatCNN model, which only
supports mean reductions and is a bit of a misnomer, because it can also
be used with transformers. This change also reimplements TextCatCNN.v2
using the new TextCatReduce.v1 layer.
* Doc fixes
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Fully specify `TextCatCNN` <-> `TextCatReduce` equivalence
* Move TextCatCNN docs to legacy, in prep for moving to spacy-legacy
* Add back a test for TextCatCNN.v2
* Replace TextCatCNN in pipe configurations and templates
* Add an infobox to the `TextCatReduce` section with an `TextCatCNN` anchor
* Add last reduction (`use_reduce_last`)
* Remove non-working TextCatCNN Netlify redirect
* Revert layer changes for the quickstart
* Revert one more quickstart change
* Remove unused import
* Fix docstring
* Fix setting name in error message
---------
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Update `TextCatBOW` to use the fixed `SparseLinear` layer
A while ago, we fixed the `SparseLinear` layer to use all available
parameters: https://github.com/explosion/thinc/pull/754
This change updates `TextCatBOW` to `v3` which uses the new
`SparseLinear_v2` layer. This results in a sizeable improvement on a
text categorization task that was tested.
While at it, this `spacy.TextCatBOW.v3` also adds the `length_exponent`
option to make it possible to change the hidden size. Ideally, we'd just
have an option called `length`. But the way that `TextCatBOW` uses
hashes results in a non-uniform distribution of parameters when the
length is not a power of two.
* Replace TexCatBOW `length_exponent` parameter by `length`
We now round up the length to the next power of two if it isn't
a power of two.
* Remove some tests for TextCatBOW.v2
* Fix missing import
* initial
* initial documentation run
* fix typo
* Remove mentions of Torchscript and quantization
Both are disabled in the initial release of `spacy-curated-transformers`.
* Fix `piece_encoder` entries
* Remove `spacy-transformers`-specific warning
* Fix duplicate entries in tables
* Doc fixes
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Remove type aliases
* Fix copy-paste typo
* Change `debug pieces` version tag to `3.7`
* Set curated transformers API version to `3.7`
* Fix transformer listener naming
* Add docs for `init fill-config-transformer`
* Update CLI command invocation syntax
* Update intro section of the pipeline component docs
* Fix source URL
* Add a note to the architectures section about the `init fill-config-transformer` CLI command
* Apply suggestions from code review
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Update CLI command name, args
* Remove hyphen from the `curated-transformers.mdx` filename
* Fix links
* Remove placeholder text
* Add text to the model/tokenizer loader sections
* Fill in the `DocTransformerOutput` section
* Formatting fixes
* Add curated transformer page to API docs sidebar
* More formatting fixes
* Remove TODO comment
* Remove outdated info about default config
* Apply suggestions from code review
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Add link to HF model hub
* `prettier`
---------
Co-authored-by: Madeesh Kannan <shadeMe@users.noreply.github.com>
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
SpaCy's HashEmbedCNN layer performs convolutions over tokens to produce
contextualized embeddings using a `MaxoutWindowEncoder` layer. These
convolutions are implemented using Thinc's `expand_window` layer, which
concatenates `window_size` neighboring sequence items on either side of
the sequence item being processed. This is repeated across `depth`
convolutional layers.
For example, consider the sequence "ABCDE" and a `MaxoutWindowEncoder`
layer with a context window of 1 and a depth of 2. We'll focus on the
token "C". We can visually represent the contextual embedding produced
for "C" as:
```mermaid
flowchart LR
A0(A<sub>0</sub>)
B0(B<sub>0</sub>)
C0(C<sub>0</sub>)
D0(D<sub>0</sub>)
E0(E<sub>0</sub>)
B1(B<sub>1</sub>)
C1(C<sub>1</sub>)
D1(D<sub>1</sub>)
C2(C<sub>2</sub>)
A0 --> B1
B0 --> B1
C0 --> B1
B0 --> C1
C0 --> C1
D0 --> C1
C0 --> D1
D0 --> D1
E0 --> D1
B1 --> C2
C1 --> C2
D1 --> C2
```
Described in words, this graph shows that before the first layer of the
convolution, the "receptive field" centered at each token consists only
of that same token. That is to say, that we have a receptive field of 1.
The first layer of the convolution adds one neighboring token on either
side to the receptive field. Since this is done on both sides, the
receptive field increases by 2, giving the first layer a receptive field
of 3. The second layer of the convolutions adds an _additional_
neighboring token on either side to the receptive field, giving a final
receptive field of 5.
However, this doesn't match the formula currently given in the docs,
which read:
> The receptive field of the CNN will be
> `depth * (window_size * 2 + 1)`, so a 4-layer network with a window
> size of `2` will be sensitive to 20 words at a time.
Substituting in our depth of 2 and window size of 1, this formula gives
us a receptive field of:
```
depth * (window_size * 2 + 1)
= 2 * (1 * 2 + 1)
= 2 * (2 + 1)
= 2 * 3
= 6
```
This not only doesn't match our computations from above, it's also an
even number! This is suspicious, since the receptive field is supposed
to be centered on a token, and not between tokens. Generally, this
formula results in an even number for any even value of `depth`.
The error in this formula is that the adjustment for the center token
is multiplied by the depth, when it should occur only once. The
corrected formula, `depth * window_size * 2 + 1`, gives the correct
value for our small example from above:
```
depth * window_size * 2 + 1
= 2 * 1 * 2 + 1
= 4 + 1
= 5
```
These changes update the docs to correct the receptive field formula and
the example receptive field size.
* Support custom token/lexeme attribute for vectors
* Fix imports
* Back off to ORTH without Vectors.attr
* Fallback if vectors.attr doesn't exist
* Update docs
* Make empty_kb() configurable.
* Format.
* Update docs.
* Be more specific in KB serialization test.
* Update KB serialization tests. Update docs.
* Remove doc update for batched candidate generation.
* Fix serialization of subclassed KB in tests.
* Format.
* Update docstring.
* Update docstring.
* Switch from pickle to json for custom field serialization.
* Rename all MDX file to `.mdx`
* Lock current node version (#11885)
* Apply Prettier (#11996)
* Minor website fixes (#11974) [ci skip]
* fix table
* Migrate to Next WEB-17 (#12005)
* Initial commit
* Run `npx create-next-app@13 next-blog`
* Install MDX packages
Following: 77b5f79a4d/packages/next-mdx/readme.md
* Add MDX to Next
* Allow Next to handle `.md` and `.mdx` files.
* Add VSCode extension recommendation
* Disabled TypeScript strict mode for now
* Add prettier
* Apply Prettier to all files
* Make sure to use correct Node version
* Add basic implementation for `MDXRemote`
* Add experimental Rust MDX parser
* Add `/public`
* Add SASS support
* Remove default pages and styling
* Convert to module
This allows to use `import/export` syntax
* Add import for custom components
* Add ability to load plugins
* Extract function
This will make the next commit easier to read
* Allow to handle directories for page creation
* Refactoring
* Allow to parse subfolders for pages
* Extract logic
* Redirect `index.mdx` to parent directory
* Disabled ESLint during builds
* Disabled typescript during build
* Remove Gatsby from `README.md`
* Rephrase Docker part of `README.md`
* Update project structure in `README.md`
* Move and rename plugins
* Update plugin for wrapping sections
* Add dependencies for plugin
* Use plugin
* Rename wrapper type
* Simplify unnessary adding of id to sections
The slugified section ids are useless, because they can not be referenced anywhere anyway. The navigation only works if the section has the same id as the heading.
* Add plugin for custom attributes on Markdown elements
* Add plugin to readd support for tables
* Add plugin to fix problem with wrapped images
For more details see this issue: https://github.com/mdx-js/mdx/issues/1798
* Add necessary meta data to pages
* Install necessary dependencies
* Remove outdated MDX handling
* Remove reliance on `InlineList`
* Use existing Remark components
* Remove unallowed heading
Before `h1` components where not overwritten and would never have worked and they aren't used anywhere either.
* Add missing components to MDX
* Add correct styling
* Fix broken list
* Fix broken CSS classes
* Implement layout
* Fix links
* Fix broken images
* Fix pattern image
* Fix heading attributes
* Rename heading attribute
`new` was causing some weird issue, so renaming it to `version`
* Update comment syntax in MDX
* Merge imports
* Fix markdown rendering inside components
* Add model pages
* Simplify anchors
* Fix default value for theme
* Add Universe index page
* Add Universe categories
* Add Universe projects
* Fix Next problem with copy
Next complains when the server renders something different then the client, therfor we move the differing logic to `useEffect`
* Fix improper component nesting
Next doesn't allow block elements inside a `<p>`
* Replace landing page MDX with page component
* Remove inlined iframe content
* Remove ability to inline HTML content in iFrames
* Remove MDX imports
* Fix problem with image inside link in MDX
* Escape character for MDX
* Fix unescaped characters in MDX
* Fix headings with logo
* Allow to export static HTML pages
* Add prebuild script
This command is automatically run by Next
* Replace `svg-loader` with `react-inlinesvg`
`svg-loader` is no longer maintained
* Fix ESLint `react-hooks/exhaustive-deps`
* Fix dropdowns
* Change code language from `cli` to `bash`
* Remove unnessary language `none`
* Fix invalid code language
`markdown_` with an underscore was used to basically turn of syntax highlighting, but using unknown languages know throws an error.
* Enable code blocks plugin
* Readd `InlineCode` component
MDX2 removed the `inlineCode` component
> The special component name `inlineCode` was removed, we recommend to use `pre` for the block version of code, and code for both the block and inline versions
Source: https://mdxjs.com/migrating/v2/#update-mdx-content
* Remove unused code
* Extract function to own file
* Fix code syntax highlighting
* Update syntax for code block meta data
* Remove unused prop
* Fix internal link recognition
There is a problem with regex between Node and browser, and since Next runs the component on both, this create an error.
`Prop `rel` did not match. Server: "null" Client: "noopener nofollow noreferrer"`
This simplifies the implementation and fixes the above error.
* Replace `react-helmet` with `next/head`
* Fix `className` problem for JSX component
* Fix broken bold markdown
* Convert file to `.mjs` to be used by Node process
* Add plugin to replace strings
* Fix custom table row styling
* Fix problem with `span` inside inline `code`
React doesn't allow a `span` inside an inline `code` element and throws an error in dev mode.
* Add `_document` to be able to customize `<html>` and `<body>`
* Add `lang="en"`
* Store Netlify settings in file
This way we don't need to update via Netlify UI, which can be tricky if changing build settings.
* Add sitemap
* Add Smartypants
* Add PWA support
* Add `manifest.webmanifest`
* Fix bug with anchor links after reloading
There was no need for the previous implementation, since the browser handles this nativly. Additional the manual scrolling into view was actually broken, because the heading would disappear behind the menu bar.
* Rename custom event
I was googeling for ages to find out what kind of event `inview` is, only to figure out it was a custom event with a name that sounds pretty much like a native one. 🫠
* Fix missing comment syntax highlighting
* Refactor Quickstart component
The previous implementation was hidding the irrelevant lines via data-props and dynamically generated CSS. This created problems with Next and was also hard to follow. CSS was used to do what React is supposed to handle.
The new implementation simplfy filters the list of children (React elements) via their props.
* Fix syntax highlighting for Training Quickstart
* Unify code rendering
* Improve error logging in Juniper
* Fix Juniper component
* Automatically generate "Read Next" link
* Add Plausible
* Use recent DocSearch component and adjust styling
* Fix images
* Turn of image optimization
> Image Optimization using Next.js' default loader is not compatible with `next export`.
We currently deploy to Netlify via `next export`
* Dont build pages starting with `_`
* Remove unused files
* Add Next plugin to Netlify
* Fix button layout
MDX automatically adds `p` tags around text on a new line and Prettier wants to put the text on a new line. Hacking with JSX string.
* Add 404 page
* Apply Prettier
* Update Prettier for `package.json`
Next sometimes wants to patch `package-lock.json`. The old Prettier setting indended with 4 spaces, but Next always indends with 2 spaces. Since `npm install` automatically uses the indendation from `package.json` for `package-lock.json` and to avoid the format switching back and forth, both files are now set to 2 spaces.
* Apply Next patch to `package-lock.json`
When starting the dev server Next would warn `warn - Found lockfile missing swc dependencies, patching...` and update the `package-lock.json`. These are the patched changes.
* fix link
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* small backslash fixes
* adjust to new style
Co-authored-by: Marcus Blättermann <marcus@essenmitsosse.de>
