Merge remote-tracking branch 'origin/develop' into rliaw-develop

.github/contributors/hertelm.md

@@ -0,0 +1,106 @@
+# spaCy contributor agreement
+
+This spaCy Contributor Agreement (**"SCA"**) is based on the
+[Oracle Contributor Agreement](http://www.oracle.com/technetwork/oca-405177.pdf).
+The SCA applies to any contribution that you make to any product or project
+managed by us (the **"project"**), and sets out the intellectual property rights
+you grant to us in the contributed materials. The term **"us"** shall mean
+[ExplosionAI GmbH](https://explosion.ai/legal). The term
+**"you"** shall mean the person or entity identified below.
+
+If you agree to be bound by these terms, fill in the information requested
+below and include the filled-in version with your first pull request, under the
+folder [`.github/contributors/`](/.github/contributors/). The name of the file
+should be your GitHub username, with the extension `.md`. For example, the user
+example_user would create the file `.github/contributors/example_user.md`.
+
+Read this agreement carefully before signing. These terms and conditions
+constitute a binding legal agreement.
+
+## Contributor Agreement
+
+1. The term "contribution" or "contributed materials" means any source code,
+object code, patch, tool, sample, graphic, specification, manual,
+documentation, or any other material posted or submitted by you to the project.
+
+2. With respect to any worldwide copyrights, or copyright applications and
+registrations, in your contribution:
+
+    * you hereby assign to us joint ownership, and to the extent that such
+    assignment is or becomes invalid, ineffective or unenforceable, you hereby
+    grant to us a perpetual, irrevocable, non-exclusive, worldwide, no-charge,
+    royalty-free, unrestricted license to exercise all rights under those
+    copyrights. This includes, at our option, the right to sublicense these same
+    rights to third parties through multiple levels of sublicensees or other
+    licensing arrangements;
+
+    * you agree that each of us can do all things in relation to your
+    contribution as if each of us were the sole owners, and if one of us makes
+    a derivative work of your contribution, the one who makes the derivative
+    work (or has it made will be the sole owner of that derivative work;
+
+    * you agree that you will not assert any moral rights in your contribution
+    against us, our licensees or transferees;
+
+    * you agree that we may register a copyright in your contribution and
+    exercise all ownership rights associated with it; and
+
+    * you agree that neither of us has any duty to consult with, obtain the
+    consent of, pay or render an accounting to the other for any use or
+    distribution of your contribution.
+
+3. With respect to any patents you own, or that you can license without payment
+to any third party, you hereby grant to us a perpetual, irrevocable,
+non-exclusive, worldwide, no-charge, royalty-free license to:
+
+    * make, have made, use, sell, offer to sell, import, and otherwise transfer
+    your contribution in whole or in part, alone or in combination with or
+    included in any product, work or materials arising out of the project to
+    which your contribution was submitted, and
+
+    * at our option, to sublicense these same rights to third parties through
+    multiple levels of sublicensees or other licensing arrangements.
+
+4. Except as set out above, you keep all right, title, and interest in your
+contribution. The rights that you grant to us under these terms are effective
+on the date you first submitted a contribution to us, even if your submission
+took place before the date you sign these terms.
+
+5. You covenant, represent, warrant and agree that:
+
+    * Each contribution that you submit is and shall be an original work of
+    authorship and you can legally grant the rights set out in this SCA;
+
+    * to the best of your knowledge, each contribution will not violate any
+    third party's copyrights, trademarks, patents, or other intellectual
+    property rights; and
+
+    * each contribution shall be in compliance with U.S. export control laws and
+    other applicable export and import laws. You agree to notify us if you
+    become aware of any circumstance which would make any of the foregoing
+    representations inaccurate in any respect. We may publicly disclose your
+    participation in the project, including the fact that you have signed the SCA.
+
+6. This SCA is governed by the laws of the State of California and applicable
+U.S. Federal law. Any choice of law rules will not apply.
+
+7. Please place an “x” on one of the applicable statement below. Please do NOT
+mark both statements:
+
+    * [x] I am signing on behalf of myself as an individual and no other person
+    or entity, including my employer, has or will have rights with respect to my
+    contributions.
+
+    * [ ] I am signing on behalf of my employer or a legal entity and I have the
+    actual authority to contractually bind that entity.
+
+## Contributor Details
+
+| Field                          | Entry                |
+|------------------------------- | -------------------- |
+| Name                           | Matthias Hertel      |
+| Company name (if applicable)   |                      |
+| Title or role (if applicable)  |                      |
+| Date                           | June 29, 2020        |
+| GitHub username                | hertelm              |
+| Website (optional)             |                      |

netlify.toml

@@ -1,6 +1,8 @@
 redirects = [
     # Netlify
     {from = "https://spacy.netlify.com/*", to="https://spacy.io/:splat", force = true },
+    # Subdomain for branches
+    {from = "https://nightly.spacy.io/*", to="https://spacy-io-develop.spacy.io/:splat", force = true, status = 200},
     # Old subdomains
     {from = "https://survey.spacy.io/*", to = "https://spacy.io", force = true},
     {from = "http://survey.spacy.io/*", to = "https://spacy.io", force = true},

setup.py

@@ -81,7 +81,7 @@ def is_new_osx():
         return False
     mac_ver = platform.mac_ver()[0]
     if mac_ver.startswith("10"):
-        minor_version = int(mac_ver.split('.')[1])
+        minor_version = int(mac_ver.split(".")[1])
         if minor_version >= 7:
             return True
         else:
@@ -158,7 +158,7 @@ def setup_package():
     ext_modules = cythonize(ext_modules, compiler_directives=COMPILER_DIRECTIVES)
 
     setup(
-        name="spacy",
+        name="spacy-nightly",
         packages=PACKAGES,
         version=about["__version__"],
         ext_modules=ext_modules,

spacy/about.py

@@ -1,6 +1,6 @@
 # fmt: off
-__title__ = "spacy"
+__title__ = "spacy-nightly"
-__version__ = "3.0.0.dev12"
+__version__ = "3.0.0a0"
 __release__ = True
 __download_url__ = "https://github.com/explosion/spacy-models/releases/download"
 __compatibility__ = "https://raw.githubusercontent.com/explosion/spacy-models/master/compatibility.json"

spacy/cli/project.py

@@ -242,12 +242,16 @@ def project_clone(
         try:
             run_command(cmd)
         except SystemExit:
-            err = f"Could not clone the repo '{repo}' into the temp dir '{tmp_dir}'"
+            err = f"Could not clone the repo '{repo}' into the temp dir '{tmp_dir}'."
             msg.fail(err)
         with (tmp_dir / ".git" / "info" / "sparse-checkout").open("w") as f:
             f.write(name)
-        run_command(["git", "-C", str(tmp_dir), "fetch"])
+        try:
-        run_command(["git", "-C", str(tmp_dir), "checkout"])
+            run_command(["git", "-C", str(tmp_dir), "fetch"])
+            run_command(["git", "-C", str(tmp_dir), "checkout"])
+        except SystemExit:
+            err = f"Could not clone '{name}' in the repo '{repo}'."
+            msg.fail(err)
         shutil.move(str(tmp_dir / Path(name).name), str(project_dir))
     msg.good(f"Cloned project '{name}' from {repo} into {project_dir}")
     for sub_dir in DIRS:
@@ -525,9 +529,9 @@ def update_dvc_config(
         outputs_no_cache = command.get("outputs_no_cache", [])
         if not deps and not outputs and not outputs_no_cache:
             continue
-        # Default to "." as the project path since dvc.yaml is auto-generated
+        # Default to the working dir as the project path since dvc.yaml is auto-generated
         # and we don't want arbitrary paths in there
-        project_cmd = ["python", "-m", NAME, "project", ".", "exec", name]
+        project_cmd = ["python", "-m", NAME, "project", "exec", name]
         deps_cmd = [c for cl in [["-d", p] for p in deps] for c in cl]
         outputs_cmd = [c for cl in [["-o", p] for p in outputs] for c in cl]
         outputs_nc_cmd = [c for cl in [["-O", p] for p in outputs_no_cache] for c in cl]

spacy/cli/train.py

@@ -339,6 +339,7 @@ def create_train_batches(nlp, corpus, cfg, randomization_index):
             yield epoch, batch
         if max_epochs >= 1 and epoch >= max_epochs:
             break
+        random.shuffle(train_examples)
 
 
 def create_evaluation_callback(nlp, optimizer, corpus, cfg):
@@ -350,13 +351,14 @@ def create_evaluation_callback(nlp, optimizer, corpus, cfg):
         )
 
         n_words = sum(len(ex.predicted) for ex in dev_examples)
+        batch_size = cfg.get("evaluation_batch_size", 128)
         start_time = timer()
 
         if optimizer.averages:
             with nlp.use_params(optimizer.averages):
-                scorer = nlp.evaluate(dev_examples, batch_size=32)
+                scorer = nlp.evaluate(dev_examples, batch_size=batch_size)
         else:
-            scorer = nlp.evaluate(dev_examples, batch_size=32)
+            scorer = nlp.evaluate(dev_examples, batch_size=batch_size)
         end_time = timer()
         wps = n_words / (end_time - start_time)
         scores = scorer.scores
@@ -479,7 +481,7 @@ def train_while_improving(
         if patience and (step - best_step) >= patience:
             break
         # Stop if we've exhausted our max steps (if specified)
-        if max_steps and (step * accumulate_gradient) >= max_steps:
+        if max_steps and step >= max_steps:
             break
 
 

spacy/gold/corpus.py

@@ -45,18 +45,22 @@ class Corpus:
 
     def make_examples(self, nlp, reference_docs, max_length=0):
         for reference in reference_docs:
-            if len(reference) >= max_length >= 1:
+            if len(reference) == 0:
-                if reference.is_sentenced:
+                continue
-                    for ref_sent in reference.sents:
+            elif max_length == 0 or len(reference) < max_length:
-                        yield Example(
-                            nlp.make_doc(ref_sent.text),
-                            ref_sent.as_doc()
-                        )
-            else:
                 yield Example(
                     nlp.make_doc(reference.text),
                     reference
                 )
+            elif reference.is_sentenced:
+                for ref_sent in reference.sents:
+                    if len(ref_sent) == 0:
+                        continue
+                    elif max_length == 0 or len(ref_sent) < max_length:
+                        yield Example(
+                            nlp.make_doc(ref_sent.text),
+                            ref_sent.as_doc()
+                        )
     
     def make_examples_gold_preproc(self, nlp, reference_docs):
         for reference in reference_docs:
@@ -65,7 +69,7 @@ class Corpus:
             else:
                 ref_sents = [reference]
             for ref_sent in ref_sents:
-                yield Example(
+                eg = Example(
                     Doc(
                         nlp.vocab, 
                         words=[w.text for w in ref_sent],
@@ -73,6 +77,8 @@ class Corpus:
                     ),
                     ref_sent
                 )
+                if len(eg.x):
+                    yield eg
 
     def read_docbin(self, vocab, locs):
         """ Yield training examples as example dicts """

spacy/ml/_precomputable_affine.py

@@ -110,6 +110,7 @@ def init(model, X=None, Y=None):
 
     ops = model.ops
     W = normal_init(ops, W.shape, mean=float(ops.xp.sqrt(1.0 / nF * nI)))
+    pad = normal_init(ops, pad.shape, mean=1.0)
     model.set_param("W", W)
     model.set_param("b", b)
     model.set_param("pad", pad)

spacy/syntax/_parser_model.pyx

@@ -339,6 +339,7 @@ cdef class precompute_hiddens:
     cdef readonly int nF, nO, nP
     cdef bint _is_synchronized
     cdef public object ops
+    cdef public object numpy_ops
     cdef np.ndarray _features
     cdef np.ndarray _cached
     cdef np.ndarray bias
@@ -368,6 +369,7 @@ cdef class precompute_hiddens:
             self.nP = 1
         self.nO = cached.shape[2]
         self.ops = lower_model.ops
+        self.numpy_ops = NumpyOps()
         assert activation in (None, "relu", "maxout")
         self.activation = activation
         self._is_synchronized = False
@@ -446,44 +448,32 @@ cdef class precompute_hiddens:
         return state_vector, backward
 
     def _nonlinearity(self, state_vector):
-        if isinstance(state_vector, numpy.ndarray):
-            ops = NumpyOps()
-        else:
-            ops = CupyOps()
-
         if self.activation == "maxout":
-            state_vector, mask = ops.maxout(state_vector)
+            return self._maxout_nonlinearity(state_vector)
         else:
-            state_vector = state_vector.reshape(state_vector.shape[:-1])
+            return self._relu_nonlinearity(state_vector)
-            if self.activation == "relu":
-                mask = state_vector >= 0.
-                state_vector *= mask
-            else:
-                mask = None
 
-        def backprop_nonlinearity(d_best):
+    def _maxout_nonlinearity(self, state_vector):
-            if isinstance(d_best, numpy.ndarray):
+        state_vector, mask = self.numpy_ops.maxout(state_vector)
-                ops = NumpyOps()
+        # We're outputting to CPU, but we need this variable on GPU for the
-            else:
+        # backward pass.
-                ops = CupyOps()
+        mask = self.ops.asarray(mask)
-            if mask is not None:
+
-                mask_ = ops.asarray(mask)
+        def backprop_maxout(d_best):
-            # This will usually be on GPU
+            return self.ops.backprop_maxout(d_best, mask, self.nP)
-            d_best = ops.asarray(d_best)
+        
-            # Fix nans (which can occur from unseen classes.)
+        return state_vector, backprop_maxout
-            try:
+
-                d_best[ops.xp.isnan(d_best)] = 0.
+    def _relu_nonlinearity(self, state_vector):
-            except:
+        state_vector = state_vector.reshape((state_vector.shape[0], -1))
-                print(ops.xp.isnan(d_best))
+        mask = state_vector >= 0.
-                raise
+        state_vector *= mask
-            if self.activation == "maxout":
+        # We're outputting to CPU, but we need this variable on GPU for the
-                mask_ = ops.asarray(mask)
+        # backward pass.
-                return ops.backprop_maxout(d_best, mask_, self.nP)
+        mask = self.ops.asarray(mask)
-            elif self.activation == "relu":
+
-                mask_ = ops.asarray(mask)
+        def backprop_relu(d_best):
-                d_best *= mask_
+            d_best *= mask
-                d_best = d_best.reshape((d_best.shape + (1,)))
+            return d_best.reshape((d_best.shape + (1,)))
-                return d_best
+ 
-            else:
+        return state_vector, backprop_relu
-                return d_best.reshape((d_best.shape + (1,)))
-        return state_vector, backprop_nonlinearity

spacy/syntax/arc_eager.pyx

@@ -742,21 +742,14 @@ cdef class ArcEager(TransitionSystem):
         if n_gold < 1:
             raise ValueError
 
-    def get_oracle_sequence(self, Example example):
+    def get_oracle_sequence_from_state(self, StateClass state, ArcEagerGold gold, _debug=None):
-        cdef StateClass state
+        cdef int i
-        cdef ArcEagerGold gold
-        states, golds, n_steps = self.init_gold_batch([example])
-        if not golds:
-            return []
-
         cdef Pool mem = Pool()
         # n_moves should not be zero at this point, but make sure to avoid zero-length mem alloc
         assert self.n_moves > 0
         costs = <float*>mem.alloc(self.n_moves, sizeof(float))
         is_valid = <int*>mem.alloc(self.n_moves, sizeof(int))
 
-        state = states[0]
-        gold = golds[0]
         history = []
         debug_log = []
         failed = False
@@ -772,18 +765,21 @@ cdef class ArcEager(TransitionSystem):
                     history.append(i)
                     s0 = state.S(0)
                     b0 = state.B(0)
-                    debug_log.append(" ".join((
+                    if _debug:
-                        self.get_class_name(i),
+                        example = _debug
-                        "S0=", (example.x[s0].text if s0 >= 0 else "__"),
+                        debug_log.append(" ".join((
-                        "B0=", (example.x[b0].text if b0 >= 0 else "__"),
+                            self.get_class_name(i),
-                        "S0 head?", str(state.has_head(state.S(0))),
+                            "S0=", (example.x[s0].text if s0 >= 0 else "__"),
-                    )))
+                            "B0=", (example.x[b0].text if b0 >= 0 else "__"),
+                            "S0 head?", str(state.has_head(state.S(0))),
+                        )))
                     action.do(state.c, action.label)
                     break
             else:
                 failed = False
                 break
         if failed:
+            example = _debug
             print("Actions")
             for i in range(self.n_moves):
                 print(self.get_class_name(i))

spacy/syntax/nn_parser.pyx

@@ -63,7 +63,9 @@ cdef class Parser:
         self.model = model
         if self.moves.n_moves != 0:
             self.set_output(self.moves.n_moves)
-        self.cfg = cfg
+        self.cfg = dict(cfg)
+        self.cfg.setdefault("update_with_oracle_cut_size", 100)
+        self.cfg.setdefault("normalize_gradients_with_batch_size", True)
         self._multitasks = []
         for multitask in cfg.get("multitasks", []):
             self.add_multitask_objective(multitask)
@@ -263,22 +265,32 @@ cdef class Parser:
         free(is_valid)
 
     def update(self, examples, drop=0., set_annotations=False, sgd=None, losses=None):
+        cdef StateClass state
         if losses is None:
             losses = {}
         losses.setdefault(self.name, 0.)
         for multitask in self._multitasks:
             multitask.update(examples, drop=drop, sgd=sgd)
+        n_examples = len([eg for eg in examples if self.moves.has_gold(eg)])
+        if n_examples == 0:
+            return losses
         set_dropout_rate(self.model, drop)
         # Prepare the stepwise model, and get the callback for finishing the batch
         model, backprop_tok2vec = self.model.begin_update(
             [eg.predicted for eg in examples])
-        # Chop sequences into lengths of this many transitions, to make the
+        if self.cfg["update_with_oracle_cut_size"] >= 1:
-        # batch uniform length. We randomize this to overfit less.
+            # Chop sequences into lengths of this many transitions, to make the
-        cut_gold = numpy.random.choice(range(20, 100))
+            # batch uniform length. We randomize this to overfit less.
-        states, golds, max_steps = self._init_gold_batch(
+            cut_size = self.cfg["update_with_oracle_cut_size"]
-            examples,
+            states, golds, max_steps = self._init_gold_batch(
-            max_length=cut_gold
+                examples,
-        )
+                max_length=numpy.random.choice(range(5, cut_size))
+            )
+        else:
+            states, golds, _ = self.moves.init_gold_batch(examples)
+            max_steps = max([len(eg.x) for eg in examples])
+        if not states:
+            return losses
         all_states = list(states)
         states_golds = zip(states, golds)
         for _ in range(max_steps):
@@ -287,6 +299,17 @@ cdef class Parser:
             states, golds = zip(*states_golds)
             scores, backprop = model.begin_update(states)
             d_scores = self.get_batch_loss(states, golds, scores, losses)
+            if self.cfg["normalize_gradients_with_batch_size"]:
+                # We have to be very careful how we do this, because of the way we
+                # cut up the batch. We subdivide long sequences. If we normalize
+                # naively, we end up normalizing by sequence length, which
+                # is bad: that would mean that states in long sequences
+                # consistently get smaller gradients. Imagine if we have two
+                # sequences, one length 1000, one length 20. If we cut up
+                # the 1k sequence so that we have a "batch" of 50 subsequences,
+                # we don't want the gradients to get 50 times smaller!
+                d_scores /= n_examples
+
             backprop(d_scores)
             # Follow the predicted action
             self.transition_states(states, scores)
@@ -384,8 +407,6 @@ cdef class Parser:
             cpu_log_loss(c_d_scores,
                 costs, is_valid, &scores[i, 0], d_scores.shape[1])
             c_d_scores += d_scores.shape[1]
-        if len(states):
-            d_scores /= len(states)
         if losses is not None:
             losses.setdefault(self.name, 0.)
             losses[self.name] += (d_scores**2).sum()
@@ -428,7 +449,7 @@ cdef class Parser:
                 if component is self:
                     break
                 if hasattr(component, "pipe"):
-                    doc_sample = list(component.pipe(doc_sample))
+                    doc_sample = list(component.pipe(doc_sample, batch_size=8))
                 else:
                     doc_sample = [component(doc) for doc in doc_sample]
         if doc_sample:
@@ -498,40 +519,49 @@ cdef class Parser:
         return self
 
     def _init_gold_batch(self, examples, min_length=5, max_length=500):
-        """Make a square batch, of length equal to the shortest doc. A long
+        """Make a square batch, of length equal to the shortest transition
+        sequence or a cap. A long
         doc will get multiple states. Let's say we have a doc of length 2*N,
         where N is the shortest doc. We'll make two states, one representing
         long_doc[:N], and another representing long_doc[N:]."""
         cdef:
+            StateClass start_state
             StateClass state
             Transition action
         all_states = self.moves.init_batch([eg.predicted for eg in examples])
         kept = []
+        max_length_seen = 0
         for state, eg in zip(all_states, examples):
             if self.moves.has_gold(eg) and not state.is_final():
                 gold = self.moves.init_gold(state, eg)
-                kept.append((eg, state, gold))
+                oracle_actions = self.moves.get_oracle_sequence_from_state(
-        max_length = max(min_length, min(max_length, min([len(eg.x) for eg in examples])))
+                    state.copy(), gold)
-        max_moves = 0
+                kept.append((eg, state, gold, oracle_actions))
+                min_length = min(min_length, len(oracle_actions))
+                max_length_seen = max(max_length, len(oracle_actions))
+        if not kept:
+            return [], [], 0
+        max_length = max(min_length, min(max_length, max_length_seen))
         states = []
         golds = []
-        for eg, state, gold in kept:
+        cdef int clas
-            oracle_actions = self.moves.get_oracle_sequence(eg)
+        max_moves = 0
-            start = 0
+        for eg, state, gold, oracle_actions in kept:
-            while start < len(eg.predicted):
+            for i in range(0, len(oracle_actions), max_length):
-                state = state.copy()
+                start_state = state.copy()
                 n_moves = 0
-                while state.B(0) < start and not state.is_final():
+                for clas in oracle_actions[i:i+max_length]:
-                    action = self.moves.c[oracle_actions.pop(0)]
+                    action = self.moves.c[clas]
                     action.do(state.c, action.label)
                     state.c.push_hist(action.clas)
                     n_moves += 1
-                has_gold = self.moves.has_gold(eg, start=start,
+                    if state.is_final():
-                                               end=start+max_length)
+                        break
-                if not state.is_final() and has_gold:
+                max_moves = max(max_moves, n_moves)
-                    states.append(state)
+                if self.moves.has_gold(eg, start_state.B(0), state.B(0)):
+                    states.append(start_state)
                     golds.append(gold)
                     max_moves = max(max_moves, n_moves)
-                start += min(max_length, len(eg.x)-start)
+                if state.is_final():
-            max_moves = max(max_moves, len(oracle_actions))
+                    break
         return states, golds, max_moves

spacy/syntax/transition_system.pyx

@@ -60,20 +60,25 @@ cdef class TransitionSystem:
             states.append(state)
             offset += len(doc)
         return states
-
+    
     def get_oracle_sequence(self, Example example, _debug=False):
+        states, golds, _ = self.init_gold_batch([example])
+        if not states:
+            return []
+        state = states[0]
+        gold = golds[0]
+        if _debug:
+            return self.get_oracle_sequence_from_state(state, gold, _debug=example)
+        else:
+            return self.get_oracle_sequence_from_state(state, gold)
+
+    def get_oracle_sequence_from_state(self, StateClass state, gold, _debug=None):
         cdef Pool mem = Pool()
         # n_moves should not be zero at this point, but make sure to avoid zero-length mem alloc
         assert self.n_moves > 0
         costs = <float*>mem.alloc(self.n_moves, sizeof(float))
         is_valid = <int*>mem.alloc(self.n_moves, sizeof(int))
 
-        cdef StateClass state
-        states, golds, n_steps = self.init_gold_batch([example])
-        if not states:
-            return []
-        state = states[0]
-        gold = golds[0]
         history = []
         debug_log = []
         while not state.is_final():
@@ -82,9 +87,10 @@ cdef class TransitionSystem:
                 if is_valid[i] and costs[i] <= 0:
                     action = self.c[i]
                     history.append(i)
-                    s0 = state.S(0)
-                    b0 = state.B(0)
                     if _debug:
+                        s0 = state.S(0)
+                        b0 = state.B(0)
+                        example = _debug
                         debug_log.append(" ".join((
                             self.get_class_name(i),
                             "S0=", (example.x[s0].text if s0 >= 0 else "__"),
@@ -95,6 +101,7 @@ cdef class TransitionSystem:
                     break
             else:
                 if _debug:
+                    example = _debug
                     print("Actions")
                     for i in range(self.n_moves):
                         print(self.get_class_name(i))

website/docs/api/phrasematcher.md

@@ -91,7 +91,7 @@ Match a stream of documents, yielding them in turn.
 > ```python
 >   from spacy.matcher import PhraseMatcher
 >   matcher = PhraseMatcher(nlp.vocab)
->   for doc in matcher.pipe(texts, batch_size=50):
+>   for doc in matcher.pipe(docs, batch_size=50):
 >       pass
 > ```
 

website/docs/api/scorer.md

@@ -46,19 +46,19 @@ Update the evaluation scores from a single [`Doc`](/api/doc) /
 
 ## Properties
 
-| Name                                                | Type  | Description                                                                                                |
+| Name                                                | Type  | Description                                                                            |
-| --------------------------------------------------- | ----- | ---------------------------------------------------------------------------------------------------------- |
+| --------------------------------------------------- | ----- | -------------------------------------------------------------------------------------- |
-| `token_acc`                                         | float | Tokenization accuracy.                                                                                     |
+| `token_acc`                                         | float | Tokenization accuracy.                                                                 |
-| `tags_acc`                                          | float | Part-of-speech tag accuracy (fine grained tags, i.e. `Token.tag`).                                         |
+| `tags_acc`                                          | float | Part-of-speech tag accuracy (fine grained tags, i.e. `Token.tag`).                     |
-| `uas`                                               | float | Unlabelled dependency score.                                                                               |
+| `uas`                                               | float | Unlabelled dependency score.                                                           |
-| `las`                                               | float | Labelled dependency score.                                                                                 |
+| `las`                                               | float | Labelled dependency score.                                                             |
-| `ents_p`                                            | float | Named entity accuracy (precision).                                                                         |
+| `ents_p`                                            | float | Named entity accuracy (precision).                                                     |
-| `ents_r`                                            | float | Named entity accuracy (recall).                                                                            |
+| `ents_r`                                            | float | Named entity accuracy (recall).                                                        |
-| `ents_f`                                            | float | Named entity accuracy (F-score).                                                                           |
+| `ents_f`                                            | float | Named entity accuracy (F-score).                                                       |
-| `ents_per_type` <Tag variant="new">2.1.5</Tag>      | dict  | Scores per entity label. Keyed by label, mapped to a dict of `p`, `r` and `f` scores.                      |
+| `ents_per_type` <Tag variant="new">2.1.5</Tag>      | dict  | Scores per entity label. Keyed by label, mapped to a dict of `p`, `r` and `f` scores.  |
 | `textcat_f` <Tag variant="new">3.0</Tag>            | float | F-score on positive label for binary classification, macro-averaged F-score otherwise. |
-| `textcat_auc` <Tag variant="new"3.0</Tag>           | float | Macro-averaged AUC ROC score for multilabel classification (`-1` if undefined).                            |
+| `textcat_auc` <Tag variant="new">3.0</Tag>          | float | Macro-averaged AUC ROC score for multilabel classification (`-1` if undefined).        |
-| `textcats_f_per_cat` <Tag variant="new">3.0</Tag>   | dict  | F-scores per textcat label, keyed by label.                                                                |
+| `textcats_f_per_cat` <Tag variant="new">3.0</Tag>   | dict  | F-scores per textcat label, keyed by label.                                            |
-| `textcats_auc_per_cat` <Tag variant="new">3.0</Tag> | dict  | ROC AUC scores per textcat label, keyed by label.                                                          |
+| `textcats_auc_per_cat` <Tag variant="new">3.0</Tag> | dict  | ROC AUC scores per textcat label, keyed by label.                                      |
-| `las_per_type` <Tag variant="new">2.2.3</Tag>       | dict  | Labelled dependency scores, keyed by label.                                                                |
+| `las_per_type` <Tag variant="new">2.2.3</Tag>       | dict  | Labelled dependency scores, keyed by label.                                            |
-| `scores`                                            | dict  | All scores, keyed by type.                                                                                 |
+| `scores`                                            | dict  | All scores, keyed by type.                                                             |

website/docs/usage/rule-based-matching.md

@@ -122,7 +122,7 @@ for match_id, start, end in matches:
 ```
 
 The matcher returns a list of `(match_id, start, end)` tuples – in this case,
-`[('15578876784678163569', 0, 2)]`, which maps to the span `doc[0:2]` of our
+`[('15578876784678163569', 0, 3)]`, which maps to the span `doc[0:3]` of our
 original document. The `match_id` is the [hash value](/usage/spacy-101#vocab) of
 the string ID "HelloWorld". To get the string value, you can look up the ID in
 the [`StringStore`](/api/stringstore).

website/docs/usage/v2-3.md

@@ -161,10 +161,18 @@ debugging your tokenizer configuration.
 
 spaCy's custom warnings have been replaced with native Python
 [`warnings`](https://docs.python.org/3/library/warnings.html). Instead of
-setting `SPACY_WARNING_IGNORE`, use the
+setting `SPACY_WARNING_IGNORE`, use the [`warnings`
-[`warnings` filters](https://docs.python.org/3/library/warnings.html#the-warnings-filter)
+filters](https://docs.python.org/3/library/warnings.html#the-warnings-filter)
 to manage warnings.
 
+```diff
+import spacy
++ import warnings
+
+- spacy.errors.SPACY_WARNING_IGNORE.append('W007')
++ warnings.filterwarnings("ignore", message=r"\\[W007\\]", category=UserWarning)
+```
+
 #### Normalization tables
 
 The normalization tables have moved from the language data in
@@ -174,6 +182,65 @@ If you're adding data for a new language, the normalization table should be
 added to `spacy-lookups-data`. See
 [adding norm exceptions](/usage/adding-languages#norm-exceptions).
 
+#### No preloaded vocab for models with vectors
+
+To reduce the initial loading time, the lexemes in `nlp.vocab` are no longer
+loaded on initialization for models with vectors. As you process texts, the
+lexemes will be added to the vocab automatically, just as in small models
+without vectors.
+
+To see the number of unique vectors and number of words with vectors, see
+`nlp.meta['vectors']`, for example for `en_core_web_md` there are `20000`
+unique vectors and `684830` words with vectors:
+
+```python
+{
+    'width': 300,
+    'vectors': 20000,
+    'keys': 684830,
+    'name': 'en_core_web_md.vectors'
+}
+```
+
+If required, for instance if you are working directly with word vectors rather
+than processing texts, you can load all lexemes for words with vectors at once:
+
+```python
+for orth in nlp.vocab.vectors:
+    _ = nlp.vocab[orth]
+```
+
+If your workflow previously iterated over `nlp.vocab`, a similar alternative
+is to iterate over words with vectors instead:
+
+```diff
+- lexemes = [w for w in nlp.vocab]
++ lexemes = [nlp.vocab[orth] for orth in nlp.vocab.vectors]
+```
+
+Be aware that the set of preloaded lexemes in a v2.2 model is not equivalent to
+the set of words with vectors. For English, v2.2 `md/lg` models have 1.3M
+provided lexemes but only 685K words with vectors. The vectors have been
+updated for most languages in v2.2, but the English models contain the same
+vectors for both v2.2 and v2.3.
+
+#### Lexeme.is_oov and Token.is_oov
+
+<Infobox title="Important note" variant="warning">
+
+Due to a bug, the values for `is_oov` are reversed in v2.3.0, but this will be
+fixed in the next patch release v2.3.1.
+
+</Infobox>
+
+In v2.3, `Lexeme.is_oov` and `Token.is_oov` are `True` if the lexeme does not
+have a word vector. This is equivalent to `token.orth not in
+nlp.vocab.vectors`.
+
+Previously in v2.2, `is_oov` corresponded to whether a lexeme had stored
+probability and cluster features. The probability and cluster features are no
+longer included in the provided medium and large models (see the next section).
+
 #### Probability and cluster features
 
 > #### Load and save extra prob lookups table
@@ -201,6 +268,28 @@ model vocab, which will take a few seconds on initial loading. When you save
 this model after loading the `prob` table, the full `prob` table will be saved
 as part of the model vocab.
 
+To load the probability table into a provided model, first make sure you have
+`spacy-lookups-data` installed. To load the table, remove the empty provided
+`lexeme_prob` table and then access `Lexeme.prob` for any word to load the
+table from `spacy-lookups-data`:
+
+```diff
++ # prerequisite: pip install spacy-lookups-data
+import spacy
+
+nlp = spacy.load("en_core_web_md")
+
+# remove the empty placeholder prob table
++ if nlp.vocab.lookups_extra.has_table("lexeme_prob"):
++     nlp.vocab.lookups_extra.remove_table("lexeme_prob")
+
+# access any `.prob` to load the full table into the model
+assert nlp.vocab["a"].prob == -3.9297883511
+
+# if desired, save this model with the probability table included
+nlp.to_disk("/path/to/model")
+```
+
 If you'd like to include custom `cluster`, `prob`, or `sentiment` tables as part
 of a new model, add the data to
 [`spacy-lookups-data`](https://github.com/explosion/spacy-lookups-data) under
@@ -218,3 +307,39 @@ When you initialize a new model with [`spacy init-model`](/api/cli#init-model),
 the `prob` table from `spacy-lookups-data` may be loaded as part of the
 initialization. If you'd like to omit this extra data as in spaCy's provided
 v2.3 models, use the new flag `--omit-extra-lookups`.
+
+#### Tag maps in provided models vs. blank models
+
+The tag maps in the provided models may differ from the tag maps in the spaCy
+library. You can access the tag map in a loaded model under
+`nlp.vocab.morphology.tag_map`.
+
+The tag map from `spacy.lang.lg.tag_map` is still used when a blank model is
+initialized. If you want to provide an alternate tag map, update
+`nlp.vocab.morphology.tag_map` after initializing the model or if you're using
+the [train CLI](/api/cli#train), you can use the new `--tag-map-path` option to
+provide in the tag map as a JSON dict.
+
+If you want to export a tag map from a provided model for use with the train
+CLI, you can save it as a JSON dict. To only use string keys as required by
+JSON and to make it easier to read and edit, any internal integer IDs need to
+be converted back to strings:
+
+```python
+import spacy
+import srsly
+
+nlp = spacy.load("en_core_web_sm")
+tag_map = {}
+
+# convert any integer IDs to strings for JSON
+for tag, morph in nlp.vocab.morphology.tag_map.items():
+    tag_map[tag] = {}
+    for feat, val in morph.items():
+        feat = nlp.vocab.strings.as_string(feat)
+        if not isinstance(val, bool):
+            val = nlp.vocab.strings.as_string(val)
+        tag_map[tag][feat] = val
+
+srsly.write_json("tag_map.json", tag_map)
+```

website/docs/usage/v3.md

@@ -0,0 +1,17 @@
+---
+title: What's New in v3.0
+teaser: New features, backwards incompatibilities and migration guide
+menu:
+  - ['Summary', 'summary']
+  - ['New Features', 'features']
+  - ['Backwards Incompatibilities', 'incompat']
+  - ['Migrating from v2.x', 'migrating']
+---
+
+## Summary {#summary}
+
+## New Features {#features}
+
+## Backwards Incompatibilities {#incompat}
+
+## Migrating from v2.x {#migrating}

website/gatsby-config.js

@@ -15,6 +15,11 @@ const universe = require('./meta/universe.json')
 
 const DEFAULT_TEMPLATE = path.resolve('./src/templates/index.js')
 
+const isNightly = !!+process.env.SPACY_NIGHTLY || site.nightlyBranches.includes(process.env.BRANCH)
+const favicon = isNightly ? `src/images/icon_nightly.png` : `src/images/icon.png`
+const binderBranch = isNightly ? 'nightly' : site.binderBranch
+const siteUrl = isNightly ? site.siteUrlNightly : site.siteUrl
+
 module.exports = {
     siteMetadata: {
         ...site,
@@ -22,6 +27,9 @@ module.exports = {
         sidebars,
         ...models,
         universe,
+        nightly: isNightly,
+        binderBranch,
+        siteUrl,
     },
 
     plugins: [
@@ -128,7 +136,7 @@ module.exports = {
                 background_color: site.theme,
                 theme_color: site.theme,
                 display: `minimal-ui`,
-                icon: `src/images/icon.png`,
+                icon: favicon,
             },
         },
         {
@@ -140,6 +148,23 @@ module.exports = {
                 respectDNT: true,
             },
         },
+        {
+            resolve: 'gatsby-plugin-robots-txt',
+            options: {
+                host: siteUrl,
+                sitemap: `${siteUrl}/sitemap.xml`,
+                // If we're in a special state (nightly, legacy) prevent indexing
+                resolveEnv: () => (isNightly ? 'development' : 'production'),
+                env: {
+                    production: {
+                        policy: [{ userAgent: '*', allow: '/' }],
+                    },
+                    development: {
+                        policy: [{ userAgent: '*', disallow: ['/'] }],
+                    },
+                },
+            },
+        },
         `gatsby-plugin-offline`,
     ],
 }

website/meta/languages.json

@@ -78,11 +78,14 @@
             "name": "Japanese",
             "models": ["ja_core_news_sm", "ja_core_news_md", "ja_core_news_lg"],
             "dependencies": [
+                { "name": "Unidic", "url": "http://unidic.ninjal.ac.jp/back_number#unidic_cwj" },
+                { "name": "Mecab", "url": "https://github.com/taku910/mecab" },
                 {
                     "name": "SudachiPy",
                     "url": "https://github.com/WorksApplications/SudachiPy"
                 }
             ],
+            "example": "これは文章です。",
             "has_examples": true
         },
         {
@@ -191,17 +194,6 @@
             "example": "นี่คือประโยค",
             "has_examples": true
         },
-        {
-            "code": "ja",
-            "name": "Japanese",
-            "dependencies": [
-                { "name": "Unidic", "url": "http://unidic.ninjal.ac.jp/back_number#unidic_cwj" },
-                { "name": "Mecab", "url": "https://github.com/taku910/mecab" },
-                { "name": "fugashi", "url": "https://github.com/polm/fugashi" }
-            ],
-            "example": "これは文章です。",
-            "has_examples": true
-        },
         {
             "code": "ko",
             "name": "Korean",

website/meta/sidebars.json

@@ -8,11 +8,7 @@
                     { "text": "Installation", "url": "/usage" },
                     { "text": "Models & Languages", "url": "/usage/models" },
                     { "text": "Facts & Figures", "url": "/usage/facts-figures" },
-                    { "text": "spaCy 101", "url": "/usage/spacy-101" },
+                    { "text": "New in v3.0", "url": "/usage/v3" }
-                    { "text": "New in v2.3", "url": "/usage/v2-3" },
-                    { "text": "New in v2.2", "url": "/usage/v2-2" },
-                    { "text": "New in v2.1", "url": "/usage/v2-1" },
-                    { "text": "New in v2.0", "url": "/usage/v2" }
                 ]
             },
             {

website/meta/site.json

@@ -3,6 +3,8 @@
     "description": "spaCy is a free open-source library for Natural Language Processing in Python. It features NER, POS tagging, dependency parsing, word vectors and more.",
     "slogan": "Industrial-strength Natural Language Processing in Python",
     "siteUrl": "https://spacy.io",
+    "siteUrlNightly": "https://nightly.spacy.io",
+    "nightlyBranches": ["spacy.io-develop"],
     "email": "contact@explosion.ai",
     "company": "Explosion AI",
     "companyUrl": "https://explosion.ai",

website/package.json

@@ -16,7 +16,7 @@
         "autoprefixer": "^9.4.7",
         "classnames": "^2.2.6",
         "codemirror": "^5.43.0",
-        "gatsby": "^2.1.18",
+        "gatsby": "^2.11.1",
         "gatsby-image": "^2.0.29",
         "gatsby-mdx": "^0.3.6",
         "gatsby-plugin-catch-links": "^2.0.11",
@@ -25,6 +25,7 @@
         "gatsby-plugin-offline": "^2.0.24",
         "gatsby-plugin-react-helmet": "^3.0.6",
         "gatsby-plugin-react-svg": "^2.0.0",
+        "gatsby-plugin-robots-txt": "^1.5.1",
         "gatsby-plugin-sass": "^2.0.10",
         "gatsby-plugin-sharp": "^2.0.20",
         "gatsby-plugin-sitemap": "^2.0.5",
@@ -52,6 +53,7 @@
     "scripts": {
         "build": "gatsby build",
         "dev": "gatsby develop",
+        "dev:nightly": "BRANCH=spacy.io-develop npm run dev",
         "lint": "eslint **",
         "clear": "rm -rf .cache",
         "test": "echo \"Write tests! -> https://gatsby.app/unit-testing\""

website/src/components/button.js

@@ -27,7 +27,7 @@ Button.defaultProps = {
 }
 
 Button.propTypes = {
-    to: PropTypes.string.isRequired,
+    to: PropTypes.string,
     variant: PropTypes.oneOf(['primary', 'secondary', 'tertiary']),
     large: PropTypes.bool,
     icon: PropTypes.string,

website/src/components/icon.js

@@ -19,6 +19,7 @@ import { ReactComponent as NoIcon } from '../images/icons/no.svg'
 import { ReactComponent as NeutralIcon } from '../images/icons/neutral.svg'
 import { ReactComponent as OfflineIcon } from '../images/icons/offline.svg'
 import { ReactComponent as SearchIcon } from '../images/icons/search.svg'
+import { ReactComponent as MoonIcon } from '../images/icons/moon.svg'
 
 import classes from '../styles/icon.module.sass'
 
@@ -41,6 +42,7 @@ const icons = {
     neutral: NeutralIcon,
     offline: OfflineIcon,
     search: SearchIcon,
+    moon: MoonIcon,
 }
 
 const Icon = ({ name, width, height, inline, variant, className }) => {

website/src/components/landing.js

@@ -2,7 +2,9 @@ import React, { Fragment } from 'react'
 import classNames from 'classnames'
 
 import pattern from '../images/pattern_blue.jpg'
+import patternNightly from '../images/pattern_nightly.jpg'
 import patternOverlay from '../images/pattern_landing.jpg'
+import patternOverlayNightly from '../images/pattern_landing_nightly.jpg'
 import logoSvgs from '../images/logos'
 
 import Grid from './grid'
@@ -14,9 +16,10 @@ import Link from './link'
 import { chunkArray } from './util'
 import classes from '../styles/landing.module.sass'
 
-export const LandingHeader = ({ style = {}, children }) => {
+export const LandingHeader = ({ nightly, style = {}, children }) => {
-    const wrapperStyle = { backgroundImage: `url(${pattern})` }
+    const overlay = nightly ? patternOverlayNightly : patternOverlay
-    const contentStyle = { backgroundImage: `url(${patternOverlay})`, ...style }
+    const wrapperStyle = { backgroundImage: `url(${nightly ? patternNightly : pattern})` }
+    const contentStyle = { backgroundImage: `url(${overlay})`, ...style }
     return (
         <header className={classes.header}>
             <div className={classes.headerWrapper} style={wrapperStyle}>

website/src/components/main.js

@@ -5,15 +5,22 @@ import classNames from 'classnames'
 import patternBlue from '../images/pattern_blue.jpg'
 import patternGreen from '../images/pattern_green.jpg'
 import patternPurple from '../images/pattern_purple.jpg'
+import patternNightly from '../images/pattern_nightly.jpg'
 import classes from '../styles/main.module.sass'
 
-const patterns = { blue: patternBlue, green: patternGreen, purple: patternPurple }
+const patterns = {
+    blue: patternBlue,
+    green: patternGreen,
+    purple: patternPurple,
+    nightly: patternNightly,
+}
 
 export const Content = ({ Component = 'div', className, children }) => (
     <Component className={classNames(classes.content, className)}>{children}</Component>
 )
 
 const Main = ({ sidebar, asides, wrapContent, theme, footer, children }) => {
+    const pattern = patterns[theme]
     const mainClassNames = classNames(classes.root, {
         [classes.withSidebar]: sidebar,
         [classes.withAsides]: asides,
@@ -23,10 +30,7 @@ const Main = ({ sidebar, asides, wrapContent, theme, footer, children }) => {
         <main className={mainClassNames}>
             {wrapContent ? <Content Component="article">{children}</Content> : children}
             {asides && (
-                <div
+                <div className={classes.asides} style={{ backgroundImage: `url(${pattern}` }} />
-                    className={classes.asides}
-                    style={{ backgroundImage: `url(${patterns[theme]}` }}
-                />
             )}
             {footer}
         </main>

website/src/components/seo.js

@@ -6,6 +6,7 @@ import { StaticQuery, graphql } from 'gatsby'
 import socialImageDefault from '../images/social_default.jpg'
 import socialImageApi from '../images/social_api.jpg'
 import socialImageUniverse from '../images/social_universe.jpg'
+import socialImageNightly from '../images/social_nightly.jpg'
 
 function getPageTitle(title, sitename, slogan, sectionTitle) {
     if (sectionTitle && title) {
@@ -17,13 +18,14 @@ function getPageTitle(title, sitename, slogan, sectionTitle) {
     return `${sitename} · ${slogan}`
 }
 
-function getImage(section) {
+function getImage(section, nightly) {
+    if (nightly) return socialImageNightly
     if (section === 'api') return socialImageApi
     if (section === 'universe') return socialImageUniverse
     return socialImageDefault
 }
 
-const SEO = ({ description, lang, title, section, sectionTitle, bodyClass }) => (
+const SEO = ({ description, lang, title, section, sectionTitle, bodyClass, nightly }) => (
     <StaticQuery
         query={query}
         render={data => {
@@ -35,7 +37,7 @@ const SEO = ({ description, lang, title, section, sectionTitle, bodyClass }) =>
                 siteMetadata.slogan,
                 sectionTitle
             )
-            const socialImage = siteMetadata.siteUrl + getImage(section)
+            const socialImage = siteMetadata.siteUrl + getImage(section, nightly)
             const meta = [
                 {
                     name: 'description',

website/src/components/tag.js

@@ -11,6 +11,9 @@ const Tag = ({ spaced, variant, tooltip, children }) => {
         const isValid = isString(children) && !isNaN(children)
         const version = isValid ? Number(children).toFixed(1) : children
         const tooltipText = `This feature is new and was introduced in spaCy v${version}`
+        // TODO: we probably want to handle this more elegantly, but the idea is
+        // that we can hide tags referring to old versions
+        // const hideTag = version.startsWith('2')
         return (
             <TagTemplate spaced={spaced} tooltip={tooltipText}>
                 v{version}

website/src/images/icons/moon.svg

@@ -0,0 +1,3 @@
+<svg xmlns="http://www.w3.org/2000/svg" width="32" height="32" viewBox="0 0 32 32">
+<path d="M10.895 7.574c0 7.55 5.179 13.67 11.567 13.67 1.588 0 3.101-0.38 4.479-1.063-1.695 4.46-5.996 7.636-11.051 7.636-6.533 0-11.83-5.297-11.83-11.83 0-4.82 2.888-8.959 7.023-10.803-0.116 0.778-0.188 1.573-0.188 2.39z"></path>
+</svg>

website/src/pages/404.js

@@ -0,0 +1,47 @@
+import React from 'react'
+import { window } from 'browser-monads'
+import { graphql } from 'gatsby'
+
+import Template from '../templates/index'
+import { LandingHeader, LandingTitle } from '../components/landing'
+import Button from '../components/button'
+
+export default ({ data, location }) => {
+    const { nightly } = data.site.siteMetadata
+    const pageContext = { title: '404 Error', searchExclude: true, isIndex: false }
+    return (
+        <Template data={data} pageContext={pageContext} location={location}>
+            <LandingHeader style={{ minHeight: 400 }} nightly={nightly}>
+                <LandingTitle>
+                    Ooops, this page
+                    <br />
+                    does not exist!
+                </LandingTitle>
+                <br />
+                <Button onClick={() => window.history.go(-1)} variant="tertiary">
+                    Click here to go back
+                </Button>
+            </LandingHeader>
+        </Template>
+    )
+}
+
+export const pageQuery = graphql`
+    query {
+        site {
+            siteMetadata {
+                nightly
+                title
+                description
+                navigation {
+                    text
+                    url
+                }
+                docSearch {
+                    apiKey
+                    indexName
+                }
+            }
+        }
+    }
+`

website/src/pages/404.md

@@ -1,7 +0,0 @@
----
-title: 404 Error
----
-
-import Error from 'widgets/404.js'
-
-<Error />

website/src/styles/alert.module.sass

@@ -3,11 +3,14 @@
     bottom: 0
     left: 0
     width: 100%
-    background: var(--color-subtle-light)
+    background: var(--color-back)
     z-index: 100
     font: var(--font-size-sm)/var(--line-height-md) var(--font-primary)
     text-align: center
     padding: 1rem
+    box-shadow: var(--box-shadow)
+    border-top: 2px solid
+    color: var(--color-theme)
 
 .warning
     --alert-bg: var(--color-yellow-light)

website/src/styles/layout.sass

@@ -47,6 +47,11 @@
     --color-theme-purple-light: hsla(255, 61%, 54%, 0.06)
     --color-theme-purple-opaque: hsla(255, 61%, 54%, 0.11)
 
+    --color-theme-nightly: hsl(257, 99%, 67%)
+    --color-theme-nightly-dark: hsl(257, 99%, 57%)
+    --color-theme-nightly-light: hsla(257, 99%, 67%, 0.06)
+    --color-theme-nightly-opaque: hsla(257, 99%, 67%, 0.11)
+
     // Regular colors
     --color-back: hsl(0, 0%, 100%)
     --color-front: hsl(213, 15%, 12%)
@@ -106,6 +111,12 @@
     --color-theme-light: var(--color-theme-purple-light)
     --color-theme-opaque: var(--color-theme-purple-opaque)
 
+.theme-nightly
+    --color-theme: var(--color-theme-nightly)
+    --color-theme-dark: var(--color-theme-nightly-dark)
+    --color-theme-light: var(--color-theme-nightly-light)
+    --color-theme-opaque: var(--color-theme-nightly-opaque)
+
 
 /* Fonts */
 

website/src/styles/sidebar.module.sass

@@ -22,6 +22,9 @@ $crumb-bar: 2px
     & > *
         padding: 0 2rem 0.35rem
 
+    &:last-child
+        margin-bottom: 5rem
+
 .label
     color: var(--color-dark)
     font: bold var(--font-size-lg)/var(--line-height-md) var(--font-secondary)

website/src/templates/docs.js

@@ -31,7 +31,7 @@ const Docs = ({ pageContext, children }) => (
                 theme,
                 version,
             } = pageContext
-            const { sidebars = [], modelsRepo, languages } = site.siteMetadata
+            const { sidebars = [], modelsRepo, languages, nightly } = site.siteMetadata
             const isModels = section === 'models'
             const sidebar = pageContext.sidebar
                 ? { items: pageContext.sidebar }
@@ -83,7 +83,7 @@ const Docs = ({ pageContext, children }) => (
                     {sidebar && <Sidebar items={sidebar.items} pageMenu={pageMenu} slug={slug} />}
                     <Main
                         section={section}
-                        theme={theme}
+                        theme={nightly ? 'nightly' : theme}
                         sidebar
                         asides
                         wrapContent
@@ -146,6 +146,7 @@ const query = graphql`
                     models
                     starters
                 }
+                nightly
                 sidebars {
                     section
                     items {

website/src/templates/index.js

@@ -75,10 +75,23 @@ const scopeComponents = {
     InlineCode,
 }
 
-const AlertSpace = () => {
+const AlertSpace = ({ nightly }) => {
     const isOnline = useOnlineStatus()
     return (
         <>
+            {nightly && (
+                <Alert
+                    title="You're viewing the pre-release docs."
+                    icon="moon"
+                    closeOnClick={false}
+                >
+                    The page reflects{' '}
+                    <Link to="https://pypi.org/project/spacy-nightly/">
+                        <InlineCode>spacy-nightly</InlineCode>
+                    </Link>
+                    , not the latest <Link to="https://spacy.io">stable version</Link>.
+                </Alert>
+            )}
             {!isOnline && (
                 <Alert title="Looks like you're offline." icon="offline" variant="warning">
                     But don't worry, your visited pages should be saved for you.
@@ -130,9 +143,10 @@ class Layout extends React.Component {
         const { data, pageContext, location, children } = this.props
         const { file, site = {} } = data || {}
         const mdx = file ? file.childMdx : null
-        const { title, section, sectionTitle, teaser, theme = 'blue', searchExclude } = pageContext
-        const bodyClass = classNames(`theme-${theme}`, { 'search-exclude': !!searchExclude })
         const meta = site.siteMetadata || {}
+        const { title, section, sectionTitle, teaser, theme = 'blue', searchExclude } = pageContext
+        const uiTheme = meta.nightly ? 'nightly' : theme
+        const bodyClass = classNames(`theme-${uiTheme}`, { 'search-exclude': !!searchExclude })
         const isDocs = ['usage', 'models', 'api', 'styleguide'].includes(section)
         const content = !mdx ? null : (
             <MDXProvider components={mdxComponents}>
@@ -148,8 +162,9 @@ class Layout extends React.Component {
                     section={section}
                     sectionTitle={sectionTitle}
                     bodyClass={bodyClass}
+                    nightly={meta.nightly}
                 />
-                <AlertSpace />
+                <AlertSpace nightly={meta.nightly} />
                 <Navigation
                     title={meta.title}
                     items={meta.navigation}
@@ -167,11 +182,11 @@ class Layout extends React.Component {
                         mdxComponents={mdxComponents}
                     />
                 ) : (
-                    <>
+                    <div>
                         {children}
                         {content}
                         <Footer wide />
-                    </>
+                    </div>
                 )}
             </>
         )
@@ -184,6 +199,7 @@ export const pageQuery = graphql`
     query($slug: String!) {
         site {
             siteMetadata {
+                nightly
                 title
                 description
                 navigation {

website/src/templates/universe.js

@@ -30,8 +30,8 @@ function filterResources(resources, data) {
     return sorted.filter(res => (res.category || []).includes(data.id))
 }
 
-const UniverseContent = ({ content = [], categories, pageContext, location, mdxComponents }) => {
+const UniverseContent = ({ content = [], categories, theme, pageContext, mdxComponents }) => {
-    const { theme, data = {} } = pageContext
+    const { data = {} } = pageContext
     const filteredResources = filterResources(content, data)
     const activeData = data ? content.find(({ id }) => id === data.id) : null
     const markdownComponents = { ...mdxComponents, code: InlineCode }
@@ -302,15 +302,16 @@ const Universe = ({ pageContext, location, mdxComponents }) => (
     <StaticQuery
         query={query}
         render={data => {
-            const content = data.site.siteMetadata.universe.resources
+            const { universe, nightly } = data.site.siteMetadata
-            const categories = data.site.siteMetadata.universe.categories
+            const theme = nightly ? 'nightly' : pageContext.theme
             return (
                 <UniverseContent
-                    content={content}
+                    content={universe.resources}
-                    categories={categories}
+                    categories={universe.categories}
                     pageContext={pageContext}
                     location={location}
                     mdxComponents={mdxComponents}
+                    theme={theme}
                 />
             )
         }}
@@ -323,6 +324,7 @@ const query = graphql`
     query UniverseQuery {
         site {
             siteMetadata {
+                nightly
                 universe {
                     resources {
                         type

website/src/widgets/404.js

@@ -1,19 +0,0 @@
-import React from 'react'
-import { window } from 'browser-monads'
-
-import { LandingHeader, LandingTitle } from '../components/landing'
-import Button from '../components/button'
-
-export default () => (
-    <LandingHeader style={{ minHeight: 400 }}>
-        <LandingTitle>
-            Ooops, this page
-            <br />
-            does not exist!
-        </LandingTitle>
-        <br />
-        <Button onClick={() => window.history.go(-1)} variant="tertiary">
-            Click here to go back
-        </Button>
-    </LandingHeader>
-)

website/src/widgets/landing.js

@@ -68,7 +68,7 @@ const Landing = ({ data }) => {
     const counts = getCounts(data.languages)
     return (
         <>
-            <LandingHeader>
+            <LandingHeader nightly={data.nightly}>
                 <LandingTitle>
                     Industrial-Strength
                     <br />
@@ -268,6 +268,7 @@ const landingQuery = graphql`
     query LandingQuery {
         site {
             siteMetadata {
+                nightly
                 repo
                 languages {
                     models