Change fasttext-wheel to fasttext in requirements.txt#470
Merged
SunnyHaze merged 2 commits intoOpenDCAI:mainfrom Feb 9, 2026
Merged
Change fasttext-wheel to fasttext in requirements.txt#470SunnyHaze merged 2 commits intoOpenDCAI:mainfrom
SunnyHaze merged 2 commits intoOpenDCAI:mainfrom
Conversation
Contributor
Author
|
Here's the diff between the two dependency versions, if you want to review: diff -u -r fasttext-wheel-0.9.2/PKG-INFO fasttext-0.9.3/PKG-INFO
--- fasttext-wheel-0.9.2/PKG-INFO 2023-05-05 07:58:47.102100400 -0400
+++ fasttext-0.9.3/PKG-INFO 2024-06-12 05:28:01.705583800 -0400
@@ -1,6 +1,6 @@
Metadata-Version: 2.1
-Name: fasttext-wheel
-Version: 0.9.2
+Name: fasttext
+Version: 0.9.3
Summary: fasttext Python bindings
Home-page: https://github.com/facebookresearch/fastText
Author: Onur Celebi
@@ -21,6 +21,9 @@
Classifier: Operating System :: Unix
Classifier: Operating System :: MacOS
License-File: LICENSE
+Requires-Dist: pybind11>=2.2
+Requires-Dist: setuptools>=0.7.0
+Requires-Dist: numpy
fastText |CircleCI|
===================
diff -u -r fasttext-wheel-0.9.2/python/fasttext_module/fasttext/FastText.py fasttext-0.9.3/python/fasttext_module/fasttext/FastText.py
--- fasttext-wheel-0.9.2/python/fasttext_module/fasttext/FastText.py 2023-05-05 07:58:14.000000000 -0400
+++ fasttext-0.9.3/python/fasttext_module/fasttext/FastText.py 2024-06-12 04:55:55.000000000 -0400
@@ -12,7 +12,6 @@
import fasttext_pybind as fasttext
import numpy as np
import multiprocessing
-import sys
from itertools import chain
loss_name = fasttext.loss_name
@@ -24,11 +23,7 @@
displayed_errors = {}
-def eprint(*args, **kwargs):
- print(*args, file=sys.stderr, **kwargs)
-
-
-class _Meter(object):
+class _Meter:
def __init__(self, fasttext_model, meter):
self.f = fasttext_model
self.m = meter
@@ -81,7 +76,7 @@
return recall
-class _FastText(object):
+class _FastText:
"""
This class defines the API to inspect models and should not be used to
create objects. It will be returned by functions such as load_model or
@@ -102,10 +97,26 @@
def set_args(self, args=None):
if args:
- arg_names = ['lr', 'dim', 'ws', 'epoch', 'minCount',
- 'minCountLabel', 'minn', 'maxn', 'neg', 'wordNgrams',
- 'loss', 'bucket', 'thread', 'lrUpdateRate', 't',
- 'label', 'verbose', 'pretrainedVectors']
+ arg_names = [
+ "lr",
+ "dim",
+ "ws",
+ "epoch",
+ "minCount",
+ "minCountLabel",
+ "minn",
+ "maxn",
+ "neg",
+ "wordNgrams",
+ "loss",
+ "bucket",
+ "thread",
+ "lrUpdateRate",
+ "t",
+ "label",
+ "verbose",
+ "pretrainedVectors",
+ ]
for arg_name in arg_names:
setattr(self, arg_name, getattr(args, arg_name))
@@ -131,21 +142,18 @@
whitespace (space, newline, tab, vertical tab) and the control
characters carriage return, formfeed and the null character.
"""
- if text.find('\n') != -1:
- raise ValueError(
- "predict processes one line at a time (remove \'\\n\')"
- )
+ if text.find("\n") != -1:
+ raise ValueError("predict processes one line at a time (remove '\\n')")
text += "\n"
dim = self.get_dimension()
b = fasttext.Vector(dim)
self.f.getSentenceVector(b, text)
return np.array(b)
- def get_nearest_neighbors(self, word, k=10, on_unicode_error='strict'):
+ def get_nearest_neighbors(self, word, k=10, on_unicode_error="strict"):
return self.f.getNN(word, k, on_unicode_error)
- def get_analogies(self, wordA, wordB, wordC, k=10,
- on_unicode_error='strict'):
+ def get_analogies(self, wordA, wordB, wordC, k=10, on_unicode_error="strict"):
return self.f.getAnalogies(wordA, wordB, wordC, k, on_unicode_error)
def get_word_id(self, word):
@@ -168,7 +176,7 @@
"""
return self.f.getSubwordId(subword)
- def get_subwords(self, word, on_unicode_error='strict'):
+ def get_subwords(self, word, on_unicode_error="strict"):
"""
Given a word, get the subwords and their indicies.
"""
@@ -184,7 +192,7 @@
self.f.getInputVector(b, ind)
return np.array(b)
- def predict(self, text, k=1, threshold=0.0, on_unicode_error='strict'):
+ def predict(self, text, k=1, threshold=0.0, on_unicode_error="strict"):
"""
Given a string, get a list of labels and a list of
corresponding probabilities. k controls the number
@@ -208,17 +216,16 @@
"""
def check(entry):
- if entry.find('\n') != -1:
- raise ValueError(
- "predict processes one line at a time (remove \'\\n\')"
- )
+ if entry.find("\n") != -1:
+ raise ValueError("predict processes one line at a time (remove '\\n')")
entry += "\n"
return entry
if type(text) == list:
text = [check(entry) for entry in text]
all_labels, all_probs = self.f.multilinePredict(
- text, k, threshold, on_unicode_error)
+ text, k, threshold, on_unicode_error
+ )
return all_labels, all_probs
else:
@@ -249,7 +256,7 @@
raise ValueError("Can't get quantized Matrix")
return np.array(self.f.getOutputMatrix())
- def get_words(self, include_freq=False, on_unicode_error='strict'):
+ def get_words(self, include_freq=False, on_unicode_error="strict"):
"""
Get the entire list of words of the dictionary optionally
including the frequency of the individual words. This
@@ -262,7 +269,7 @@
else:
return pair[0]
- def get_labels(self, include_freq=False, on_unicode_error='strict'):
+ def get_labels(self, include_freq=False, on_unicode_error="strict"):
"""
Get the entire list of labels of the dictionary optionally
including the frequency of the individual labels. Unsupervised
@@ -280,17 +287,15 @@
else:
return self.get_words(include_freq)
- def get_line(self, text, on_unicode_error='strict'):
+ def get_line(self, text, on_unicode_error="strict"):
"""
Split a line of text into words and labels. Labels must start with
the prefix used to create the model (__label__ by default).
"""
def check(entry):
- if entry.find('\n') != -1:
- raise ValueError(
- "get_line processes one line at a time (remove \'\\n\')"
- )
+ if entry.find("\n") != -1:
+ raise ValueError("get_line processes one line at a time (remove '\\n')")
entry += "\n"
return entry
@@ -336,7 +341,7 @@
thread=None,
verbose=None,
dsub=2,
- qnorm=False
+ qnorm=False,
):
"""
Quantize the model reducing the size of the model and
@@ -356,8 +361,7 @@
if input is None:
input = ""
self.f.quantize(
- input, qout, cutoff, retrain, epoch, lr, thread, verbose, dsub,
- qnorm
+ input, qout, cutoff, retrain, epoch, lr, thread, verbose, dsub, qnorm
)
def set_matrices(self, input_matrix, output_matrix):
@@ -365,8 +369,9 @@
Set input and output matrices. This function assumes you know what you
are doing.
"""
- self.f.setMatrices(input_matrix.astype(np.float32),
- output_matrix.astype(np.float32))
+ self.f.setMatrices(
+ input_matrix.astype(np.float32), output_matrix.astype(np.float32)
+ )
@property
def words(self):
@@ -437,46 +442,45 @@
def load_model(path):
"""Load a model given a filepath and return a model object."""
- eprint("Warning : `load_model` does not return WordVectorModel or SupervisedModel any more, but a `FastText` object which is very similar.")
return _FastText(model_path=path)
unsupervised_default = {
- 'model': "skipgram",
- 'lr': 0.05,
- 'dim': 100,
- 'ws': 5,
- 'epoch': 5,
- 'minCount': 5,
- 'minCountLabel': 0,
- 'minn': 3,
- 'maxn': 6,
- 'neg': 5,
- 'wordNgrams': 1,
- 'loss': "ns",
- 'bucket': 2000000,
- 'thread': multiprocessing.cpu_count() - 1,
- 'lrUpdateRate': 100,
- 't': 1e-4,
- 'label': "__label__",
- 'verbose': 2,
- 'pretrainedVectors': "",
- 'seed': 0,
- 'autotuneValidationFile': "",
- 'autotuneMetric': "f1",
- 'autotunePredictions': 1,
- 'autotuneDuration': 60 * 5, # 5 minutes
- 'autotuneModelSize': ""
+ "model": "skipgram",
+ "lr": 0.05,
+ "dim": 100,
+ "ws": 5,
+ "epoch": 5,
+ "minCount": 5,
+ "minCountLabel": 0,
+ "minn": 3,
+ "maxn": 6,
+ "neg": 5,
+ "wordNgrams": 1,
+ "loss": "ns",
+ "bucket": 2000000,
+ "thread": multiprocessing.cpu_count() - 1,
+ "lrUpdateRate": 100,
+ "t": 1e-4,
+ "label": "__label__",
+ "verbose": 2,
+ "pretrainedVectors": "",
+ "seed": 0,
+ "autotuneValidationFile": "",
+ "autotuneMetric": "f1",
+ "autotunePredictions": 1,
+ "autotuneDuration": 60 * 5, # 5 minutes
+ "autotuneModelSize": "",
}
def read_args(arg_list, arg_dict, arg_names, default_values):
param_map = {
- 'min_count': 'minCount',
- 'word_ngrams': 'wordNgrams',
- 'lr_update_rate': 'lrUpdateRate',
- 'label_prefix': 'label',
- 'pretrained_vectors': 'pretrainedVectors'
+ "min_count": "minCount",
+ "word_ngrams": "wordNgrams",
+ "lr_update_rate": "lrUpdateRate",
+ "label_prefix": "label",
+ "pretrained_vectors": "pretrainedVectors",
}
ret = {}
@@ -512,22 +516,45 @@
repository such as the dataset pulled by classification-example.sh.
"""
supervised_default = unsupervised_default.copy()
- supervised_default.update({
- 'lr': 0.1,
- 'minCount': 1,
- 'minn': 0,
- 'maxn': 0,
- 'loss': "softmax",
- 'model': "supervised"
- })
-
- arg_names = ['input', 'lr', 'dim', 'ws', 'epoch', 'minCount',
- 'minCountLabel', 'minn', 'maxn', 'neg', 'wordNgrams', 'loss', 'bucket',
- 'thread', 'lrUpdateRate', 't', 'label', 'verbose', 'pretrainedVectors',
- 'seed', 'autotuneValidationFile', 'autotuneMetric',
- 'autotunePredictions', 'autotuneDuration', 'autotuneModelSize']
- args, manually_set_args = read_args(kargs, kwargs, arg_names,
- supervised_default)
+ supervised_default.update(
+ {
+ "lr": 0.1,
+ "minCount": 1,
+ "minn": 0,
+ "maxn": 0,
+ "loss": "softmax",
+ "model": "supervised",
+ }
+ )
+
+ arg_names = [
+ "input",
+ "lr",
+ "dim",
+ "ws",
+ "epoch",
+ "minCount",
+ "minCountLabel",
+ "minn",
+ "maxn",
+ "neg",
+ "wordNgrams",
+ "loss",
+ "bucket",
+ "thread",
+ "lrUpdateRate",
+ "t",
+ "label",
+ "verbose",
+ "pretrainedVectors",
+ "seed",
+ "autotuneValidationFile",
+ "autotuneMetric",
+ "autotunePredictions",
+ "autotuneDuration",
+ "autotuneModelSize",
+ ]
+ args, manually_set_args = read_args(kargs, kwargs, arg_names, supervised_default)
a = _build_args(args, manually_set_args)
ft = _FastText(args=a)
fasttext.train(ft.f, a)
@@ -549,11 +576,29 @@
dataset pulled by the example script word-vector-example.sh, which is
part of the fastText repository.
"""
- arg_names = ['input', 'model', 'lr', 'dim', 'ws', 'epoch', 'minCount',
- 'minCountLabel', 'minn', 'maxn', 'neg', 'wordNgrams', 'loss', 'bucket',
- 'thread', 'lrUpdateRate', 't', 'label', 'verbose', 'pretrainedVectors']
- args, manually_set_args = read_args(kargs, kwargs, arg_names,
- unsupervised_default)
+ arg_names = [
+ "input",
+ "model",
+ "lr",
+ "dim",
+ "ws",
+ "epoch",
+ "minCount",
+ "minCountLabel",
+ "minn",
+ "maxn",
+ "neg",
+ "wordNgrams",
+ "loss",
+ "bucket",
+ "thread",
+ "lrUpdateRate",
+ "t",
+ "label",
+ "verbose",
+ "pretrainedVectors",
+ ]
+ args, manually_set_args = read_args(kargs, kwargs, arg_names, unsupervised_default)
a = _build_args(args, manually_set_args)
ft = _FastText(args=a)
fasttext.train(ft.f, a)
@@ -562,12 +607,18 @@
def cbow(*kargs, **kwargs):
- raise Exception("`cbow` is not supported any more. Please use `train_unsupervised` with model=`cbow`. For more information please refer to https://fasttext.cc/blog/2019/06/25/blog-post.html#2-you-were-using-the-unofficial-fasttext-module")
+ raise Exception(
+ "`cbow` is not supported any more. Please use `train_unsupervised` with model=`cbow`. For more information please refer to https://fasttext.cc/blog/2019/06/25/blog-post.html#2-you-were-using-the-unofficial-fasttext-module"
+ )
def skipgram(*kargs, **kwargs):
- raise Exception("`skipgram` is not supported any more. Please use `train_unsupervised` with model=`skipgram`. For more information please refer to https://fasttext.cc/blog/2019/06/25/blog-post.html#2-you-were-using-the-unofficial-fasttext-module")
+ raise Exception(
+ "`skipgram` is not supported any more. Please use `train_unsupervised` with model=`skipgram`. For more information please refer to https://fasttext.cc/blog/2019/06/25/blog-post.html#2-you-were-using-the-unofficial-fasttext-module"
+ )
def supervised(*kargs, **kwargs):
- raise Exception("`supervised` is not supported any more. Please use `train_supervised`. For more information please refer to https://fasttext.cc/blog/2019/06/25/blog-post.html#2-you-were-using-the-unofficial-fasttext-module")
+ raise Exception(
+ "`supervised` is not supported any more. Please use `train_supervised`. For more information please refer to https://fasttext.cc/blog/2019/06/25/blog-post.html#2-you-were-using-the-unofficial-fasttext-module"
+ )
diff -u -r fasttext-wheel-0.9.2/python/fasttext_module/fasttext/pybind/fasttext_pybind.cc fasttext-0.9.3/python/fasttext_module/fasttext/pybind/fasttext_pybind.cc
--- fasttext-wheel-0.9.2/python/fasttext_module/fasttext/pybind/fasttext_pybind.cc 2023-05-05 07:58:14.000000000 -0400
+++ fasttext-0.9.3/python/fasttext_module/fasttext/pybind/fasttext_pybind.cc 2024-06-12 04:55:55.000000000 -0400
@@ -78,8 +78,9 @@
} else if (type == fasttext::entry_type::label && wid >= 0) {
labels.push_back(castToPythonString(token, onUnicodeError));
}
- if (token == fasttext::Dictionary::EOS)
+ if (token == fasttext::Dictionary::EOS) {
break;
+}
}
return std::pair<std::vector<py::str>, std::vector<py::str>>(words, labels);
}
@@ -199,28 +200,29 @@
.def("scoreVsTrue", &fasttext::Meter::scoreVsTrue)
.def(
"precisionRecallCurveLabel",
- py::overload_cast<int32_t>(
- &fasttext::Meter::precisionRecallCurve, py::const_))
+ (std::vector<std::pair<double, double>>(fasttext::Meter::*)(int32_t)
+ const) &
+ fasttext::Meter::precisionRecallCurve)
.def(
"precisionRecallCurve",
- py::overload_cast<>(
- &fasttext::Meter::precisionRecallCurve, py::const_))
+ (std::vector<std::pair<double, double>>(fasttext::Meter::*)() const) &
+ fasttext::Meter::precisionRecallCurve)
.def(
"precisionAtRecallLabel",
- py::overload_cast<int32_t, double>(
- &fasttext::Meter::precisionAtRecall, py::const_))
+ (double (fasttext::Meter::*)(int32_t, double) const) &
+ fasttext::Meter::precisionAtRecall)
.def(
"precisionAtRecall",
- py::overload_cast<double>(
- &fasttext::Meter::precisionAtRecall, py::const_))
+ (double (fasttext::Meter::*)(double) const) &
+ fasttext::Meter::precisionAtRecall)
.def(
"recallAtPrecisionLabel",
- py::overload_cast<int32_t, double>(
- &fasttext::Meter::recallAtPrecision, py::const_))
+ (double (fasttext::Meter::*)(int32_t, double) const) &
+ fasttext::Meter::recallAtPrecision)
.def(
"recallAtPrecision",
- py::overload_cast<double>(
- &fasttext::Meter::recallAtPrecision, py::const_));
+ (double (fasttext::Meter::*)(double) const) &
+ fasttext::Meter::recallAtPrecision);
py::class_<fasttext::FastText>(m, "fasttext")
.def(py::init<>())
Only in fasttext-0.9.3/python/fasttext_module: fasttext.egg-info
Only in fasttext-wheel-0.9.2/python/fasttext_module: fasttext_wheel.egg-info
diff -u -r fasttext-wheel-0.9.2/setup.py fasttext-0.9.3/setup.py
--- fasttext-wheel-0.9.2/setup.py 2023-05-05 07:58:44.000000000 -0400
+++ fasttext-0.9.3/setup.py 2024-06-12 05:27:28.000000000 -0400
@@ -21,33 +21,36 @@
import platform
import io
-__version__ = '0.9.2'
+__version__ = "0.9.3"
FASTTEXT_SRC = "src"
# Based on https://github.com/pybind/python_example
-class get_pybind_include(object):
+
+class get_pybind_include:
"""Helper class to determine the pybind11 include path
The purpose of this class is to postpone importing pybind11
until it is actually installed, so that the ``get_include()``
- method can be invoked. """
+ method can be invoked."""
def __init__(self, user=False):
try:
- import pybind11
+ pass
except ImportError:
- if subprocess.call([sys.executable, '-m', 'pip', 'install', 'pybind11']):
- raise RuntimeError('pybind11 install failed.')
+ if subprocess.call([sys.executable, "-m", "pip", "install", "pybind11"]):
+ raise RuntimeError("pybind11 install failed.")
self.user = user
def __str__(self):
import pybind11
+
return pybind11.get_include(self.user)
+
try:
- coverage_index = sys.argv.index('--coverage')
+ coverage_index = sys.argv.index("--coverage")
except ValueError:
coverage = False
else:
@@ -55,7 +58,7 @@
coverage = True
fasttext_src_files = map(str, os.listdir(FASTTEXT_SRC))
-fasttext_src_cc = list(filter(lambda x: x.endswith('.cc'), fasttext_src_files))
+fasttext_src_cc = list(filter(lambda x: x.endswith(".cc"), fasttext_src_files))
fasttext_src_cc = list(
map(lambda x: str(os.path.join(FASTTEXT_SRC, x)), fasttext_src_cc)
@@ -63,10 +66,11 @@
ext_modules = [
Extension(
- str('fasttext_pybind'),
+ str("fasttext_pybind"),
[
- str('python/fasttext_module/fasttext/pybind/fasttext_pybind.cc'),
- ] + fasttext_src_cc,
+ str("python/fasttext_module/fasttext/pybind/fasttext_pybind.cc"),
+ ]
+ + fasttext_src_cc,
include_dirs=[
# Path to pybind11 headers
get_pybind_include(),
@@ -74,9 +78,12 @@
# Path to fasttext source code
FASTTEXT_SRC,
],
- language='c++',
- extra_compile_args=["-O0 -fno-inline -fprofile-arcs -pthread -march=native" if coverage else
- "-O3 -funroll-loops -pthread -march=native"],
+ language="c++",
+ extra_compile_args=[
+ "-O0 -fno-inline -fprofile-arcs -pthread -march=native"
+ if coverage
+ else "-O3 -funroll-loops -pthread -march=native"
+ ],
),
]
@@ -88,8 +95,9 @@
the specified compiler.
"""
import tempfile
- with tempfile.NamedTemporaryFile('w', suffix='.cpp') as f:
- f.write('int main (int argc, char **argv) { return 0; }')
+
+ with tempfile.NamedTemporaryFile("w", suffix=".cpp") as f:
+ f.write("int main (int argc, char **argv) { return 0; }")
try:
compiler.compile([f.name], extra_postargs=flags)
except setuptools.distutils.errors.CompileError:
@@ -98,58 +106,53 @@
def cpp_flag(compiler):
- """Return the -std=c++[11/14] compiler flag.
- The c++14 is preferred over c++11 (when it is available).
- """
- standards = ['-std=c++14', '-std=c++11']
+ """Return the -std=c++17 compiler flag."""
+ standards = ["-std=c++17"]
for standard in standards:
if has_flag(compiler, [standard]):
return standard
- raise RuntimeError(
- 'Unsupported compiler -- at least C++11 support '
- 'is needed!'
- )
+ raise RuntimeError("Unsupported compiler -- at least C++17 support " "is needed!")
class BuildExt(build_ext):
"""A custom build extension for adding compiler-specific options."""
+
c_opts = {
- 'msvc': ['/EHsc'],
- 'unix': [],
+ "msvc": ["/EHsc"],
+ "unix": [],
}
def build_extensions(self):
- if sys.platform == 'darwin':
- mac_osx_version = float('.'.join(platform.mac_ver()[0].split('.')[:2]))
- os.environ['MACOSX_DEPLOYMENT_TARGET'] = str(mac_osx_version)
- all_flags = ['-stdlib=libc++', '-mmacosx-version-min=10.7']
+ if sys.platform == "darwin":
+ mac_osx_version = float(".".join(platform.mac_ver()[0].split(".")[:2]))
+ os.environ["MACOSX_DEPLOYMENT_TARGET"] = str(mac_osx_version)
+ all_flags = ["-stdlib=libc++", "-mmacosx-version-min=10.7"]
if has_flag(self.compiler, [all_flags[0]]):
- self.c_opts['unix'] += [all_flags[0]]
+ self.c_opts["unix"] += [all_flags[0]]
elif has_flag(self.compiler, all_flags):
- self.c_opts['unix'] += all_flags
+ self.c_opts["unix"] += all_flags
else:
raise RuntimeError(
- 'libc++ is needed! Failed to compile with {} and {}.'.
- format(" ".join(all_flags), all_flags[0])
+ "libc++ is needed! Failed to compile with {} and {}.".format(
+ " ".join(all_flags), all_flags[0]
+ )
)
ct = self.compiler.compiler_type
opts = self.c_opts.get(ct, [])
extra_link_args = []
if coverage:
- coverage_option = '--coverage'
+ coverage_option = "--coverage"
opts.append(coverage_option)
extra_link_args.append(coverage_option)
- if ct == 'unix':
+ if ct == "unix":
opts.append('-DVERSION_INFO="%s"' % self.distribution.get_version())
opts.append(cpp_flag(self.compiler))
- if has_flag(self.compiler, ['-fvisibility=hidden']):
- opts.append('-fvisibility=hidden')
- elif ct == 'msvc':
- opts.append(
- '/DVERSION_INFO=\\"%s\\"' % self.distribution.get_version()
- )
+ if has_flag(self.compiler, ["-fvisibility=hidden"]):
+ opts.append("-fvisibility=hidden")
+ elif ct == "msvc":
+ opts.append('/DVERSION_INFO=\\"%s\\"' % self.distribution.get_version())
for ext in self.extensions:
ext.extra_compile_args = opts
ext.extra_link_args = extra_link_args
@@ -161,43 +164,43 @@
Use pandoc to generate rst from md.
pandoc --from=markdown --to=rst --output=python/README.rst python/README.md
"""
- with io.open("python/README.rst", encoding='utf-8') as fid:
+ with io.open("python/README.rst", encoding="utf-8") as fid:
return fid.read()
setup(
- name='fasttext-wheel',
+ name="fasttext",
version=__version__,
- author='Onur Celebi',
- author_email='celebio@fb.com',
- description='fasttext Python bindings',
+ author="Onur Celebi",
+ author_email="celebio@fb.com",
+ description="fasttext Python bindings",
long_description=_get_readme(),
ext_modules=ext_modules,
- url='https://github.com/facebookresearch/fastText',
- license='MIT',
+ url="https://github.com/facebookresearch/fastText",
+ license="MIT",
classifiers=[
- 'Development Status :: 3 - Alpha',
- 'Intended Audience :: Developers',
- 'Intended Audience :: Science/Research',
- 'License :: OSI Approved :: MIT License',
- 'Programming Language :: Python :: 2.7',
- 'Programming Language :: Python :: 3.4',
- 'Programming Language :: Python :: 3.5',
- 'Programming Language :: Python :: 3.6',
- 'Topic :: Software Development',
- 'Topic :: Scientific/Engineering',
- 'Operating System :: Microsoft :: Windows',
- 'Operating System :: POSIX',
- 'Operating System :: Unix',
- 'Operating System :: MacOS',
+ "Development Status :: 3 - Alpha",
+ "Intended Audience :: Developers",
+ "Intended Audience :: Science/Research",
+ "License :: OSI Approved :: MIT License",
+ "Programming Language :: Python :: 2.7",
+ "Programming Language :: Python :: 3.4",
+ "Programming Language :: Python :: 3.5",
+ "Programming Language :: Python :: 3.6",
+ "Topic :: Software Development",
+ "Topic :: Scientific/Engineering",
+ "Operating System :: Microsoft :: Windows",
+ "Operating System :: POSIX",
+ "Operating System :: Unix",
+ "Operating System :: MacOS",
],
- install_requires=['pybind11>=2.2', "setuptools >= 0.7.0", "numpy"],
- cmdclass={'build_ext': BuildExt},
+ install_requires=["pybind11>=2.2", "setuptools >= 0.7.0", "numpy"],
+ cmdclass={"build_ext": BuildExt},
packages=[
- str('fasttext'),
- str('fasttext.util'),
- str('fasttext.tests'),
+ str("fasttext"),
+ str("fasttext.util"),
+ str("fasttext.tests"),
],
- package_dir={str(''): str('python/fasttext_module')},
+ package_dir={str(""): str("python/fasttext_module")},
zip_safe=False,
)
Only in fasttext-0.9.3/src: aligned.h
diff -u -r fasttext-wheel-0.9.2/src/args.cc fasttext-0.9.3/src/args.cc
--- fasttext-wheel-0.9.2/src/args.cc 2023-05-05 07:58:14.000000000 -0400
+++ fasttext-0.9.3/src/args.cc 2024-06-12 04:55:55.000000000 -0400
@@ -8,7 +8,8 @@
#include "args.h"
-#include <stdlib.h>
+#include <cstdlib>
+#include <cstdint>
#include <iostream>
#include <stdexcept>
@@ -106,7 +107,7 @@
}
void Args::parseArgs(const std::vector<std::string>& args) {
- std::string command(args[1]);
+ const std::string& command(args[1]);
if (command == "supervised") {
model = model_name::sup;
loss = loss_name::softmax;
@@ -401,13 +402,13 @@
} else if (autotuneMetric == "f1") {
return metric_name::f1score;
} else if (autotuneMetric.substr(0, 18) == "precisionAtRecall:") {
- size_t semicolon = autotuneMetric.find(":", 18);
+ size_t semicolon = autotuneMetric.find(':', 18);
if (semicolon != std::string::npos) {
return metric_name::precisionAtRecallLabel;
}
return metric_name::precisionAtRecall;
} else if (autotuneMetric.substr(0, 18) == "recallAtPrecision:") {
- size_t semicolon = autotuneMetric.find(":", 18);
+ size_t semicolon = autotuneMetric.find(':', 18);
if (semicolon != std::string::npos) {
return metric_name::recallAtPrecisionLabel;
}
@@ -424,7 +425,7 @@
} else if (
metric == metric_name::precisionAtRecallLabel ||
metric == metric_name::recallAtPrecisionLabel) {
- size_t semicolon = autotuneMetric.find(":", 18);
+ size_t semicolon = autotuneMetric.find(':', 18);
label = autotuneMetric.substr(semicolon + 1);
} else {
return label;
@@ -444,7 +445,7 @@
metric == metric_name::recallAtPrecisionLabel ||
metric == metric_name::recallAtPrecision) {
size_t firstSemicolon = 18; // semicolon position in "precisionAtRecall:"
- size_t secondSemicolon = autotuneMetric.find(":", firstSemicolon);
+ size_t secondSemicolon = autotuneMetric.find(':', firstSemicolon);
const std::string valueStr =
autotuneMetric.substr(firstSemicolon, secondSemicolon - firstSemicolon);
value = std::stof(valueStr) / 100.0;
diff -u -r fasttext-wheel-0.9.2/src/densematrix.cc fasttext-0.9.3/src/densematrix.cc
--- fasttext-wheel-0.9.2/src/densematrix.cc 2023-05-05 07:58:14.000000000 -0400
+++ fasttext-0.9.3/src/densematrix.cc 2024-06-12 04:55:55.000000000 -0400
@@ -15,6 +15,10 @@
#include "utils.h"
#include "vector.h"
+#if defined(__AVX512F__) || defined(__AVX__) || defined(__SSE__)
+#include <immintrin.h>
+#endif
+
namespace fasttext {
DenseMatrix::DenseMatrix() : DenseMatrix(0, 0) {}
@@ -146,6 +150,92 @@
}
}
+/* Abstract over AVX512F, AVX, and SSE intrinsics, using the one available on this machine. */
+#if defined(__AVX512F__)
+using Register = __m512;
+inline Register Add(Register first, Register second) { return _mm512_add_ps(first, second); }
+inline Register Set1(float to) { return _mm512_set1_ps(to); }
+inline Register Multiply(Register first, Register second) { return _mm512_mul_ps(first, second); }
+#elif defined(__AVX__)
+using Register = __m256;
+inline Register Add(Register first, Register second) { return _mm256_add_ps(first, second); }
+inline Register Set1(float to) { return _mm256_set1_ps(to); }
+inline Register Multiply(Register first, Register second) { return _mm256_mul_ps(first, second); }
+#elif defined(__SSE__)
+using Register = __m128;
+inline Register Add(Register first, Register second) { return _mm_add_ps(first, second); }
+inline Register Set1(float to) { return _mm_set1_ps(to); }
+inline Register Multiply(Register first, Register second) { return _mm_mul_ps(first, second); }
+#endif
+
+/* Faster routine for averaging rows of a matrix on x86.
+ * The idea here is to keep the accumulators in registers if possible. */
+#if defined(__AVX512F__) || defined(__AVX__) || defined(__SSE__)
+template <unsigned Cols> void averageRowsFast(Vector& x, const std::vector<int32_t>& rows, const DenseMatrix &matrix) {
+ // Columns must be a multiple of how many floats fit in a register.
+ static_assert(Cols % (sizeof(Register) / 4) == 0);
+ constexpr unsigned RegisterCount = Cols / (sizeof(Register) / 4);
+ // These should be aligned by aligned.h
+ assert(reinterpret_cast<uintptr_t>(x.data()) % sizeof(Register) == 0);
+ assert(reinterpret_cast<uintptr_t>(matrix.data()) % sizeof(Register) == 0);
+
+ // Guard against empty list of rows with default NaN behavior.
+ if (rows.empty()) {
+ x.zero();
+ x.mul(1.0 / rows.size());
+ return;
+ }
+
+ // Copy the first row to accumulation registers.
+ Register accum[RegisterCount];
+ auto row = rows.cbegin();
+ const Register *base = reinterpret_cast<const Register*>(matrix.data() + matrix.cols() * *row);
+ for (unsigned i = 0; i < RegisterCount; ++i) {
+ accum[i] = base[i];
+ }
+ // Add the rows after the first.
+ for (++row; row != rows.cend(); ++row) {
+ base = reinterpret_cast<const Register*>(matrix.data() + matrix.cols() * *row);
+ for (unsigned i = 0; i < RegisterCount; ++i) {
+ accum[i] = Add(accum[i], base[i]);
+ }
+ }
+ // Multiply by (1.0 / rows.size()) and write to x.
+ Register mul = Set1(1.0 / rows.size());
+ for (unsigned i = 0; i < RegisterCount; ++i) {
+ reinterpret_cast<Register*>(x.data())[i] = Multiply(accum[i], mul);
+ }
+}
+#endif
+
+void DenseMatrix::averageRowsToVector(Vector& x, const std::vector<int32_t>& rows) const {
+#if defined(__AVX512F__) || defined(__AVX__) || defined(__SSE__)
+ switch (cols()) {
+ case 512:
+ // Maximum number that can fit all in registers on AVX512F.
+ averageRowsFast<512>(x, rows, *this);
+ return;
+ case 256:
+ averageRowsFast<256>(x, rows, *this);
+ return;
+ case 64:
+ averageRowsFast<64>(x, rows, *this);
+ return;
+ case 32:
+ averageRowsFast<32>(x, rows, *this);
+ return;
+ case 16:
+ averageRowsFast<16>(x, rows, *this);
+ return;
+ }
+#endif
+ x.zero();
+ for (auto it = rows.cbegin(); it != rows.cend(); ++it) {
+ addRowToVector(x, *it);
+ }
+ x.mul(1.0 / rows.size());
+}
+
void DenseMatrix::save(std::ostream& out) const {
out.write((char*)&m_, sizeof(int64_t));
out.write((char*)&n_, sizeof(int64_t));
@@ -155,7 +245,7 @@
void DenseMatrix::load(std::istream& in) {
in.read((char*)&m_, sizeof(int64_t));
in.read((char*)&n_, sizeof(int64_t));
- data_ = std::vector<real>(m_ * n_);
+ data_ = intgemm::AlignedVector<real>(m_ * n_);
in.read((char*)data_.data(), m_ * n_ * sizeof(real));
}
diff -u -r fasttext-wheel-0.9.2/src/densematrix.h fasttext-0.9.3/src/densematrix.h
--- fasttext-wheel-0.9.2/src/densematrix.h 2023-05-05 07:58:14.000000000 -0400
+++ fasttext-0.9.3/src/densematrix.h 2024-06-12 04:55:55.000000000 -0400
@@ -15,6 +15,7 @@
#include <stdexcept>
#include <vector>
+#include "aligned.h"
#include "matrix.h"
#include "real.h"
@@ -24,7 +25,7 @@
class DenseMatrix : public Matrix {
protected:
- std::vector<real> data_;
+ intgemm::AlignedVector<real> data_;
void uniformThread(real, int, int32_t);
public:
@@ -71,6 +72,7 @@
void addVectorToRow(const Vector&, int64_t, real) override;
void addRowToVector(Vector& x, int32_t i) const override;
void addRowToVector(Vector& x, int32_t i, real a) const override;
+ void averageRowsToVector(Vector& x, const std::vector<int32_t>& rows) const override;
void save(std::ostream&) const override;
void load(std::istream&) override;
void dump(std::ostream&) const override;
diff -u -r fasttext-wheel-0.9.2/src/dictionary.cc fasttext-0.9.3/src/dictionary.cc
--- fasttext-wheel-0.9.2/src/dictionary.cc 2023-05-05 07:58:14.000000000 -0400
+++ fasttext-0.9.3/src/dictionary.cc 2024-06-12 04:55:55.000000000 -0400
@@ -42,11 +42,11 @@
load(in);
}
-int32_t Dictionary::find(const std::string& w) const {
+int32_t Dictionary::find(const std::string_view w) const {
return find(w, hash(w));
}
-int32_t Dictionary::find(const std::string& w, uint32_t h) const {
+int32_t Dictionary::find(const std::string_view w, uint32_t h) const {
int32_t word2intsize = word2int_.size();
int32_t id = h % word2intsize;
while (word2int_[id] != -1 && words_[word2int_[id]].word != w) {
@@ -126,12 +126,12 @@
return rand > pdiscard_[id];
}
-int32_t Dictionary::getId(const std::string& w, uint32_t h) const {
+int32_t Dictionary::getId(const std::string_view w, uint32_t h) const {
int32_t id = find(w, h);
return word2int_[id];
}
-int32_t Dictionary::getId(const std::string& w) const {
+int32_t Dictionary::getId(const std::string_view w) const {
int32_t h = find(w);
return word2int_[h];
}
@@ -142,7 +142,7 @@
return words_[id].type;
}
-entry_type Dictionary::getType(const std::string& w) const {
+entry_type Dictionary::getType(const std::string_view w) const {
return (w.find(args_->label) == 0) ? entry_type::label : entry_type::word;
}
@@ -160,7 +160,7 @@
// Since all fasttext models that were already released were trained
// using signed char, we fixed the hash function to make models
// compatible whatever compiler is used.
-uint32_t Dictionary::hash(const std::string& str) const {
+uint32_t Dictionary::hash(const std::string_view str) const {
uint32_t h = 2166136261;
for (size_t i = 0; i < str.size(); i++) {
h = h ^ uint32_t(int8_t(str[i]));
@@ -324,11 +324,16 @@
void Dictionary::addSubwords(
std::vector<int32_t>& line,
- const std::string& token,
+ const std::string_view token,
int32_t wid) const {
if (wid < 0) { // out of vocab
if (token != EOS) {
- computeSubwords(BOW + token + EOW, line);
+ std::string concat;
+ concat.reserve(BOW.size() + token.size() + EOW.size());
+ concat += BOW;
+ concat.append(token.data(), token.size());
+ concat += EOW;
+ computeSubwords(concat, line);
}
} else {
if (args_->maxn <= 0) { // in vocab w/o subwords
@@ -390,6 +395,51 @@
uint32_t h = hash(token);
int32_t wid = getId(token, h);
entry_type type = wid < 0 ? getType(token) : getType(wid);
+
+ ntokens++;
+ if (type == entry_type::word) {
+ addSubwords(words, token, wid);
+ word_hashes.push_back(h);
+ } else if (type == entry_type::label && wid >= 0) {
+ labels.push_back(wid - nwords_);
+ }
+ if (token == EOS) {
+ break;
+ }
+ }
+ addWordNgrams(words, word_hashes, args_->wordNgrams);
+ return ntokens;
+}
+
+namespace {
+bool readWordNoNewline(std::string_view& in, std::string_view& word) {
+ const std::string_view spaces(" \n\r\t\v\f\0");
+ std::string_view::size_type begin = in.find_first_not_of(spaces);
+ if (begin == std::string_view::npos) {
+ in.remove_prefix(in.size());
+ return false;
+ }
+ in.remove_prefix(begin);
+ word = in.substr(0, in.find_first_of(spaces));
+ in.remove_prefix(word.size());
+ return true;
+}
+} // namespace
+
+int32_t Dictionary::getStringNoNewline(
+ std::string_view in,
+ std::vector<int32_t>& words,
+ std::vector<int32_t>& labels) const {
+ std::vector<int32_t> word_hashes;
+ std::string_view token;
+ int32_t ntokens = 0;
+
+ words.clear();
+ labels.clear();
+ while (readWordNoNewline(in, token)) {
+ uint32_t h = hash(token);
+ int32_t wid = getId(token, h);
+ entry_type type = wid < 0 ? getType(token) : getType(wid);
ntokens++;
if (type == entry_type::word) {
diff -u -r fasttext-wheel-0.9.2/src/dictionary.h fasttext-0.9.3/src/dictionary.h
--- fasttext-wheel-0.9.2/src/dictionary.h 2023-05-05 07:58:14.000000000 -0400
+++ fasttext-0.9.3/src/dictionary.h 2024-06-12 04:55:55.000000000 -0400
@@ -13,6 +13,7 @@
#include <ostream>
#include <random>
#include <string>
+#include <string_view>
#include <unordered_map>
#include <vector>
@@ -36,13 +37,13 @@
static const int32_t MAX_VOCAB_SIZE = 30000000;
static const int32_t MAX_LINE_SIZE = 1024;
- int32_t find(const std::string&) const;
- int32_t find(const std::string&, uint32_t h) const;
+ int32_t find(const std::string_view) const;
+ int32_t find(const std::string_view, uint32_t h) const;
void initTableDiscard();
void initNgrams();
void reset(std::istream&) const;
void pushHash(std::vector<int32_t>&, int32_t) const;
- void addSubwords(std::vector<int32_t>&, const std::string&, int32_t) const;
+ void addSubwords(std::vector<int32_t>&, const std::string_view, int32_t) const;
std::shared_ptr<Args> args_;
std::vector<int32_t> word2int_;
@@ -71,10 +72,10 @@
int32_t nwords() const;
int32_t nlabels() const;
int64_t ntokens() const;
- int32_t getId(const std::string&) const;
- int32_t getId(const std::string&, uint32_t h) const;
+ int32_t getId(const std::string_view) const;
+ int32_t getId(const std::string_view, uint32_t h) const;
entry_type getType(int32_t) const;
- entry_type getType(const std::string&) const;
+ entry_type getType(const std::string_view) const;
bool discard(int32_t, real) const;
std::string getWord(int32_t) const;
const std::vector<int32_t>& getSubwords(int32_t) const;
@@ -87,7 +88,7 @@
const std::string&,
std::vector<int32_t>&,
std::vector<std::string>* substrings = nullptr) const;
- uint32_t hash(const std::string& str) const;
+ uint32_t hash(const std::string_view str) const;
void add(const std::string&);
bool readWord(std::istream&, std::string&) const;
void readFromFile(std::istream&);
@@ -99,6 +100,8 @@
const;
int32_t getLine(std::istream&, std::vector<int32_t>&, std::minstd_rand&)
const;
+ int32_t getStringNoNewline(std::string_view, std::vector<int32_t>&,
+ std::vector<int32_t>&) const;
void threshold(int64_t, int64_t);
void prune(std::vector<int32_t>&);
bool isPruned() {
diff -u -r fasttext-wheel-0.9.2/src/fasttext.cc fasttext-0.9.3/src/fasttext.cc
--- fasttext-wheel-0.9.2/src/fasttext.cc 2023-05-05 07:58:14.000000000 -0400
+++ fasttext-0.9.3/src/fasttext.cc 2024-06-12 04:55:55.000000000 -0400
@@ -532,7 +532,7 @@
if (ngrams[i] >= 0) {
vec.addRow(*input_, ngrams[i]);
}
- result.push_back(std::make_pair(substrings[i], std::move(vec)));
+ result.emplace_back(substrings[i], std::move(vec));
}
return result;
}
@@ -609,7 +609,7 @@
const std::string& wordA,
const std::string& wordB,
const std::string& wordC) {
- Vector query = Vector(args_->dim);
+ Vector query(args_->dim);
query.zero();
Vector buffer(args_->dim);
diff -u -r fasttext-wheel-0.9.2/src/matrix.h fasttext-0.9.3/src/matrix.h
--- fasttext-wheel-0.9.2/src/matrix.h 2023-05-05 07:58:14.000000000 -0400
+++ fasttext-0.9.3/src/matrix.h 2024-06-12 04:55:55.000000000 -0400
@@ -36,6 +36,7 @@
virtual void addVectorToRow(const Vector&, int64_t, real) = 0;
virtual void addRowToVector(Vector& x, int32_t i) const = 0;
virtual void addRowToVector(Vector& x, int32_t i, real a) const = 0;
+ virtual void averageRowsToVector(Vector& x, const std::vector<int32_t>& rows) const = 0;
virtual void save(std::ostream&) const = 0;
virtual void load(std::istream&) = 0;
virtual void dump(std::ostream&) const = 0;
diff -u -r fasttext-wheel-0.9.2/src/meter.cc fasttext-0.9.3/src/meter.cc
--- fasttext-wheel-0.9.2/src/meter.cc 2023-05-05 07:58:14.000000000 -0400
+++ fasttext-0.9.3/src/meter.cc 2024-06-12 04:55:55.000000000 -0400
@@ -39,9 +39,9 @@
labelMetrics_[prediction.second].scoreVsTrue.emplace_back(score, gold);
}
- if (falseNegativeLabels_) {
- for (const auto& label : labels) {
- labelMetrics_[label].gold++;
+ for (const auto& label : labels) {
+ labelMetrics_[label].gold++;
+ if (falseNegativeLabels_) {
if (!utils::containsSecond(predictions, label)) {
labelMetrics_[label].scoreVsTrue.emplace_back(falseNegativeScore, 1.0);
}
diff -u -r fasttext-wheel-0.9.2/src/model.cc fasttext-0.9.3/src/model.cc
--- fasttext-wheel-0.9.2/src/model.cc 2023-05-05 07:58:14.000000000 -0400
+++ fasttext-0.9.3/src/model.cc 2024-06-12 04:55:55.000000000 -0400
@@ -42,11 +42,7 @@
void Model::computeHidden(const std::vector<int32_t>& input, State& state)
const {
Vector& hidden = state.hidden;
- hidden.zero();
- for (auto it = input.cbegin(); it != input.cend(); ++it) {
- hidden.addRow(*wi_, *it);
- }
- hidden.mul(1.0 / input.size());
+ wi_->averageRowsToVector(hidden, input);
}
void Model::predict(
diff -u -r fasttext-wheel-0.9.2/src/quantmatrix.cc fasttext-0.9.3/src/quantmatrix.cc
--- fasttext-wheel-0.9.2/src/quantmatrix.cc 2023-05-05 07:58:14.000000000 -0400
+++ fasttext-0.9.3/src/quantmatrix.cc 2024-06-12 04:55:55.000000000 -0400
@@ -80,6 +80,14 @@
pq_->addcode(x, codes_.data(), i, norm);
}
+void QuantMatrix::averageRowsToVector(Vector& x, const std::vector<int32_t>& rows) const {
+ x.zero();
+ for (auto it = rows.cbegin(); it != rows.cend(); ++it) {
+ addRowToVector(x, *it);
+ }
+ x.mul(1.0 / rows.size());
+}
+
void QuantMatrix::save(std::ostream& out) const {
out.write((char*)&qnorm_, sizeof(qnorm_));
out.write((char*)&m_, sizeof(m_));
diff -u -r fasttext-wheel-0.9.2/src/quantmatrix.h fasttext-0.9.3/src/quantmatrix.h
--- fasttext-wheel-0.9.2/src/quantmatrix.h 2023-05-05 07:58:14.000000000 -0400
+++ fasttext-0.9.3/src/quantmatrix.h 2024-06-12 04:55:55.000000000 -0400
@@ -52,6 +52,7 @@
void addVectorToRow(const Vector&, int64_t, real) override;
void addRowToVector(Vector& x, int32_t i) const override;
void addRowToVector(Vector& x, int32_t i, real a) const override;
+ void averageRowsToVector(Vector& x, const std::vector<int32_t>& rows) const override;
void save(std::ostream&) const override;
void load(std::istream&) override;
void dump(std::ostream&) const override;
diff -u -r fasttext-wheel-0.9.2/src/vector.h fasttext-0.9.3/src/vector.h
--- fasttext-wheel-0.9.2/src/vector.h 2023-05-05 07:58:14.000000000 -0400
+++ fasttext-0.9.3/src/vector.h 2024-06-12 04:55:55.000000000 -0400
@@ -12,6 +12,7 @@
#include <ostream>
#include <vector>
+#include "aligned.h"
#include "real.h"
namespace fasttext {
@@ -20,12 +21,12 @@
class Vector {
protected:
- std::vector<real> data_;
+ intgemm::AlignedVector<real> data_;
public:
explicit Vector(int64_t);
Vector(const Vector&) = default;
- Vector(Vector&&) noexcept = default;
+ Vector(Vector&&) = default;
Vector& operator=(const Vector&) = default;
Vector& operator=(Vector&&) = default;
|
Collaborator
|
Hi, thanks for your interest in our project. Regarding this issue: as you may have noticed, installing For this PR, however, For example: # requirements.txt
...
fasttext; python_version >= "3.13"
fasttext-wheel; python_version < "3.13"
...This keeps the current default behavior for existing Python versions while enabling Python 3.13 support. |
ckane
force-pushed
the
update-fasttext
branch
from
5508c6c to
8305e8d
Compare
ckane
force-pushed
the
update-fasttext
branch
from
8305e8d to
94e6971
Compare
Contributor
Author
Ok thanks - I have re-pushed the change with the recommended fixes, they look like a great solution to me |
Collaborator
|
Hi, it should trigger a GitHub action CI here for automatic checking, but now I didn't see any signal about it. I am not sure if this is because of a |
Collaborator
|
OK i fixed the auto CI/CD issue by adding a commit, let's wait and see. If everything works well, I will merge this PR. Thanks! |
Collaborator
|
Test passed, looks good to me, thanks! |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The fastText project has been archived, and it looks like the latest version is now only in the
fasttextpackage in PyPI, notfasttext-wheel. This change switches to usingfasttext(0.9.3).The reason for this is because the wheel published as
fasttext-wheel==0.9.2doesn't build properly for me under Python 3.13, but the version hosted asfasttext==0.9.3does.