New repo layout based on RFC

.gitignore

@@ -141,7 +141,13 @@ __pypackages__/
 .scrapy
 
 # Sphinx documentation
-docs/_build/
+docs/build
+# sphinx-gallery
+docs/source/auto_examples/
+docs/source/gen_modules/
+docs/source/generated/
+# pytorch-sphinx-theme gets installed here
+docs/src
 
 # Spyder project settings
 .spyderproject

README.md

@@ -28,7 +28,7 @@ from torchtune.llm.llama2.transformer import TransformerDecoder
 ```
 
 ### Coding Style
-TorchTBD uses pre-commit hooks to ensure style consistency and prevent common mistakes. Enable it by:
+`torchtune` uses pre-commit hooks to ensure style consistency and prevent common mistakes. Enable it by:
 
 ```
 pre-commit install
@@ -41,3 +41,20 @@ You can also run this manually on every file using:
 ```
 pre-commit run --all-files
 ```
+
+### Build docs
+
+From the `docs` folder:
+
+Install dependencies:
+
+```
+pip install -r requirements.txt
+```
+
+Then:
+
+```
+make html
+# Now open build/html/index.html
+```

docs/Makefile

@@ -0,0 +1,43 @@
+# Minimal makefile for Sphinx documentation
+#
+
+ifneq ($(EXAMPLES_PATTERN),)
+    EXAMPLES_PATTERN_OPTS := -D sphinx_gallery_conf.filename_pattern="$(EXAMPLES_PATTERN)"
+endif
+
+# You can set these variables from the command line.
+SPHINXOPTS    = -W -j auto $(EXAMPLES_PATTERN_OPTS)
+SPHINXBUILD   = sphinx-build
+SPHINXPROJ    = torchvision
+SOURCEDIR     = source
+BUILDDIR      = build
+
+# Put it first so that "make" without argument is like "make help".
+help:
+	@$(SPHINXBUILD) -M help "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)
+
+docset: html
+	doc2dash --name $(SPHINXPROJ) --icon $(SOURCEDIR)/_static/img/pytorch-logo-flame.png --enable-js --online-redirect-url http://pytorch.org/vision/ --force $(BUILDDIR)/html/
+
+	# Manually fix because Zeal doesn't deal well with `icon.png`-only at 2x resolution.
+	cp $(SPHINXPROJ).docset/icon.png $(SPHINXPROJ).docset/icon@2x.png
+	convert $(SPHINXPROJ).docset/icon@2x.png -resize 16x16 $(SPHINXPROJ).docset/icon.png
+
+html-noplot:  # Avoids running the gallery examples, which may take time
+	$(SPHINXBUILD) -D plot_gallery=0 -b html "${SOURCEDIR}" "$(BUILDDIR)"/html
+	@echo
+	@echo "Build finished. The HTML pages are in $(BUILDDIR)/html."
+
+clean:
+	rm -rf $(BUILDDIR)/*
+	rm -rf $(SOURCEDIR)/auto_examples/  # sphinx-gallery
+	rm -rf $(SOURCEDIR)/gen_modules/  # sphinx-gallery
+	rm -rf $(SOURCEDIR)/generated/  # autosummary
+	rm -rf $(SOURCEDIR)/models/generated  # autosummary
+
+.PHONY: help Makefile docset
+
+# Catch-all target: route all unknown targets to Sphinx using the new
+# "make mode" option.  $(O) is meant as a shortcut for $(SPHINXOPTS).
+%: Makefile
+	@$(SPHINXBUILD) -M $@ "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)

docs/license_header.txt

@@ -0,0 +1,5 @@
+Copyright (c) Meta Platforms, Inc. and affiliates.
+All rights reserved.
+
+This source code is licensed under the BSD-style license found in the
+LICENSE file in the root directory of this source tree.

docs/requirements.txt

@@ -0,0 +1,4 @@
+sphinx-gallery>=0.11.1
+sphinx==5.0.0
+sphinx_copybutton
+-e git+https://github.com/pytorch/pytorch_sphinx_theme.git#egg=pytorch_sphinx_theme

docs/source/conf.py

@@ -0,0 +1,190 @@
+#!/usr/bin/env python3
+#
+# PyTorch documentation build configuration file, created by
+# sphinx-quickstart on Fri Dec 23 13:31:47 2016.
+#
+# This file is execfile()d with the current directory set to its
+# containing dir.
+#
+# Note that not all possible configuration values are present in this
+# autogenerated file.
+#
+# All configuration values have a default; values that are commented out
+# serve to show the default.
+
+# If extensions (or modules to document with autodoc) are in another directory,
+# add these directories to sys.path here. If the directory is relative to the
+# documentation root, use os.path.abspath to make it absolute, like shown here.
+#
+# import os
+# import sys
+# sys.path.insert(0, os.path.abspath('.'))
+
+import os
+import sys
+import textwrap
+from copy import copy
+from pathlib import Path
+
+import pytorch_sphinx_theme
+
+sys.path.append(os.path.abspath("."))
+
+# -- General configuration ------------------------------------------------
+
+# Required version of sphinx is set from docs/requirements.txt
+
+# Add any Sphinx extension module names here, as strings. They can be
+# extensions coming with Sphinx (named 'sphinx.ext.*') or your custom
+# ones.
+extensions = [
+    "sphinx.ext.autodoc",
+    "sphinx.ext.autosummary",
+    "sphinx.ext.doctest",
+    "sphinx.ext.intersphinx",
+    "sphinx.ext.todo",
+    "sphinx.ext.mathjax",
+    "sphinx.ext.napoleon",
+    "sphinx.ext.viewcode",
+    "sphinx.ext.duration",
+    "sphinx_copybutton",
+]
+
+napoleon_use_ivar = True
+napoleon_numpy_docstring = False
+napoleon_google_docstring = True
+
+
+# Add any paths that contain templates here, relative to this directory.
+templates_path = ["_templates"]
+
+# The suffix(es) of source filenames.
+# You can specify multiple suffix as a list of string:
+#
+source_suffix = {
+    ".rst": "restructuredtext",
+}
+
+# The master toctree document.
+master_doc = "index"
+
+# General information about the project.
+project = "TorchTune"
+copyright = "2023-present, TorchTune Contributors"
+author = "Torch Contributors"
+
+# The language for content autogenerated by Sphinx. Refer to documentation
+# for a list of supported languages.
+#
+# This is also used if you do content translation via gettext catalogs.
+# Usually you set "language" from the command line for these cases.
+language = "en"
+
+# List of patterns, relative to source directory, that match files and
+# directories to ignore when looking for source files.
+# This patterns also effect to html_static_path and html_extra_path
+exclude_patterns = []
+
+# The name of the Pygments (syntax highlighting) style to use.
+pygments_style = "sphinx"
+
+# If true, `todo` and `todoList` produce output, else they produce nothing.
+todo_include_todos = True
+
+
+# -- Options for HTML output ----------------------------------------------
+
+# The theme to use for HTML and HTML Help pages.  See the documentation for
+# a list of builtin themes.
+#
+html_theme = "pytorch_sphinx_theme"
+html_theme_path = [pytorch_sphinx_theme.get_html_theme_path()]
+
+# Theme options are theme-specific and customize the look and feel of a theme
+# further.  For a list of options available for each theme, see the
+# documentation.
+#
+html_theme_options = {
+    "collapse_navigation": False,
+    "display_version": True,
+    "logo_only": True,
+    "pytorch_project": "docs",
+    "navigation_with_keys": True,
+    "analytics_id": "GTM-T8XT4PS",
+}
+
+html_logo = "_static/img/pytorch-logo-dark.svg"
+
+# Add any paths that contain custom static files (such as style sheets) here,
+# relative to this directory. They are copied after the builtin static files,
+# so a file named "default.css" will overwrite the builtin "default.css".
+html_static_path = ["_static"]
+
+# -- Options for HTMLHelp output ------------------------------------------
+
+# Output file base name for HTML help builder.
+htmlhelp_basename = "PyTorchdoc"
+
+
+autosummary_generate = True
+
+# Example configuration for intersphinx: refer to the Python standard library.
+intersphinx_mapping = {
+    "python": ("https://docs.python.org/3/", None),
+    "torch": ("https://pytorch.org/docs/stable/", None),
+    "numpy": ("https://numpy.org/doc/stable/", None),
+    "PIL": ("https://pillow.readthedocs.io/en/stable/", None),
+    "matplotlib": ("https://matplotlib.org/stable/", None),
+}
+
+# -- A patch that prevents Sphinx from cross-referencing ivar tags -------
+# See http://stackoverflow.com/a/41184353/3343043
+
+from docutils import nodes
+from sphinx import addnodes
+from sphinx.util.docfields import TypedField
+
+
+def patched_make_field(self, types, domain, items, **kw):
+    # `kw` catches `env=None` needed for newer sphinx while maintaining
+    #  backwards compatibility when passed along further down!
+
+    # type: (list, unicode, tuple) -> nodes.field  # noqa: F821
+    def handle_item(fieldarg, content):
+        par = nodes.paragraph()
+        par += addnodes.literal_strong("", fieldarg)  # Patch: this line added
+        # par.extend(self.make_xrefs(self.rolename, domain, fieldarg,
+        #                           addnodes.literal_strong))
+        if fieldarg in types:
+            par += nodes.Text(" (")
+            # NOTE: using .pop() here to prevent a single type node to be
+            # inserted twice into the doctree, which leads to
+            # inconsistencies later when references are resolved
+            fieldtype = types.pop(fieldarg)
+            if len(fieldtype) == 1 and isinstance(fieldtype[0], nodes.Text):
+                typename = "".join(n.astext() for n in fieldtype)
+                typename = typename.replace("int", "python:int")
+                typename = typename.replace("long", "python:long")
+                typename = typename.replace("float", "python:float")
+                typename = typename.replace("type", "python:type")
+                par.extend(self.make_xrefs(self.typerolename, domain, typename, addnodes.literal_emphasis, **kw))
+            else:
+                par += fieldtype
+            par += nodes.Text(")")
+        par += nodes.Text(" -- ")
+        par += content
+        return par
+
+    fieldname = nodes.field_name("", self.label)
+    if len(items) == 1 and self.can_collapse:
+        fieldarg, content = items[0]
+        bodynode = handle_item(fieldarg, content)
+    else:
+        bodynode = self.list_type()
+        for fieldarg, content in items:
+            bodynode += nodes.list_item("", handle_item(fieldarg, content))
+    fieldbody = nodes.field_body("", bodynode)
+    return nodes.field("", fieldname, fieldbody)
+
+
+TypedField.make_field = patched_make_field

docs/source/index.rst

@@ -0,0 +1,4 @@
+TorchTune
+=========
+
+Hello.

torchtune/llm/scripts/recipes/finetune_alpaca.py → recipes/finetune_llm.py

@@ -8,9 +8,10 @@
 from torch.optim.optimizer import Optimizer
 from torch.utils.data import DataLoader
 
-from torchtune.llm.datasets import get_dataset
-from torchtune.llm.llama2.tokenizer import Tokenizer
-from torchtune.llm.llama2.transformer import TransformerDecoder
+from torchtune.datasets import get_dataset
+from torchtune.models.llama2.tokenizer import Tokenizer
+from torchtune.models.llama2.transformer import TransformerDecoder
+
 from tqdm import tqdm
 
 

torchtune/llm/requirements.txt → requirements.txt

@@ -1,2 +1,3 @@
 sentencepiece==0.1.99
 tqdm==4.66.1
+datasets==2.15.0

torchtune/llm/scripts/checkpoint/convert_llama2_to_native.py → scripts/llama2_checkpoint/convert_llama2_to_native.py

@@ -14,10 +14,10 @@
 from typing import Any, Dict, Optional
 
 import torch
-from tests.llm.llama2.scripts.compare_decoder import Transformer
+from tests.blocks.llama2.scripts.compare_decoder import Transformer
 
 from torch import Tensor
-from torchtune.llm.llama2.transformer import TransformerDecoder
+from torchtune.models.llama2.transformer import TransformerDecoder
 
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)

torchtune/llm/scripts/examples/README.md → scripts/llama2_inference/README.md

@@ -1,4 +1,4 @@
-### torchtune.llm Examples 
+### torchtune Examples 
 
 This directory contains examples and documentation for getting started with using torchtune's LLM utilities.
 
@@ -8,4 +8,4 @@ This directory contains examples and documentation for getting started with usin
 
 1. Install the additional requirements within your conda env via `pip install requirements.txt` (run this command in this directory)
 2. Obtain a native Llama-2 7b checkpoint stored on your local machine. This can for example be produced via our conversion utilities, see the [README](https://github.com/pytorch-labs/torchtune/blob/main/torchtune/llm/scripts/checkpoint/README.md) for more details.
-3. Run the inference example within your torchtune install:  `python -m torchtune.llm.scripts.examples.inference --native-checkpoint-path <your checkpoint path> --tokenizer-path <your tokenizer path>`
+3. Run the inference example within your torchtune install:  `python -m examples.inference --native-checkpoint-path <your checkpoint path> --tokenizer-path <your tokenizer path>`