wgit functionalities and skeleton, move to subparsers and addition of…

… Repo Class (#1001)

* [feat] Adding wgit within fairscale/experimental/wgit.

* [feat] adding experimental wgit

* [feat] wgit init functionalities and skeleton for the rest.

* adapted the suggested changes

* repo class working

* [feat] wgit functionalities and skeleton. Addition of subparsers and repo class along with some changes.

* [feat] wgit functionalities and skeleton, move to subparsers and addition of Repo Class

* [feat] wgit functionalities and skeleton, move to subparsers and addition of Repo Class

* [docs] changed a comment in .gitignore

* [refactor] changed the sequene of tests in ci_test_list2.txt

.gitignore

@@ -12,6 +12,10 @@
 *.egg-info/
 .testmondata
 
+# experimental weigit
+experimental/.gitignore
+.wgit
+
 # Build and release
 build/
 dist/

experimental/__init__.py

@@ -2,3 +2,5 @@
 #
 # This source code is licensed under the BSD license found in the
 # LICENSE file in the root directory of this source tree.
+
+from .wgit import cli, weigit_api

experimental/wgit/__init__.py

@@ -4,5 +4,6 @@
 # LICENSE file in the root directory of this source tree.
 
 from .cli import main
+from .weigit_api import WeiGit
 
 __version__ = "0.0.1"

experimental/wgit/cli.py

@@ -5,18 +5,91 @@
 
 import argparse
 
+import experimental.wgit as wgit
+import experimental.wgit.weigit_api as weigit_api
+
 
 def main(argv=None):
-    desc = "WeiGit checkpoint tracking"
+    desc = "WeiGit: A git-like tool for model weight tracking"
+
+    # top level parser and corresponding subparser
     parser = argparse.ArgumentParser(description=desc)
+    subparsers = parser.add_subparsers(dest="command")
+
+    # Version
+    version_parser = subparsers.add_parser("version", description="Display version")
+    version_parser.set_defaults(command="version", subcommand="")
+
+    # Repo
+    init_parser = subparsers.add_parser("init", description="Initialize a weigit repo")
+    init_parser.add_argument("init", action="store_true", help="initialize the repo")
+
+    status_parser = subparsers.add_parser("status", description="Shows the repo's current status")
+    status_parser.add_argument("status", action="store_true", help="Show the repo's current status")
+
+    add_parser = subparsers.add_parser("add", description="add a file to the staged changeset (default: none)")
+    add_parser.add_argument(
+        "add",
+        default="",
+        type=str,
+        metavar="FILE_PATH",
+        help="add a file to the staged changeset (default: none)",
+    )
+
+    commit_parser = subparsers.add_parser("commit", description="Commits the staged changes")
+    commit_parser.add_argument("commit", action="store_true", help="Commit the staged changes")
+    commit_parser.add_argument(
+        "-m",
+        "--message",
+        default="",
+        type=str,
+        metavar="MESSAGE",
+        required=True,
+        help="commit message",
+    )
 
-    # flags
-    parser.add_argument("-i", "--init", action="store_true", help="Initialize a weigit repository!")
+    checkout_parser = subparsers.add_parser("checkout", description="checkout from a commit")
+    checkout_parser.add_argument(
+        "checkout",
+        default="",
+        type=str,
+        metavar="FILE_SHA1",
+        help="checkout from a commit",
+    )
+
+    log_parser = subparsers.add_parser("log", description="Show the history log of the repo or optionally of a file.")
+    log_parser.add_argument("log", action="store_true", help="Show the repo's history log")
+    log_parser.add_argument(
+        "-f",
+        "--file",
+        default="",
+        type=str,
+        metavar="FILE_PATH",
+        help="Show the history log of a file",
+    )
 
     args = parser.parse_args(argv)
 
-    if args.init:
-        print("Hello World, Wgit has been initialized!")
+    if args.command == "init":
+        weigit = weigit_api.WeiGit()
+
+    if args.command == "add":
+        weigit_api.WeiGit.add(args.add)
+
+    if args.command == "status":
+        weigit_api.WeiGit.status()
+
+    if args.command == "log":
+        weigit_api.WeiGit.log(args.file)
+
+    if args.command == "commit":
+        weigit_api.WeiGit.commit(args.message)
+
+    if args.command == "checkout":
+        weigit_api.WeiGit.checkout()
+
+    if args.command == "version":
+        print(wgit.__version__)
 
 
 if __name__ == "__main__":

experimental/wgit/utils.py

@@ -0,0 +1,8 @@
+from enum import Enum
+
+
+class ExitCode(Enum):
+    CLEAN = 0
+    FILE_EXISTS_ERROR = 1
+
+    ERROR = -1  # unknown errors

experimental/wgit/weigit_api.py

@@ -0,0 +1,158 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+#
+# This source code is licensed under the BSD license found in the
+# LICENSE file in the root directory of this source tree.
+
+
+import os
+import pathlib
+import sys
+
+import pygit2
+
+from experimental.wgit.utils import ExitCode
+
+
+class WeiGit:
+    def __init__(self) -> None:
+        """
+        Planned Features:
+            1. create the wgit directory. Error, if already dir exists.
+            2. SHA1Store.init()
+            3. Create SHA1 .wgit/sha1_ref_count.json
+            3. Initialize a .git directory within the .wgit using `git init`.
+            4. add a .gitignore within the .wgit directory, so that the git repo within will ignore `sha1_ref_count.json`
+        """
+
+        # Make .wgit directory. If already exists, we error out
+        try:
+            os.mkdir(".wgit")
+        except FileExistsError:
+            sys.stderr.write("An exception occured while wgit initialization: WeiGit already Initialized\n")
+            sys.exit(ExitCode.FILE_EXISTS_ERROR)
+
+        # if no .wgit dir then initialize the following
+        SHA1_store()
+
+        # create sha1_ref_count and a .gitignore:
+        # In general sha1_ref_count is only create only if .wgit already exists
+        try:
+            ref_count_json = ".wgit/sha1_ref_count.json"
+            with open(ref_count_json, "w") as f:
+                pass
+        except FileExistsError as error:
+            sys.stderr.write(f"An exception occured while creating {ref_count_json}: {repr(error)}\n")
+            sys.exit(ExitCode.FILE_EXISTS_ERROR)
+
+        # Make the .wgit a git repo
+        try:
+            pygit2.init_repository(".wgit/.git", False)
+        except BaseException as error:
+            sys.stderr.write(f"An exception occurred while initializing .wgit/.git: {repr(error)}\n")
+            sys.exit(ExitCode.ERROR)
+
+        # add a .gitignore:
+        try:
+            gitignore = ".wgit/.gitignore"
+            with open(gitignore, "w") as f:
+                f.write("sha1_ref_count.json")
+        except FileExistsError as error:
+            sys.stderr.write(f"An exception occured while creating {gitignore}: {repr(error)}\n")
+            sys.exit(ExitCode.FILE_EXISTS_ERROR)
+
+    @staticmethod
+    def add(file):
+        if Repo(os.getcwd()).exists():
+            print("wgit added")
+
+    @staticmethod
+    def status():
+        if Repo(os.getcwd()).exists():
+            print("wgit status")
+
+    @staticmethod
+    def log(file):
+        if Repo(os.getcwd()).exists():
+            if file:
+                print(f"wgit log of the file: {file}")
+            else:
+                print("wgit log")
+
+    @staticmethod
+    def commit(message):
+        if Repo(os.getcwd()).exists():
+            if message:
+                print(f"commited with message: {message}")
+            else:
+                print("wgit commit")
+
+    @staticmethod
+    def checkout():
+        if Repo(os.getcwd()).exists():
+            print("wgit checkout")
+
+    @staticmethod
+    def compression():
+        print("Not Implemented!")
+
+    @staticmethod
+    def checkout_by_steps():
+        print("Not Implemented!")
+
+
+class SHA1_store:
+    """
+    Planned Features:
+        1. def init
+        2. def add <file or data> -> SHA1
+        3. def remove (SHA1)
+        4. def add_ref(children_SHA1, parent_SHA1)
+        5. def read(SHA1): ->
+        6. def lookup(SHA1): -> file path to the data. NotFound Exception if not found.
+    """
+
+    def __init__(self) -> None:
+        pass
+
+
+class Repo:
+    """
+    Designates the weigit repo, which is identified by a path to the repo.
+    """
+
+    def __init__(self, check_dir) -> None:
+        self.repo_path = None
+        self.check_dir = os.path.realpath(check_dir)
+
+    def exists(self):
+        def weigit_repo_exists(check_dir):
+            """
+            checks if the input path to dir (check_dir) is a valid weigit repo
+            with .git and sha1_ref_count in the repo.
+            """
+            is_wgit_in_curr = pathlib.Path(os.path.join(check_dir, ".wgit")).exists()
+            is_refcount_in_wgit = pathlib.Path(os.path.join(check_dir, ".wgit/sha1_ref_count.json")).exists()
+            is_git_in_wgit = pathlib.Path(os.path.join(check_dir, ".wgit/.git")).exists()
+            return is_wgit_in_curr and is_refcount_in_wgit and is_git_in_wgit
+
+        if weigit_repo_exists(self.check_dir):
+            self.repo_path = os.path.join(self.check_dir, ".wgit")
+        else:
+            while self.check_dir != os.getcwd():
+                self.check_dir = os.path.dirname(self.check_dir)
+
+                if weigit_repo_exists(self.check_dir):
+                    self.repo_path = os.path.join(self.check_dir, ".wgit")
+                    break
+
+        if self.repo_path is None:
+            print("Initialize a weigit repo first!!")
+            is_exist = False
+        else:
+            is_exist = True
+        return is_exist
+
+    def get_repo_path(self):
+        if self.repo_path is None:
+            self.exists()
+        return self.repo_path

requirements-dev.txt

@@ -33,3 +33,6 @@ numpy == 1.21.5
 
 # For layerwise gradient scaler
 sklearn >= 0.0
+
+# For weigit
+pygit2==1.9.2

tests/ci_test_list_2.txt

@@ -1,3 +1,12 @@
+tests/experimental/nn/test_mevo.py
+tests/experimental/nn/test_multiprocess_pipe.py
+tests/experimental/nn/test_sync_batchnorm.py
+tests/experimental/nn/ampnet_pipe_process/test_ampnet_pipe.py
+tests/experimental/nn/test_offload.py
+tests/experimental/nn/test_auto_shard.py
+tests/experimental/optim/test_dynamic_loss_scaler.py
+tests/experimental/tooling/test_layer_memory_tracker.py
+tests/experimental/nn/test_ssd_offload.py
 tests/nn/data_parallel/test_fsdp_shared_weights_mevo.py
 tests/nn/data_parallel/test_fsdp_shared_weights.py
 tests/nn/data_parallel/test_fsdp_pre_backward_hook.py
@@ -41,13 +50,4 @@ tests/nn/pipe/test_dependency.py
 tests/nn/pipe/test_stream.py
 tests/nn/moe/test_moe_layer.py
 tests/nn/moe/test_top2gating.py
-tests/experimental/nn/test_mevo.py
-tests/experimental/nn/test_multiprocess_pipe.py
-tests/experimental/nn/test_sync_batchnorm.py
-tests/experimental/nn/ampnet_pipe_process/test_ampnet_pipe.py
-tests/experimental/nn/test_offload.py
-tests/experimental/nn/test_auto_shard.py
-tests/experimental/optim/test_dynamic_loss_scaler.py
-tests/experimental/tooling/test_layer_memory_tracker.py
-tests/experimental/nn/test_ssd_offload.py
 tests/nn/data_parallel/test_fsdp_offload.py