Merge pull request #2 from AlignmentResearch/chore/pull-in-upstream

Chore/pull in upstream

Author: Adrià Garriga-Alonso

.github/workflows/nightly-release.yml

@@ -40,7 +40,7 @@ jobs:
       - name: setup python
         uses: actions/setup-python@v2
         with:
-          python-version: '3.7'
+          python-version: '3.8'
           architecture: x64
 
       - name: install deps

.github/workflows/pull-request.yml

@@ -24,14 +24,14 @@ on:
       - reopened
       - edited
 jobs:
-  validate-naming-convention:
-    name: Pull Request's title matches naming convention
-    runs-on: ubuntu-latest
-    steps:
-      - uses: deepakputhraya/action-pr-title@master
-        with:
-          regex: '^\[(?:feat|fix|doc|refactor|deprecation)\]\s[A-Z].*(?<!\.)$'
-          github_token: ${{ github.token }}
+#  validate-naming-convention:
+#    name: Pull Request's title matches naming convention
+#    runs-on: ubuntu-latest
+#    steps:
+#      - uses: deepakputhraya/action-pr-title@master
+#        with:
+#          regex: '^\[(?:feat|fix|doc|refactor|deprecation)\]\s[A-Z].*(?<!\.)$'
+#          github_token: ${{ github.token }}
   run-checks:
     if: github.event.pull_request.draft == false && github.event.action != 'edited'
     runs-on: ubuntu-latest
@@ -68,8 +68,8 @@ jobs:
 
   storage-performance-checks:
     needs: run-checks
-    concurrency: perf-tests
-    runs-on: [self-hosted, performance-tests]
+    concurrency: storage-performance-checks
+    runs-on: [self-hosted, perf-tests]
     name: Performance tests
     steps:
       - name: checkout

.github/workflows/python-package.yml

@@ -96,6 +96,7 @@ jobs:
           python -m pip install -r requirements.txt
 
       - name: Build bdist wheels for 'cp37-cp37m'
+        if: matrix.manylinux-version == 'manylinux_2_24_x86_64'
         uses: nick-fields/retry@v2
         with:
           max_attempts: 3

CHANGELOG.md

@@ -1,5 +1,44 @@
 # Changelog
 
+## Unreleased:
+
+### Fixes: 
+- Fix issues with tag false reassignment (mihran113)
+
+## 3.29.1 May 8, 2025:
+
+### Enhancements:
+- Constant indexing of in-progress runs (alberttorosyan)
+- Fallback to union view if index db is missing (alberttorosyan, mihran113)
+
+
+### Fixes:
+- Fix min/max calculation for single point metrics (mihran113)
+- Aim web ui integration in jupyter/colab (larissapoghosyan)
+- Package publishing for Linux/Python 3.7 (alberttorosyan)
+
+## 3.29.0 May 8, 2025 (Yanked)
+
+## 3.28.0 Mar 21, 2025
+
+### Enhancements:
+- Skip metrics check when run is known to yield false result (alberttorosyan)
+- Remove metric version check to improve performance of metric retrieval (mihran113)
+- Move indexing thread to main process of `aim up` (alberttorosyan)
+- Add AimCallback implementation for hugging face distributed runs (VassilisVassiliadis)
+- Add py.typed marker to allow usage of existing type annotations (bluenote10)
+
+
+### Fixes:
+- Decrease client resources keep-alive time (mihran113)
+- Fix connection of data points on epoch alignment (mihran113)
+- Resolve issue with adding duplicate tags to the same run (mihran113)
+- Improve error messages for remote tracking server (mihran113)
+- Fix spurious assertion error in message stream parsing (qzed)
+- Correct indentation on query proxy object return statement (alberttorosyan)
+- Fix typing issues in S3ArtifactStorage implementation (sbatchelder)
+
+
 ## 3.27.0 Dec 18, 2024
 
 ### Enhancements:

aim/VERSION

@@ -1 +1 @@
-3.27.0
+3.29.1

aim/acme.py

@@ -1,2 +1,3 @@
 # Alias to SDK acme interface
-from aim.sdk.adapters.acme import AimCallback, AimWriter  # noqa F401
+from aim.sdk.adapters.acme import AimCallback as AimCallback
+from aim.sdk.adapters.acme import AimWriter as AimWriter

aim/cli/convert/commands.py

@@ -40,7 +40,7 @@ def convert_tensorboard(ctx, logdir, flat, no_cache):
 @click.option('--flat', '-f', required=False, is_flag=True, default=False)
 def convert_tensorflow(ctx, logdir, flat):
     click.secho(
-        "WARN: Command 'tf' is deprecated and will be removed in future releases," " please use 'tensorboard' instead.",
+        "WARN: Command 'tf' is deprecated and will be removed in future releases, please use 'tensorboard' instead.",
         fg='red',
     )
     repo_inst = ctx.obj['repo_inst']

aim/cli/init/commands.py

@@ -20,15 +20,15 @@ def init(repo, yes, skip_if_exists):
     re_init = False
     if Repo.exists(repo_path):
         if yes and skip_if_exists:
-            raise click.BadParameter('Conflicting init options.' 'Either specify -y/--yes or -s/--skip-if-exists')
+            raise click.BadParameter('Conflicting init options.Either specify -y/--yes or -s/--skip-if-exists')
         elif yes:
             re_init = True
         elif skip_if_exists:
             click.echo('Repo exists at {}. Skipped initialization.'.format(repo_path))
             return
         else:
             re_init = click.confirm(
-                'Aim repository is already initialized. ' 'Do you want to re-initialize to empty Aim repository?'
+                'Aim repository is already initialized. Do you want to re-initialize to empty Aim repository?'
             )
         if not re_init:
             return

aim/cli/manager/manager.py

@@ -33,18 +33,18 @@ def check_startup_success():
         import requests
 
         server_path = 'http://{}:{}{}'.format(args['--host'], args['--port'], args['--base-path'])
-        status_api = f'{server_path}/api/projects/status'
-        retry_count = 5
-        sleep_interval = 1
+        status_api = f'{server_path}/api/projects/'
+        retry_count = 10
+        sleep_interval = 0.1
         for _ in range(retry_count):
+            time.sleep(sleep_interval)
+            sleep_interval *= 2
             try:
                 response = requests.get(status_api)
                 if response.status_code == 200:
                     return True
             except Exception:
                 pass
-            sleep_interval += 1
-            time.sleep(sleep_interval)
 
         return False
 

aim/cli/runs/commands.py

@@ -195,7 +195,7 @@ def update_metrics(ctx, yes):
     index_manager = RepoIndexManager.get_index_manager(repo)
     hashes = repo.list_all_runs()
     for run_hash in tqdm.tqdm(hashes, desc='Updating runs', total=len(hashes)):
-        meta_tree = repo.request_tree('meta', run_hash, read_only=False, from_union=False)
+        meta_tree = repo.request_tree('meta', run_hash, read_only=False)
         meta_run_tree = meta_tree.subtree(('meta', 'chunks', run_hash))
         try:
             # check if the Run has already been updated.