8 Releases published by 1 person

• v20250422-162548

  published Apr 22, 2025

• v20250423-171917

  published Apr 23, 2025

• v20250425-175031

  published Apr 25, 2025

• v20250428-175124

  published Apr 28, 2025

• v20250501-150719

  published May 1, 2025

• v20250509-185555

  published May 9, 2025

• v20250510-161323

  published May 10, 2025

• v20250513-130103

  published May 13, 2025

73 Pull requests merged by 18 people

• pr label event table in dynamo

  #6644 merged May 20, 2025

• [Cherry-Pick] ci: Pin more nvidia-container dependencies #6637

  #6645 merged May 20, 2025

• Fix s3 update workflow after: #6639

  #6643 merged May 20, 2025

• Keep nightly dev20250310 for ExecuTorch

  #6639 merged May 20, 2025

• Check if nvidia-container-toolkit is installed correctly

  #6638 merged May 16, 2025

• ci: Pin more nvidia-container dependencies

  #6637 merged May 16, 2025

• demo

  #6611 merged May 15, 2025

• Link to hud disabled view in GH issue

  #6636 merged May 15, 2025

• [ez][CH][drci] Improve recent_pr_workflows_query perf

  #6634 merged May 15, 2025

• [ez][HUD] Make main page column width wider

  #6631 merged May 15, 2025

• Add h100 workflow dispatch to HUD

  #6635 merged May 15, 2025

• [BE] Upgrade XPU support package to 2025.1 in Windows CICD

  #6553 merged May 14, 2025

• [Mergebot] Adding ciflow/pull in PR without pull and lint workflows

  #6610 merged May 14, 2025

• Added AMD CPU CI instances in scale-config.yml

  #6629 merged May 14, 2025

• Fix typo after #6612

  #6632 merged May 14, 2025

• Removing ephemeral experiment from scale-config.yml and updating validate_scale_config.py

  #6605 merged May 13, 2025

• Fix scaleUpChron check for queue time using max_queue_time_minutes

  #6618 merged May 13, 2025

• [Uitlization] Aggreggation Report Api

  #6609 merged May 12, 2025

• Make every runner in scale-config.yml to be ephemeral

  #6628 merged May 12, 2025

• [Cherry-Pick] Fix python 3.13t builds on aarch64 #6624

  #6627 merged May 12, 2025

• Fix python 3.13t builds on aarch64

  #6624 merged May 12, 2025

• [XPU] Update runtime pypi package dependencies for 2025.1

  #6612 merged May 12, 2025

• Trigger scale-up-cron lambda more frequently

  #6621 merged May 10, 2025

• Update pr_time_benchmarks log classification rule

  #6615 merged May 10, 2025

• [release-only] Update pinned docker images for vision and audio

  #6622 merged May 9, 2025

• [BE] Improve autoscaler logging about how many instances are being provisioned

  #6619 merged May 9, 2025

• Revert "Temporarily shrink the interval we wait for during chaos testing"

  #6617 merged May 9, 2025

• Temporarily shrink the interval we wait for during chaos testing

  #6616 merged May 9, 2025

• [release] Bump Candidate to 2.7.1, update stable and nightly as well

  #6614 merged May 9, 2025

• Release 2.7.1 bump candidate version

  #6613 merged May 9, 2025

• [bug] logic handle device_ID

  #6600 merged May 8, 2025

• Revert "[Mergebot] Adding ciflow/pull in PR without pull and lint workflows"

  #6606 merged May 8, 2025

• [Mergebot] Adding ciflow/pull in PR without pull and lint workflows

  #6604 merged May 8, 2025

• Pin to SHA for actions outside of PyTorch

  #6591 merged May 7, 2025

• add workflow util summary

  #6602 merged May 6, 2025

• Improvements to test times queries

  #6597 merged May 2, 2025

• [ez][HUD] Expand log search to other repos

  #6599 merged May 2, 2025

• Send metrics before callback() for scaleUpChron

  #6596 merged May 1, 2025

• [tritonbench] Exclude low_mem_dropout from the data to avoid zeros.

  #6594 merged May 1, 2025

• Separate private and public devices on benchmark dashboard

  #6595 merged May 1, 2025

• [Device ID] visualize benchmark graph with group_key

  #6593 merged May 1, 2025

• Upgrade nightly wheels to ROCm6.4

  #6592 merged Apr 30, 2025

• Maybe better syntax for calculate docker image with custom tag prefix

  #6586 merged Apr 30, 2025

• update quickstart UI to remove conda

  #6585 merged Apr 29, 2025

• Delete TorchBench from NavBar

  #6587 merged Apr 29, 2025

• [hud][ch][drci] add api to cache ch queries + cache issues query

  #6578 merged Apr 28, 2025

• [ch] optimize master_commit_red_percent query

  #6580 merged Apr 28, 2025

• Fix aarch64 cpu/cuda include logic

  #6584 merged Apr 28, 2025

• Add deep dive videos for the autoscaler

  #6488 merged Apr 28, 2025

• Add device type (public, private) and id

  #6579 merged Apr 28, 2025

• job cancellation dashboard

  #6577 merged Apr 25, 2025

• [hud][ch] fix query_execution_metrics page

  #6571 merged Apr 25, 2025

• Fix scaleupchron metrics

  #6572 merged Apr 25, 2025

• [S3] Fix up networkx index for UV

  #6575 merged Apr 25, 2025

• [HUD] Hack for rerun disable tests and mem leak regex

  #6566 merged Apr 25, 2025

• [Release] fbgemm_gpu and fbgemm_gpu_genai. Cleanup release sripts

  #6570 merged Apr 25, 2025

• [Queue Time Analysis] Search bar improvement

  #6563 merged Apr 24, 2025

• [tritonbench] Add benchmark picker

  #6560 merged Apr 24, 2025

• [release] Update promotion scripts to download.pytorch.org

  #6565 merged Apr 24, 2025

• [validations] Fix stable cuda validation for pypi release. Cleanup unused code.

  #6558 merged Apr 24, 2025

• Wdvr/darkmode fix queuetime dark mode colors

  #6562 merged Apr 23, 2025

• [Queue Time Analysis] Dynamically resize echarts

  #6561 merged Apr 23, 2025

• Single View Table Improvement for benchmark

  #6556 merged Apr 23, 2025

• Queue time histogram [First Iteration]

  #6531 merged Apr 23, 2025

• [Autoscaler] Add logging to better understand why a scale down occurred

  #6557 merged Apr 23, 2025

• [CH] optimize master_commit_red_jobs by eliminating join and adding pre-filtering

  #6552 merged Apr 22, 2025

• [Release 2.7] Update Binary Build Matrix

  #6554 merged Apr 22, 2025

• [CH] optimize hud_query by eliminating joins

  #6550 merged Apr 22, 2025

• [Cherry-PIck] Repair Manylinux 2.28 wheels #6549

  #6551 merged Apr 22, 2025

• Repair Manylinux 2.28 wheels

  #6549 merged Apr 22, 2025

• Fix log in lambda

  #6547 merged Apr 21, 2025

• Fix tutorials stats workflow to checkout the correct repo

  #6548 merged Apr 21, 2025

• [Release 2.7] Update pypi staging scripts

  #6546 merged Apr 21, 2025

11 Pull requests opened by 8 people

• Double-check runner status when it is free with different API using GHA API runner id

  #6564 opened Apr 24, 2025

• fix loading page's default behaviour

  #6576 opened Apr 25, 2025

• Add source code indexing

  #6590 opened Apr 30, 2025

• Slightly better typing and doc comment for query clickhouse

  #6607 opened May 8, 2025

• Update to 12.8.1 for windows AMI

  #6620 opened May 9, 2025

• [DRAFT] Test use uv for building wheels

  #6626 opened May 12, 2025

• Introduce uv based wheel build

  #6630 opened May 12, 2025

• Bump undici from 5.28.4 to 5.29.0 in /setup-ssh

  #6633 opened May 15, 2025

• Bump aws-actions/configure-aws-credentials from 1.7.0 to 4.2.1

  #6640 opened May 19, 2025

• Bump setuptools from 70.0.0 to 78.1.1 in /tools/torchci

  #6641 opened May 19, 2025

• Add Amazon EC2 M8g Instances

  #6642 opened May 20, 2025

13 Issues closed by 7 people

• [Pytorch] There are 1 Recurrently Failing Jobs on pytorch/pytorch nightly

  #3867 closed May 18, 2025

• Cost dashboard has gaps in April for LF

  #6569 closed May 12, 2025

• Http status 403 on dr. CI HUD API fetch in trymerge

  #6598 closed May 2, 2025

• [feature]:

  #6589 closed May 1, 2025

• [feature]:

  #6588 closed May 1, 2025

• Better max autotune discovery on pt2 dashboard

  #6175 closed May 1, 2025

• Compilers HUD should show all graphs on page

  #4031 closed May 1, 2025

• Record failed benchmark runs in the database

  #6294 closed May 1, 2025

• Torchbench benchmark page crashes when selecting hashes

  #6567 closed Apr 29, 2025

• [dr ci] broken trunk not working?

  #6414 closed Apr 29, 2025

• Flaky-bot claims test is not flaky and it starts failing immediately afterwards

  #6533 closed Apr 29, 2025

• Allow using tag prefixes in calculate-docker-image

  #6048 closed Apr 23, 2025

• Change conda pruning mechanism to keep last 2 versions of packages instead of 1

  #3916 closed Apr 22, 2025

9 Issues opened by 8 people

• Better ClickHouse query testing: Mock tests with real data

  #6608 opened May 8, 2025

• Add extensions `flash_attention` and `vllm` as test of new PyTorch releases for known issues of compat of their binaries and of possibility of compiling these from source

  #6603 opened May 6, 2025

• Bots should tell developers when to rebase their PRs

  #6601 opened May 5, 2025

• Automate % of runners on Linux Foundation experiment

  #6583 opened Apr 28, 2025

• [release] UV validations

  #6582 opened Apr 28, 2025

• [release] Poetry validations

  #6581 opened Apr 28, 2025

• Job names too long they are getting cut off by github

  #6574 opened Apr 25, 2025

• lots of perf dashboards are just empty

  #6568 opened Apr 24, 2025

• [ExecuTorch] UI single view issue

  #6555 opened Apr 22, 2025

11 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• Enable windows arm64 builds for audio and vision

  #6352 commented on May 20, 2025 • 7 new comments

• [Pytorch] There are 37 Recurrently Failing Jobs on pytorch/builder main

  #4226 commented on May 1, 2025 • 0 new comments

• Wrong datapoints on PT2 dashboard for retried workflows

  #6173 commented on May 1, 2025 • 0 new comments

• Support for h100 CI

  #6535 commented on May 7, 2025 • 0 new comments

• Inductor dashboard features tracking

  #4410 commented on May 7, 2025 • 0 new comments

• DO NOT CLOSE IT / GitHub Workflow Runner Determinator

  #5132 commented on May 14, 2025 • 0 new comments

• [Pytorch] There are 2 machines with long queues

  #4232 commented on May 20, 2025 • 0 new comments

• [Pytorch] There are 4 Recurrently Failing Jobs on pytorch/pytorch main

  #4194 commented on May 20, 2025 • 0 new comments

• Fix wrong benchmark data associated with commits

  #6366 commented on May 19, 2025 • 0 new comments

• Draft - Test Workflow arm64 logic

  #6369 commented on May 2, 2025 • 0 new comments

• Masquerade ephemeral runners as non-ephemeral ones

  #6545 commented on Apr 21, 2025 • 0 new comments