v0.72.0

What's new

• Fixed blocking error when reloading operation details page
• Adjust undesired scrolling and updated help text in Operations Graph
• Fix disconnected nodes in Device Operations graph
• Fix non unique key in device operations rendering
• Added more clear error message if the port is unavailable
• Added a new SSH identity file field to remote connection dialogs
• Updated and expanded remote sync documentation
• Source file path now displayed when viewing file contents
• Added navigation to source file contents to scroll to top, bottom and the highlighted line
• Errors when viewing source file contents are now output into the UI
• In Operation Details, tensors with tooltips are now highlighted with a styled underline when their attributes are available
• Recommended TTNN config is now detailed and explained in the documentation

What's Changed

• Fix/Instance is requested too frequently by @dcblundell in #1173
• Feature/Add file path to stack trace source by @dcblundell in #1205
• Show more user friendly error message when socket cannot be bound by @smountenay-tt in #1209
• Fix edge case error when reloading the page by @aidemsined in #1212
• Feature/Add scroll to source line by @dcblundell in #1210
• Use unique keys for op input/output lists by @aidemsined in #1216
• Feature/Indicate tensor tooltip exists by @dcblundell in #1179
• Feature/Improve stack trace source error messaging by @dcblundell in #1207
• Improve SSH key handling by @smountenay-tt in #1218
• Adjust operation graph UI text and layout by @aidemsined in #1220
• Refactor device ops graph construction & styling by @aidemsined in #1217
• Docs/Add descriptions for environment variables by @dcblundell in #1208

Full Changelog: v0.71.0...v0.72.0

v0.71.0

What's New

• Device Operations Graph rewritten to enhance readibility
• Updated tt-perf-report dependency and added support for new data fields
• Replaced deprecated StackByIn0 grouping in the performance table with more flexible GroupBy options
• Rewrote regexes used in finding the source file from operation stack trace to fix compatibility with certain systems
• Improved default report path strategy to be more agnostic and added new environment varibles to set it
• Automatic dependency license checking added to commit workflow

Data Migration

• Default report path has changed to use /Users/<your_user>/.ttnn-visualizer/reports so you can either move any existing reports there or set it locally with the env var APP_DATA_DIRECTORY.

What's Changed

• Fix/License check not running correctly by @dcblundell in #1193
• Feature/Support latest perf report version by @dcblundell in #1196
• Refactor/Modify regex to improve compatibility by @dcblundell in #1181
• Feature/Replace StackByIn0 with GroupBy by @dcblundell in #1200
• Change default paths for app db and reports by @smountenay-tt in #1201
• Device Operations Graph readability improvement - big refactor by @aidemsined in #1202

Full Changelog: v0.70.0...v0.71.0

v0.70.0

What's New

• Added support for tt-perf-report's Tracing Mode in the performance table
• Reorganized the Tensor and Buffer tooltip layout for clarity
• Moved and improved fetch retry logic to Axios Interceptor
• Fixed a conflicting peer dependency with vis packages and updated other FE dependencies
• Improved edge rendering and fixed overlapping edges in Operations Graph
• Added compact view option to Operations Graph
• Enhanced tensor details rendering in Graph
• Fixed display for unsupported architecture in NPE
• Added indexes to report database to improve handling of very large report databases
• Removed use of the num_storage_cores field and ensured backwards compatibility with report databases that still have it
• Added support for reading remote synced report's mesh descriptor

What's Changed

• Add compact view toggle to OperationGraph by @aidemsined in #1163
• Improve edge rendering in OperationGraphComponent by @aidemsined in #1162
• Chore/Rename function by @dcblundell in #1158
• Feature/Use cssVar across app by @dcblundell in #1154
• Fix/Conflicting peer dependency by @dcblundell in #1155
• Feature/Improve JSON refetch by @dcblundell in #1164
• Remove num_storage_cores field from device schema by @smountenay-tt in #1175
• Remove placeholder remote mesh descriptor code by @smountenay-tt in #1174
• Show toast for unsupported architecture in NPEView - fixes react error by @aidemsined in #1171
• Enhance tensor details with MemoryTag and readable dtype by @aidemsined in #1169
• Feature/Tracing mode by @dcblundell in #1166
• Feature/Tensor tooltip layout improved by @dcblundell in #1172

Full Changelog: v0.69.0...v0.70.0

v0.69.0

What's New

• Included paths in error message when app fails to launch with --profiler-path or --performance-path
• Toast notifications now have several states (info/success/warning/error) which have been applied to existing usages in the appliation
• Standardized memory size across the app to KiB for improved consistency
• Added a scroll shade indicator to the legend for better visual context
• Enabled deselection of remote reports when changing remote or switching to a locally uploaded report
• Highlighted input and output nodes and edges in the graph view
• Fixed potential word wrapping issues in various data tables
• Fixed focus selection behavior for disconnected operations
• Fixed invalid graph configuration options
• Physical chip mesh coordinate mapping files will now be copied to TTNN memory reports (tt-metal)

What's Changed

• Feature/Include paths when report preload paths fail by @dcblundell in #1143
• Feature/Expand Toast UI states by @dcblundell in #1136
• Feature/Use KiB for sizes by @dcblundell in #1126
• Fix placement of improvedLayout option in graph config by @aidemsined in #1151
• Feature/Add scroll shade to legend by @dcblundell in #1140
• Feature/Deselect remote reports by @dcblundell in #1138
• Highlight input/output nodes and edges in graph by @aidemsined in #1144
• Fix/Word break in tables by @dcblundell in #1141
• Fix focus node selection for disconnected operations by @aidemsined in #1153

Full Changelog: v0.68.0...v0.69.0

v0.68.0

What's new

• Enhanced remote folder sync data validation and expanded error feedback to the user
• Active memory and performance report locations are now tracked in a given instance e.g. (remote or local)
• User feedback button user interface updated to increase prominence
• Chip mesh coordinate mapping local api implementation and integration with cluster description where available
• Set up workflow to automatically fix detected Python code issues
• Expanded logging when TT-NN Visualizer auto launches to indicate what phase it is in and to highlight known problems
• Cleaned up front end dependency list by moving dev specific dependencies into their correct space

What's Changed

• Fix/SSH Folder Validation by @dcblundell in #1123
• Update flask:lint script to use black --fix by @aidemsined in #1124
• Feature/Add report location to instance by @dcblundell in #1122
• Chore/Moved packages to devDependencies by @dcblundell in #1129
• Chore/Expand auto launch logging by @dcblundell in #1128
• Feature/Improve feedback UX by @dcblundell in #1131
• Feature/Remote Error Handling by @dcblundell in #1125
• Load mesh descriptor by @aidemsined in #1133

Full Changelog: v0.67.0...v0.68.0

v0.67.2

What's Changed

• Fix/SSH Folder Validation by @dcblundell in #1123

Full Changelog: v0.67.1...v0.67.2

v0.67.0

What's New

• Changed browser launch to wait for server
• Allow scroll restoration on filtered lists
• Buffer summary table now hides Device column when not multi device
• Core count fallback added for Blackhole architecture
• Added additional report generation details to TT-NN Visualizer documentation
• Expanded front end test suite
• Improved file upload status information
• Improved feedback UI
• Improve torch tensor size and shape readability
• Fix for an edge case where L1 small marker is positioned incorrectly
• Improved SPDX checking support
• Added rendering tensor sizes and ability to sort by it
• Added device operation filtering in operations list

What's Changed

• Enable portal for feedback dialog and add custom style by @aidemsined in #1096
• Feature/Add core fallback for Blackhole by @dcblundell in #1094
• Feature/Change browser launch approach by @dcblundell in #1100
• Docs/Add generated report folder paths by @dcblundell in #1099
• Support torch.Size in toReadableShape function by @aidemsined in #1098
• Use L1 default memory size as fallback in useGetL1SmallMarker by @aidemsined in #1103
• Feature/Allow scroll restoration on filtered lists by @dcblundell in #1097
• Add size field to tensors list response by @smountenay-tt in #1108
• Improve SPDX license check for prior years by @aidemsined in #1114
• Feature/Hide device column when not multi device by @dcblundell in #1105
• Feature/File Overlay Formatting by @dcblundell in #1116
• Feature/Improve local folder tests by @dcblundell in #1104
• Feature/device operations rendering by @aidemsined in #1113
• Feature/Sort TensorList by size by @aidemsined in #1101
• Refactor device operation handling in OperationList by @aidemsined in #1117
• Chore/Improve SPDX check by @dcblundell in #1119

Full Changelog: v0.66.0...v0.67.0

v0.66.0

What's New

• Now disabling merge multi device option on the performance table when there is no multi device data
• Axis labels for memory plots now display formatted numbers
• Buffer sizes now formatted with units
• Architectural data for Blackhole configurations now extracted from the device log where available
• Fixed an issue extracting layout information for devices with a double digit id
• Fixed a bug which caused an error when filtering operation or tensor lists in certain conditions
• Operations graph - deallocate operations are now hidden by default
• Updated graph lib dependency and improved rendering of operation graph. updated instructions.
• Fixed edge case with NPE rendering incorrect file and making extraneous api call
• Fixed perf rendering NPE timeline availability when there isnt one
• Added core labeling and units in device operation details
• Removed cluster rendering fallback for missing chips to prevent confusing output

What's Changed

• Fix/Virtual List Out Of Bounds by @dcblundell in #1061
• Fix/Row Attributes Not Extracting by @dcblundell in #1064
• Docs/Update Troubleshooting Headers by @dcblundell in #1059
• Feature/Support Blackhole Arch Core Count by @dcblundell in #1066
• Feature/Axis label formatting by @dcblundell in #1088
• Feature/Update buffer size in tooltips by @dcblundell in #1087
• Feature/Disable Merge Multi Device Conditionally by @dcblundell in #1089
• Rename and invert deallocate filter logic by @aidemsined in #1071
• Improve NPE error handling and query logic by @aidemsined in #1067
• Improve graph usability + version bump by @aidemsined in #1090
• Clarify memory load and core labeling in UI by @aidemsined in #1092
• Handle unsupported cluster topologies in ClusterRenderer by @aidemsined in #1091

Full Changelog: v0.65.0...v0.66.0

v0.65.1

What's Changed

• Fix/Virtual List Out Of Bounds by @dcblundell in #1061

Full Changelog: v0.65.0...v0.65.1

v0.65.0

What's New

• Massively increased resolution speed for device architecture
• Can now toggle between merging device rows in performance data
• Removed unused parsing library from project
• Fixed bugs in tt-perf-report and added new optional signpost arguments
• Device operations now visible in operation graph
• Performance table formatting optimised
• Removed unused tensor size fetching to speed up loading
• Updates to TT-NN Visualizer documentation
• Added error in TT-Metal when trying to use Event Synchronisation with tracing

What's Changed

• Feature/1039 move arch and frequency to python by @aidemsined in #1043
• Docs/Update Logo Link by @dcblundell in #1044
• Show device operations in operation graph by @aidemsined in #1051
• Add section about event synchronization error to troubleshooting guide by @smountenay-tt in #1048
• Feature/Optimise Perf Table Formatting by @dcblundell in #1050
• Add footerSpan to handle footer widths by @dcblundell in #1046
• Feature/Update tt-perf-report to v1.1.14 by @dcblundell in #1054
• Removed unused tensor size fetching to speedup the load by @aidemsined in #1055
• Feature/Support no merge rows by @dcblundell in #1042
• Chore/Remove Memoisation From Perf Fetch by @dcblundell in #1057

Full Changelog: v0.64.0...v0.65.0