v0.17.1

What's Changed

• Feature/Wheel auto launch by @dcblundell in #317
• Feature/Remove TT assets by @dcblundell in #321

Full Changelog: v0.17.0...v0.17.1

v0.17.0

Release notes:

• Improved operation names rendering using filenames
• Improved Tensor rendering emphasizing shape and dtype and referencing producer operation
• Improved memory legend tensor information
• Improved operation graph functionality with the above updates and properties panel
• Added additional charts
• Unified performance chart configuration
• Fixed rendering issue when switching to and from chart tab
• Chart data can now be filtered by operation codes

What's Changed

• Combine chart configs by @dcblundell in #312
• Improve operations and tensors rendering by @aidemsined in #313
• Operation graph render improvements by @aidemsined in #314
• Allow perf data filtering by @dcblundell in #315

Full Changelog: v0.16.0...v0.17.0

v0.16.0

Release Notes

• Local and Remote report selection upgraded to include clearer messaging on the current active report/performance trace with toasts. Active reports/performance traces are updated on selection rather than with the View Report button which has been removed.
• Device kernel performance charts added to the Performance tab.
• Additional performance data parsed and returns the device architecture and frequency.
• Tensor list now includes indicators for where tensors have an excessive amount of consumers. The list can also be filtered to show only these tensors.
• Chart legends in operation details now include tensor shape and data type.

What's Changed

• Feature/read cluster desc file by @GregHattJr in #298
• Feature/local-upload-ui-updates by @dcblundell in #302
• Feature/Remove Device Selector by @dcblundell in #304
• get device architecture and frequency by @aidemsined in #306
• Fix/session store loaded files by @dcblundell in #305
• Feature/Add Device Kernel charts by @dcblundell in #307
• support for device id by @aidemsined in #308
• Feature/Add high consuming warning to tensor list by @dcblundell in #309
• show tensor details in legend by @aidemsined in #310

Full Changelog: v0.15.0...v0.16.0

v0.15.0

Release Notes:

• Readme has been updated with model running instructions and an example performance trace data
• Memory config parsing expanded to more operation arguments
• Performance section now features Device Kernel Duration/Utilization charts for Matmul operations
• Operations by type now plotted in a chart
• Performance table improvements, Matmul optimisation suggestions and high dispatch latency
• Backend decoupling of profile files from reports
• Additional codebase cleanup in preparation for open source

What's Changed

• remove reading packaje.json from incorect location by @aidemsined in #286
• Fix/minor cleanup by @dcblundell in #288
• update model running instructions by @aidemsined in #291
• Feature/Expand memory config parsing by @dcblundell in #290
• Feature/readme updates for perf reports by @dcblundell in #292
• Feature/add operation types chart by @dcblundell in #293
• Feature/Performance scatter chart by @dcblundell in #294
• Perf table improvements, Matmul and high dispatch latency by @aidemsined in #297
• Decouple Profile Files from Reports by @GregHattJr in #299
• Feature/Update op types logic in pie chart by @dcblundell in #296
• Feature/align perf report to ops by @aidemsined in #300
• Feature/Performance Mixed Data Chart by @dcblundell in #301

Full Changelog: v0.14.1...v0.15.0

v0.14.2

Release notes

• fix broken wheel

What's Changed

• remove reading packaje.json from incorect location by @aidemsined in #286

Full Changelog: v0.14.1...v0.14.2

v0.14.1

Release notes:

• Load performance data
• Intimal rendering of performance data analysis table
• Misc: SPDX updates for open source
• Fix: improved multi device support for L1 and tensor list rendering
• Fix: L1 and DRAM zoom inconsistencies
• Fix: color logic for buffer pages when data is insufficient

What's Changed

• Feature/spdx update by @aidemsined in #273
• Feature/Add layout details to buffer summary tooltips by @dcblundell in #272
• update title by @aidemsined in #275
• DRAFT: Feature/add profiler uploader by @GregHattJr in #276
• Feature/add performance route by @dcblundell in #278
• Feature/performance uploading UI by @dcblundell in #274
• initial implementation of perf table by @aidemsined in #279
• Feature/Add labels to layout metadata by @dcblundell in #280
• remove device_id api level by @aidemsined in #281
• fix L1 zoom by @aidemsined in #282
• fix color logic for buffer pages when data is insufficient (multi dev… by @aidemsined in #283

Full Changelog: v0.13.0...v0.14.1

v0.14.0

What's Changed

• Feature/spdx update by @aidemsined in #273
• Feature/Add layout details to buffer summary tooltips by @dcblundell in #272
• update title by @aidemsined in #275
• Feature/add profiler uploader by @GregHattJr in #276
• Feature/add performance route by @dcblundell in #278
• Feature/performance uploading UI by @dcblundell in #274
• initial implementation of perf table by @aidemsined in #279
• Feature/Add labels to layout metadata by @dcblundell in #280
• remove device_id api level by @aidemsined in #281
• fix L1 zoom by @aidemsined in #282

Full Changelog: v0.13.0...v0.14.0

v0.13.0

Release Notes:

• Tensor memory layout overlay for plots in operation details and buffers
• L1 Small detailed rendering
• Tensnor sharding details fixes
• Memory leak logic cleanup
• Improve memory config parsing performance
• Highlight peak memory load
• Readme cleanup in prepartion to open source

What's Changed

• fix tensor sharding shape by @aidemsined in #259
• update FE SPDX headers by @aidemsined in #261
• Feature/Remove a lot of the memory leak logic by @dcblundell in #262
• move memory config parsing to BE by @aidemsined in #264
• Feature/update readme by @dcblundell in #263
• Fixes import path for parse_memory_config import by @GregHattJr in #266
• python SPDX by @aidemsined in #267
• Feature/L1 small separate by @dcblundell in #265
• Feature/Readme add examples by @dcblundell in #268
• highlight peak memory load by @aidemsined in #260
• render Tensor memory layout overlay by @aidemsined in #269
• Feature/add-tensor-patterns-to-buffer-summary by @dcblundell in #270

Full Changelog: v0.12.0...v0.13.0

v0.12.0

Release notes

• Memory load calculation and rendering, peak memory load
• Initial support for multidevice data
• Render memory layout and sharding spec for tensors
• Fix: buffer summary toast not dismissing
• Fix: color rendering on tensix details zoom
• Fix: initial buffer focus
• Fix: toasts added to buffer summary when focusing on buffers
• Test coverage for runners/filtering/query

What's Changed

• Feature/Add device id to tensor list call by @dcblundell in #245
• Fix for correct color rendering on tensix details by @aidemsined in #247
• Feature/buffer summary toasts by @dcblundell in #244
• Feature/memory config parsing refactor by @dcblundell in #248
• Feature/multi-device-select by @dcblundell in #246
• Adds coverage for existing query runners/filtering/query construction by @GregHattJr in #255
• fix/device-operation-data-serialisation by @dcblundell in #257
• Fix/Initial buffer focus by @dcblundell in #253
• Fix/Memory config typing by @dcblundell in #254
• Memory load calculation and rendering, peak memory load by @aidemsined in #250
• render Tensor memoryLaoyut on operation details by @aidemsined in #252
• Feature/device selector op details by @dcblundell in #251

Full Changelog: v0.11.0...v0.12.0

v0.11.0

Release notes:

• Zoom for non-continuous memory allocation
• Device operations full render (memory peak load prerequisite)
• Device operations graph view
• Initial support for multi device data (device 0)
• Buffer highlight improvements

What's Changed

• Added more bold highlights to focused buffers by @dcblundell in #238
• Refactor queries to use generic filtering, adds device filtering by @GregHattJr in #240
• Device operations graph view by @aidemsined in #239
• Device operations full render by @aidemsined in #242
• Feature/non continuous dram zoom by @dcblundell in #241

Full Changelog: v0.10.0...v0.11.0