v0.1.0

What's Changed

• Update ROCm docker command by @Delaunay in #124
• Validation Layers by @Delaunay in #56
• Update voir to 2.15 by @Delaunay in #135
• Make requirement updates more stable by @breuleux in #133
• Fix instability in Whisper benchmark by @breuleux in #137
• Disable ROCm tests while we find a suitable machine replacement by @Delaunay in #140
• Add system config by @satyaog in #130
• Add metadata gathering by @Delaunay in #132
• New Metric Persistence Backend by @Delaunay in #58
• Add RWKV benchmark by @breuleux in #90
• Tweak Performance computation by @Delaunay in #144
• Add execution plan abstraction by @satyaog in #145
• Make error validation work with python exception by @Delaunay in #154
• Multi node install & prepare by @Delaunay in #153
• Docker tweaks by @Delaunay in #160
• Fix #164: makes sure all the timeout tasks are cancelled by @Delaunay in #165
• Prevent machine_metadata from throwing by @Delaunay in #163
• remove dlrm profiling by @Delaunay in #168
• Add flops benchmark by @Delaunay in #169
• Add new inference bench by @Delaunay in #174
• Autoscale by @Delaunay in #177
• Fix ${{}} in runner by @Delaunay in #175
• Use black by @Delaunay in #178
• Track pytorch version by @Delaunay in #155
• Deploy script by @Delaunay in #182
• Add dataset revision by @Delaunay in #187
• Update README.md by @Delaunay in #180
• Use node["port"] to ssh to the node by @Delaunay in #189
• Build docker container for reporting by @Delaunay in #197
• Add missing property by @Delaunay in #198
• Add git to docker by @Delaunay in #199
• Simplify name by @Delaunay in #200
• Make sure report works without GPU by @Delaunay in #201
• Tag report containers by @Delaunay in #202
• Update README.md by @Delaunay in #209
• Intel GPU Max Support + Gaudi by @Delaunay in #214
• Add Benchmate: benchmark companion lib
• Support for XPU and HPU
• New measure methods based on timing Event
• New compatibility layer for multi vendor support
• New argument placeholders
  {arch}: accelerator arch (cuda, xpu, hpu, etc...)
  {ccl}: communication collection library (nccl, rccl, ccl, hccl, etc...)
  {cpu_count}: number of CPU available on the machine
  {cpu_per_gpu} : number of CPU available per GPUs cpu_count / device_count
  {n_worker}: recommended number of workers min(cpu_per_gpu, 16)

Full Changelog: v0.0.6...v0.1.0