1.0.0: hivemind.Optimizer, improved averaging stability, better logging

What's Changed

• Fix averager speed for TCP connections by @borzunov in #373
• Fix "Too many open files" and load state freezing by @justheuristic in #371
• Prefetch while reading rpc_aggregate_part() outputs by @borzunov in #370
• Use ModeClient in libp2p DHT in case of --client_mode by @borzunov in #374
• Integrate p2pd logs and outputs into hivemind logging by @borzunov in #375
• Split compression strategies into separate classes by @justheuristic in #366
• Implement colored logs by @borzunov in #377
• Parametrize max message size for persistent connections by @deniskamazur in #376
• Make log handlers configurable, shorten entries by @borzunov in #378
• Enable log handler in benchmarks and run_server by @borzunov in #380
• Fix step_tolerance in CollaborativeOptimizer by @justheuristic in #383
• Fix pickle vulnerability by @deniskamazur in #386
• Remove arguments with default values from example instructions by @borzunov in #388
• Implement weight as part of the allreduce protocol, not matchmaking by @justheuristic in #384
• Support different AMP & buffer configurations in one experiment, fix minor bugs by @justheuristic in #389
• Fix codecov_in_develop_mode with pip>=21.2 by @justheuristic in #393
• Fix minor issues in documentation by @borzunov in #392
• Apply averager updates asynchronously by @justheuristic in #395
• Fix schema typing by @justheuristic in #396
• backport PerformanceEMA from server_side_averaging by @justheuristic in #397
• Add an option to pre-schedule averaging by @justheuristic in #398
• Move DHT to dht/dht.py, update DHT figure by @justheuristic in #399
• [hotfix] replace StepControl.can_modify with began_allreduce by @justheuristic in #402
• move PerformanceEMA to utils, TrainingAverager to optim, update utils by @justheuristic in #405
• Add GradientAverager with support for delayed averaging by @justheuristic in #404
• [hivemind.Optimizer] TrainingStateAverager by @justheuristic in #407
• Catch OSError in MPFuture by @artek0chumak in #409
• [hivemind.Optimizer] ProgressTracker by @justheuristic in #408
• Fix minor bugs in GradientAverager by @justheuristic in #410
• Make target group size optional by @justheuristic in #412
• Prepare GradScaler for hivemind.Optimizer by @justheuristic in #413
• Patch recursive cancel in StepControl by @justheuristic in #411
• Replace the invalid link to discord by @artek0chumak in #414
• Implement state sharing priority by @justheuristic in #415
• Implement core functionality of hivemind.Optimizer by @justheuristic in #403
• DHT Benchmark with asynchronous w/r by @MuXauJl11110 in #406
• Hotfix: load_state_from_peers with offload_optimizer by @justheuristic in #417
• Improve Optimizer docs, update quickstart to use Optimizer by @justheuristic in #416
• Quickstart: typos and references by @justheuristic in #420
• Remove trailing dots in log messages and errors by @borzunov in #419
• Do not log caller for INFO messages by @borzunov in #418
• Improve hivemind.optim.experimental and averager stability by @borzunov in #421
• Add minor tweaks learned from the NeurIPS demo run by @justheuristic in #422
• Improve All-Reduce fault-tolerance by @justheuristic in #423
• Fix All-Reduce fault-tolerance: catch Exception instead of BaseException by @justheuristic in #424
• Fix Task was destroeyd but is pending (put items) by @justheuristic in #427
• Use hivemind.Optimizer in examples/albert by @mryab in #426

New Contributors

• @artek0chumak made their first contribution in #409
• @MuXauJl11110 made their first contribution in #406

Full Changelog: 0.10.0...1.0.0