petals/cli
Alexander Borzunov 8a73b41a42
Make ServerState announcements work better (#93)
- Before this PR, `ServerState.JOINING` was announced only once. This announcement quickly expires in case of the full-size BLOOM, since loading blocks takes several minutes. This PR fixes it, so `ServerState.JOINING` is announced periodically in a thread until blocks are loaded.

- This PR also makes the `Server` class a non-thread, so it runs in the main thread and can catch `KeyboardInterrupt`. This is important, since if we are downloading blocks right now, we need to stop it and send the `ServerState.OFFLINE` message. Note that `ModuleContainer` is still a thread.

- (minor) For the sake of readability, I moved the `ModuleContainer.create()` definition, so it is now defined before `Server.__init__()` (this is because `.create()` is invoked first).
2022-11-28 07:44:03 +04:00
..
__init__.py add quantization script for cpu 2022-06-12 05:59:11 +03:00
config.json add minimalistic benchmarks 2022-06-14 15:18:11 +03:00
convert_model.py Reduce vocabulary size in test model, fix bug in routing when overlapped (#45) 2022-08-17 18:50:52 +03:00
deploy_server.sh integrate mixed-8bit model (#39) 2022-08-04 09:57:37 +03:00
inference_one_block.py WIP: make DistributedBloom compliant with HF interface 2022-07-07 03:11:28 +03:00
local_server_config_example.cfg deploy swarm on local & remote machines 2022-06-29 13:52:43 +03:00
remote_server_config_example.cfg deploy swarm on local & remote machines 2022-06-29 13:52:43 +03:00
run_local_servers.sh integrate mixed-8bit model (#39) 2022-08-04 09:57:37 +03:00
run_remote_servers.sh Sequential and parallel forward / backward (#36) 2022-07-23 14:32:39 +03:00
run_server.py Make ServerState announcements work better (#93) 2022-11-28 07:44:03 +04:00
speed_test.py Add automated tests (#23) 2022-07-16 01:59:23 +03:00