def run(conn: Connection,
config: dict,
model: torch.nn.Module,
log_queue: Optional[mp.Queue] = None):
log.configure(log_queue)
inference_proc = InferenceProcess(config, model)
srv = MPServer(inference_proc, conn)
srv.listen()
def run(conn: Connection,
config: dict,
model: torch.nn.Module,
log_queue: Optional[mp.Queue] = None):
log.configure(log_queue)
# print('CUDA_VISIBLE_DEVICES:', os.environ["CUDA_VISIBLE_DEVICES"])
dryrun_proc = DryRunProcess(config, model)
srv = MPServer(dryrun_proc, conn)
srv.listen()
def run(
conn: Connection,
config: dict,
model: torch.nn.Module,
optimizer_state: bytes = b"",
log_queue: Optional[mp.Queue] = None,
):
log.configure(log_queue)
training_proc = TrainingProcess(config, model, optimizer_state)
srv = MPServer(training_proc, conn)
srv.listen()
def run(
conn: Connection,
config: dict,
model_file: bytes,
model_state: bytes,
optimizer_state: bytes,
log_queue: Optional[mp.Queue] = None,
):
log.configure(log_queue)
handler = HandlerProcess(config, model_file, model_state, optimizer_state,
log_queue)
srv = MPServer(handler, conn)
srv.listen()
def _run_model_session_process(
conn: Connection, model_zip: bytes, devices: List[str], log_queue: Optional[_mp.Queue] = None
):
try:
# from: https://github.com/pytorch/pytorch/issues/973#issuecomment-346405667
import resource
rlimit = resource.getrlimit(resource.RLIMIT_NOFILE)
resource.setrlimit(resource.RLIMIT_NOFILE, (4096, rlimit[1]))
except ModuleNotFoundError:
pass # probably running on windows
if log_queue:
log.configure(log_queue)
session_proc = ModelSessionProcess(model_zip, devices)
srv = MPServer(session_proc, conn)
srv.listen()
def _srv(conn, log_queue):
log.configure(log_queue)
srv = MPServer(ApiImpl(), conn)
srv.listen()
def _run_srv(srv_cls, conn, log_queue):
log.configure(log_queue)
srv = MPServer(srv_cls(), conn)
srv.listen()
def _cancel_srv(conn, log_queue):
log.configure(log_queue)
srv = MPServer(CancelableSrv(), conn)
srv.listen()