Ejemplo n.º 1
0
    def test_submit_models(self):
        _reset()
        nni.retiarii.debug_configs.framework = 'pytorch'
        os.makedirs('generated', exist_ok=True)
        from nni.runtime import protocol
        import nni.runtime.platform.test as tt
        protocol._set_out_file(
            open('generated/debug_protocol_out_file.py', 'wb'))
        protocol._set_in_file(
            open('generated/debug_protocol_out_file.py', 'rb'))

        models = _load_mnist(2)

        advisor = RetiariiAdvisor()
        cgo_engine = CGOExecutionEngine(devices=[
            GPUDevice("test", 0),
            GPUDevice("test", 1),
            GPUDevice("test", 2),
            GPUDevice("test", 3)
        ],
                                        batch_waiting_time=0)
        set_execution_engine(cgo_engine)
        submit_models(*models)
        time.sleep(3)

        if torch.cuda.is_available() and torch.cuda.device_count() >= 2:
            cmd, data = protocol.receive()
            params = nni.load(data)

            tt.init_params(params)

            trial_thread = threading.Thread(
                target=CGOExecutionEngine.trial_execute_graph)
            trial_thread.start()
            last_metric = None
            while True:
                time.sleep(1)
                if tt._last_metric:
                    metric = tt.get_last_metric()
                    if metric == last_metric:
                        continue
                    if 'value' in metric:
                        metric['value'] = json.dumps(metric['value'])
                    advisor.handle_report_metric_data(metric)
                    last_metric = metric
                if not trial_thread.is_alive():
                    trial_thread.join()
                    break

            trial_thread.join()

        advisor.stopping = True
        advisor.default_worker.join()
        advisor.assessor_worker.join()
        cgo_engine.join()
Ejemplo n.º 2
0
def _prepare_receive(data):
    protocol._set_in_file(BytesIO(data))
Ejemplo n.º 3
0
def _restore_io():
    _in_buf.seek(0)
    _out_buf.seek(0)
    protocol._set_in_file(_in_buf)
    protocol._set_out_file(_out_buf)