FMInference / FlexLLMGen

Running large language models on a single GPU for throughput-oriented scenarios.
Apache License 2.0
9.18k stars 548 forks source link

AttributeError: 'NoneType' object has no attribute 'stream_id' #134

Open neomi-tenenbaum-huawei opened 7 months ago

neomi-tenenbaum-huawei commented 7 months ago

on file general/hooks/forward.py on FlexPrepare init (~line 64)

self.use_streams = torch.cuda.is_available() and policy.overlap self.streams = {} self.streams["prev_layer"] = torch.cuda.Stream() if self.use_streams else None self.streams["next_layer"] = torch.cuda.Stream() if self.use_streams else None self.streams["prev_batch"] = torch.cuda.Stream() if self.use_streams else None self.streams["next_batch"] = torch.cuda.Stream() if self.use_streams else None self.stream_names = list(self.streams.keys())

    for name, stream in self.streams.items():
            logger.info(f"{name}: id-{stream.stream_id}")

if 'stream' is None (self.use_streams is False), I will get the error: AttributeError: 'NoneType' object has no attribute 'stream_id'

you have to change the code to: for name, stream in self.streams.items(): if stream: logger.info(f"{name}: id-{stream.stream_id}")