slowfast在summary的时候报错

[Bradly-s]

summary.py部分代码 def main(): args = parse_args() cfg, model_name = _trim(get_config(args.config, show=False), args) print(f"Building model({model_name})...") model = build_model(cfg)

img_size = args.img_size
num_seg = args.num_seg
#NOTE: only support tsm now, will refine soon
params_info = paddle.summary(model, (1, 1, num_seg, 3, img_size, img_size))
# print(params_info)

args.FLOPs = True
if args.FLOPs:
    flops_info = paddle.flops(model, [1, 1, num_seg, 3, img_size, img_size], print_detail=True)
    print("args.FLOPs:")
    print(flops_info)

if name == "main": main()

报错： Traceback (most recent call last): File "/mnt/sdb1/swf/project/PaddleVideo/tools/summary.py", line 109, in main() File "/mnt/sdb1/swf/project/PaddleVideo/tools/summary.py", line 92, in main params_info = paddle.summary(model, (1, 1, num_seg, 3, img_size, img_size)) File "/home/pl/anaconda3/envs/swfpd/lib/python3.7/site-packages/paddle/hapi/model_summary.py", line 308, in summary result, params_info = summary_string(net, _input_size, dtypes, input) File "/home/pl/anaconda3/envs/swfpd/lib/python3.7/site-packages/decorator.py", line 232, in fun return caller(func, *(extras + args), kw) File "/home/pl/anaconda3/envs/swfpd/lib/python3.7/site-packages/paddle/base/dygraph/base.py", line 350, in _decorate_function return func(*args, kwargs) File "/home/pl/anaconda3/envs/swfpd/lib/python3.7/site-packages/paddle/hapi/model_summary.py", line 444, in summary_string model(x) File "/home/pl/anaconda3/envs/swfpd/lib/python3.7/site-packages/paddle/nn/layer/layers.py", line 1423, in call return self.forward(inputs, kwargs) File "/mnt/sdb1/swf/project/PaddleVideo/paddlevideo/modeling/framework/recognizers/base.py", line 55, in forward return self.infer_step(data_batch) File "/mnt/sdb1/swf/project/PaddleVideo/paddlevideo/modeling/framework/recognizers/recognizer3d.py", line 91, in infer_step cls_score = self.forward_net(imgs) File "/mnt/sdb1/swf/project/PaddleVideo/paddlevideo/modeling/framework/recognizers/recognizer3d.py", line 28, in forward_net feature = self.backbone(imgs) File "/home/pl/anaconda3/envs/swfpd/lib/python3.7/site-packages/paddle/nn/layer/layers.py", line 1425, in call return self._dygraph_call_func(*inputs, kwargs) File "/home/pl/anaconda3/envs/swfpd/lib/python3.7/site-packages/paddle/nn/layer/layers.py", line 1404, in _dygraph_call_func outputs = self.forward(*inputs, kwargs) File "/mnt/sdb1/swf/project/PaddleVideo/paddlevideo/modeling/backbones/resnet_slowfast.py", line 776, in forward x = self.s1(x) #VideoModelStem File "/home/pl/anaconda3/envs/swfpd/lib/python3.7/site-packages/paddle/nn/layer/layers.py", line 1425, in call return self._dygraph_call_func(*inputs, *kwargs) File "/home/pl/anaconda3/envs/swfpd/lib/python3.7/site-packages/paddle/nn/layer/layers.py", line 1404, in _dygraph_call_func outputs = self.forward(inputs, kwargs) File "/mnt/sdb1/swf/project/PaddleVideo/paddlevideo/modeling/backbones/resnet_slowfast.py", line 505, in forward self.num_pathways) AssertionError: Input tensor does not contain 2 pathway

Process finished with exit code 1

[westfish]

根据你提供的错误信息，问题出在resnet_slowfast.py文件的第505行，这里的self.num_pathways期望输入张量包含2个路径，但实际上并没有，这可能是因为你的模型配置或输入数据的问题。

另外，如果有图像、视频理解和生成的需求，可以使用我们新的跨模态工具: https://github.com/PaddlePaddle/PaddleMIX/tree/develop