unmatched iekfnets.p from dropbox

[enguang2]

Hello, The iekfnets.p file downloaded from wget "https://www.dropbox.com/s/77kq4s7ziyvsrmi/temp.zip" does not correspond with the MesNet structure. The fourth layer has different size and apparently the iekfnets.p contains more layes than the 8 layes' MesNet we see in utils_torch_filter.py

[ccsmm78]

I've met the same problem! as followings:

RuntimeError: Error(s) in loading state_dict for TORCHIEKF:
Unexpected key(s) in state_dict: "mes_net.cov_net.8.weight", "mes_net.cov_net.8.bias", "mes_net.cov_net.12.weight", "mes_net.cov_net.12.bias", "mes_net.cov_net.16.weight", "mes_net.cov_net.16.bias".
size mismatch for mes_net.cov_net.4.weight: copying a param with shape torch.Size([64, 32, 5]) from checkpoint, the shape in current model is torch.Size([32, 32, 5]).
size mismatch for mes_net.cov_net.4.bias: copying a param with shape torch.Size([64]) from checkpoint, the shape in current model is torch.Size([32]).

/home/ccsmm/git_repository/ai-imu-dr/src/utils_torch_filter.py(466)load()
465 mondict = torch.load(path_iekf)
--> 466 self.load_state_dict(mondict)
467 cprint("IEKF nets loaded", 'green')

In debugging,
ipdb> mondict.keys()
odict_keys(['initprocesscov_net.factor_initial_covariance.weight', 'initprocesscov_net.factor_process_covariance.weight', 'mes_net.cov_net.0.weight', 'mes_net.cov_net.0.bias', 'mes_net.cov_net.4.weight', 'mes_net.cov_net.4.bias', 'mes_net.cov_net.8.weight', 'mes_net.cov_net.8.bias', 'mes_net.cov_net.12.weight', 'mes_net.cov_net.12.bias', 'mes_net.cov_net.16.weight', 'mes_net.cov_net.16.bias', 'mes_net.cov_lin.0.weight', 'mes_net.cov_lin.0.bias'])

But the self is not the same as the network.
ipdb> self
TORCHIEKF(
(initprocesscov_net): InitProcessCovNet(
(factor_initial_covariance): Linear(in_features=1, out_features=6, bias=False)
(factor_process_covariance): Linear(in_features=1, out_features=6, bias=False)
(tanh): Tanh()
)
(mes_net): MesNet(
(tanh): Tanh()
(cov_net): Sequential(
(0): Conv1d(6, 32, kernel_size=(5,), stride=(1,))
(1): ReplicationPad1d((4, 4))
(2): ReLU()
(3): Dropout(p=0.5)
(4): Conv1d(32, 32, kernel_size=(5,), stride=(1,), dilation=(3,))
(5): ReplicationPad1d((4, 4))
(6): ReLU()
(7): Dropout(p=0.5)
)
(cov_lin): Sequential(
(0): Linear(in_features=32, out_features=2, bias=True)
(1): Tanh()
)
)
)
ipdb

[enguang2]

@ccsmm78 I would guess the URL points to the parameter file hasn't been updated and the parameter file actually has been altered since firstly published. Nevertheless, I am trying to train the model locally yet the training stops at early epoch(7 epoch), have you encountered same case?
:)

[Mieczmik]

@enguang2 I have the same problem. I will give info about progress later.

[Akudavale]

hello EVERYONE,

II am facing same issuee of size mismatch how to resolve it?

[TiloccaS]

hi @Mieczmik @enguang2 I noticed that I have the same problem as yours, have you solved it somehow, or there are news?

[luopengting]

I have fixed the problem, please see #88 , hope it will help you. :D @Akudavale @TiloccaS