pochih / Video-Cap

🎬 Video Captioning: ICCV '15 paper implementation
48 stars 20 forks source link

model provided has issue #2

Open Harryhaha opened 6 years ago

Harryhaha commented 6 years ago

Hi, when I load the model-2380, and run the code, it throw the error: "ValueError: Dimensions must be equal, but are 3175 and 3514 for 'xw_plus_b_1' (op: 'BiasAdd') with input shapes: [1,3175], [3514]." Something is incompatible... Could you please take a look ?

pochih commented 6 years ago

I'll take a look, please wait :D

XiangChen1994 commented 6 years ago

I also got the same error. ValueError: Dimensions must be equal, but are 3175 and 3514 for 'xw_plus_b_1' (op: 'BiasAdd') with input shapes: [1,3175], [3514]. How to solve it?

XiangChen1994 commented 6 years ago

The model trained by myself is OK.

sxs4337 commented 6 years ago

Thanks for the code for this great paper! I am getting the same error- ValueError: Dimensions must be equal, but are 3175 and 3514 for 'xw_plus_b_1' (op: 'BiasAdd') with input shapes: [1,3175], [3514].

Any updates on this? I want to try with the pre-trained model before preparing the data to train it myself. Thanks in advance.

nayyeraafaq commented 6 years ago

I got the similar error too. ValueError: Dimensions must be equal, but are 3175 and 3514 for 'xw_plus_b_1' (op: 'BiasAdd') with input shapes: [1,3175], [3514]. Did anyone to solve it? Please share. May be the error is due to probable missing files like training_label.json, val and training_feature files ? Is that right ? Where to get these ? Please reply anyone ????? Thanks

yoavaviram commented 6 years ago

Same error hear.

Annusha commented 6 years ago

Still the same error.

cskostas87 commented 6 years ago

i have the same error, did someone have resolved this?