On Tue, Aug 21, 2018 at 1:16 PM wrote:
> Sorry, one more question. We set up 4 different machines all running the
> command below except for minor differences in the momentum and the learning
> rate. Changing the momentum and learning rate in this situation, because
> it is fine tuning, shouldn
On Tue, Aug 21, 2018 at 2:47 PM Jacob Biros
wrote:
> Thank you for your response. The command that I posted previously was for
> fine tuning, so I was hoping to figure out what was causing the changes in
> accuracy when changes to the parameters should have no effect at all.j
>
> I'm not sure if
Oh, that thought hadn't even occurred to me. I will look into it!
Thank you!
On Tue, Aug 21, 2018 at 6:43 PM, Shree Devi Kumar
wrote:
> When you specify the complete network spec as in
>
> --net_spec '[1,36,0,1 Ct3,3,16 Mp3,3 Lfys48 Lfx96 Lrx96 Lfx256 O1c111]' \
>
>
> It probably treats it as
When you specify the complete network spec as in
--net_spec '[1,36,0,1 Ct3,3,16 Mp3,3 Lfys48 Lfx96 Lrx96 Lfx256 O1c111]' \
It probably treats it as a training from scratch and ignores the
continue_from.
I haven't looked at how the training command is parsed. I just followed
Ray's examples for th
Thank you for your response. The command that I posted previously was for
fine tuning, so I was hoping to figure out what was causing the changes in
accuracy when changes to the parameters should have no effect at all.j
I'm not sure if we will move forward with replacing the layers, but I will
po
>lstmtraining --model_output ./layer_from_deva/layer --continue_from
./layer_from_deva/Devanagari.lstm --append_index 5 --net_spec '[Lfx192
O1c1]' --traineddata ./sanplustrain/san/san.traineddata --train_listfile
./sanplustrain/san.training_files.txt --eval_listfile
./sanpluseval/san.training_files
Sorry, one more question. We set up 4 different machines all running the
command below except for minor differences in the momentum and the learning
rate. Changing the momentum and learning rate in this situation, because
it is fine tuning, shouldn't affect anything right? In our case though
Thank you for the prompt response!
On Tue, Aug 21, 2018 at 1:14 PM, Shree Devi Kumar
wrote:
> If you want to change parameters, please look at the replace layers
> option. With fine tuning you cannot change them.
>
> On Tue 21 Aug, 2018, 7:27 AM , wrote:
>
>> Is it possible to change the parame
If you want to change parameters, please look at the replace layers option.
With fine tuning you cannot change them.
On Tue 21 Aug, 2018, 7:27 AM , wrote:
> Is it possible to change the parameters when Fine Tuning?
>
> The documentation says "Fine tuning is the process of training an existing
>
9 matches
Mail list logo
