Tensorboardx: add_text not working unless closing writer

[JuanFMontesinos]

Hi
I've noticed (also after reading other issues) that add_text does not display it until you closes the writer.
tensorboardX 1.6
tensorboard 1.8
Is there any reason why?

[dumitrescustefan]

I have the same problem. Also, writing several times (with incremental time steps) only produces the first step on screen (after closing the SummaryWriter and refreshing tensorboard).

[BPiepmatz]

For me it is also not showing even though with writer.close()
this code produces the image below:
writer.add_text('lr', 'learning rate', lr) writer.add_text('wd', 'weight decay', wd)

tensorboardX 1.6
pytorch 0.4
tensorboard 1.12

[lanpa]

Hi all, I need some code to reproduce the issue. @BPiepmatz The third argument accepts integer (the iteration number). For your case, add_text('learning rate', str(lr), 1) makes more sense.

[BPiepmatz]

@lanpa Oh I see, same issue still appearing. This is the whole code ` learning_rates = [1e-2, 1e-3, 1e-4, 1e-5, 1e-6, 1e-7, 1e-8]
wd = 0.0002
opt_names = ['RMSprop']
nAveGrad = 5 # Average the gradient every nAveGrad iterations
nEpochs = 200 * nAveGrad # Number of epochs for train
snapshot = nEpochs # Store a model every snapshot epochs
gpu_id = 0
device = torch.device("cuda:" + str(gpu_id) if torch.cuda.is_available() else "cpu")

for lr in learning_rates:
    for opt_name in opt_names:
        model_file_name = datetime.now(gettz('Europe/Berlin')).strftime(
            '%b%d_%H-%M-%S') + '_' + socket.gethostname() + '-' + str(
            lr) + '-' + str(nEpochs) + '-' + str(wd) + '-' + str(lr) + '-' + str(opt_name)
        log_dir = os.path.join('tensorboard_log/runs', model_file_name)

        # Logging into Tensorboard
        writer = SummaryWriter(log_dir=log_dir)

        # Load Network
        print('Loading model...')
        net = CustomNet(pretrained=False)
        optimizer = choose_optimizer(opt_name, net)
        writer.add_text('lr', str(lr), 1)
        writer.add_text('wd', str(wd), 2)
        # writer.add_text('nAveGrad', nAveGrad, 2)
        # writer.add_text('Optimizer', opt_name, 3)

        x = torch.randn(1, 3, 480, 854)
        y = torch.randn(1, 3, 480, 854)
        writer.add_graph(net, (x, y))
        net.to(device)
        writer.close()
        print('Model loaded')`

Does this help?

Best

[lanpa]

@BPiepmatz It should be writer.add_text('lr', str(lr), n_iter)
I would write:

for n_iter, lr in enumerate(learning_rates):
  writer.add_text('lr', str(lr), n_iter)

[ZizhouJia]

I also have the same problem
tensorboardX 1.6
pytorch 1.0
tensorboard 1.12

[Guitaricet]

I have the very same problem - blank field instead of text.

After checking .tfevents file, everything seems to be on its place. It is hard to say definitely, but It is possible, that this is a tensorboard problem.

Part of .tfevents where I can find my texts:

loss/train^U/A  Cpç\^N"^@^@^@^@^@^@^@x^Q=ý      (>ò^H"1×A^PÐ^O*^T
^R
^Kspeed/train^U<89>×^HFMW^B3-^@^@^@^@^@^@^@<A<8a>û      2       û^H"1×A^PÐ^O*^_
^]
^Vgpu_memory_usage/train^U^@<88><9c>EÇd®t0^@^@^@^@^@^@^@<95>õÐ_ L3ô     "1×A^PÐ^O*"
 :^DtestB^N^H^G^R^D^R^B^H^AB^DtestJ^H
^F
^Dtext^P^FÙ2¯^B^@^@^@^@^@^@r¥öÔ ^P9ô    "1×A^PÐ^O* ^E
<9d>^E:^FsourceB<88>^E^H^G^R^D^R^B^H^ABý^DChavez checked his bolter and slapped home afresh magazine .
I am now married to Gweneth , who is English , and we have two children , Carl and Michelle .
In other words , NATO is very much interested in a number of joint operations with Russia .
The purpose in using such contracts and clauses is to provide the firm legal basis required for concluding an agreement acceptable to both sides .
Of course , political struggles are never - ending , but in this case the mass media are hostage to artificially created events that are often orchestrated by spin doctors .
By 2005 it became necessary to address additional user requirements .J^H
^F
^Dtexth²¢hÀ^D^@^@^@^@^@^@jú<9c>ö        )>ô     "1×A^PÐ^O*±
®       :^FtargetB<99>  ^H^G^R^D^R^B^H^AB<8e>   <s> Ð<9a>апиÑ<82>ан пÑ<80>овеÑ<80>ил его болÑ<82>еÑ<80> и вÑ<81>Ñ<82>авил новÑ<8b>й магазин . </s>
<s> СейÑ<87>аÑ<81> Ñ<8f> женаÑ<82> на Ð<93>винеÑ<82> , она англиÑ<87>анка , и Ñ<83> наÑ<81> двое деÑ<82>ей : Ð<9a>аÑ<80>л и Ð<9c>иÑ<88>елÑ<8c> . </s>
<s> Ð<94>Ñ<80>Ñ<83>гими Ñ<81>ловами , Ð<9d>Ð<90>ТÐ<9e> веÑ<81>Ñ<8c>ма заинÑ<82>еÑ<80>еÑ<81>овано в Ñ<80>Ñ<8f>де Ñ<81>овмеÑ<81>Ñ<82>нÑ<8b>Ñ<85> Ñ<81> РоÑ<81>Ñ<81>ией опеÑ<80>аÑ<86>ий , вÑ<8b>полнение коÑ<82>оÑ<80>Ñ<8b>Ñ<85> длÑ<8f> ниÑ<85> жизненно важно . </s>
<s> ЦелÑ<8c> иÑ<81>полÑ<8c>зованиÑ<8f> Ñ<82>акиÑ<85> конÑ<82>Ñ<80>акÑ<82>ов и оговоÑ<80>ок - Ñ<8d>Ñ<82>о обеÑ<81>пеÑ<87>ение пÑ<80>оÑ<87>ной пÑ<80>авовой базой , необÑ<85>одимой длÑ<8f> заклÑ<8e>Ñ<87>ениÑ<8f> пÑ<80>иемлемого длÑ<8f> обеиÑ<85> Ñ<81>Ñ<82>оÑ<80>он Ñ<81>оглаÑ<88>ениÑ<8f> . </s>
<s> Ð<9d>о СÐ<9c>Ð<98> в данном Ñ<81>лÑ<83>Ñ<87>ае Ñ<81>Ñ<82>ановÑ<8f>Ñ<82>Ñ<81>Ñ<8f> заложниками Ñ<81>озданнÑ<8b>Ñ<85> Ñ<81>пеÑ<86>иалÑ<8c>но Ñ<81>обÑ<8b>Ñ<82>ий , за коÑ<82>оÑ<80>Ñ<8b>ми Ñ<87>аÑ<81>Ñ<82>о Ñ<81>Ñ<82>оÑ<8f>Ñ<82> полиÑ<82>Ñ<82>еÑ<85>нологи . </s>
<s> Ð<9a> 2005 годÑ<83> поÑ<8f>вилаÑ<81>Ñ<8c> необÑ<85>одимоÑ<81>Ñ<82>Ñ<8c> в Ñ<80>аÑ<81>Ñ<88>иÑ<80>ении возможноÑ<81>Ñ<82>ей пÑ<80>оÑ<82>окола . </s>J^H
^F
^Dtext0wíOû^A^@^@^@^@^@^@^^<9f><98>û    ^T[ô    "1×A^PÐ^O*ì^C
é^C:
translatedBÐ^C^H^G^R^D^R^B^H^ABÅ^C<unk> <unk> и <unk> <unk> .
Я <unk> , <unk> , <unk> , <unk> , <unk> , <unk> и <unk> .
Ð<9a>Ñ<80>оме Ñ<82>ого , <unk> , <unk> , <unk> <unk> <unk> <unk> .
<unk> <unk> , <unk> и <unk> , <unk> , <unk> и <unk> , в Ñ<82>ом Ñ<87>иÑ<81>ле как <unk> <unk> , <unk> <unk> <unk> <unk> .
<unk> <unk> , но не <unk> , но не <unk> , но не <unk> , но в Ñ<8d>Ñ<82>ом Ñ<81>лÑ<83>Ñ<87>ае <unk> <unk> .
Ð<92> наÑ<81>Ñ<82>оÑ<8f>Ñ<89>ее вÑ<80>емÑ<8f> <unk> <unk> <unk> <unk> <unk> .J^H
^F
^Dtext13ûx!^@^@^@^@^@^@^@^A{ìö  uèj
"1×A^PÚ^O*^S
^Q

loss/train^UÊ<82>^OC´æ+."^@^@^@^@^@^@^@x^Q=ý    cêj
"1×A^PÚ^O*^T
^R
^Kspeed/train^U
^P¨DÞ>"^R!^@^@^@^@^@^@^@^A{ìö   a¼¢
"1×A^Pä^O*^S
^Q

[lanpa]

We will have flush() for the summarywriter after #451 is merged. Then you can call writer.flush() manually to replace close().