FluxTraining
"""
TensorBoardBackend(logdir[, tb_overwrite];
time=time(),
purge_step=nothing,
step_increment=1,
min_level=Logging.Info)
TensorBoard backend for logging callbacks. Takes the same arguments
as [`TensorBoardLogger.TBLogger`](https://julialogging.github.io/TensorBoardLogger.jl/dev/).
"""
struct
TensorBoardBackend
<:
LoggerBackend
logger
::
TBLogger
function
TensorBoardBackend
(
logdir
,
existfn
=
tb_overwrite
;
kwargs
...
)
return
new
(
TBLogger
(
logdir
,
existfn
;
kwargs
...
)
)
end
end
Base
.
show
(
io
::
IO
,
backend
::
TensorBoardBackend
)
=
print
(
io
,
"
TensorBoardBackend(
"
,
backend
.
logger
.
logdir
,
"
)
"
)
function
log_to
(
backend
::
TensorBoardBackend
,
value
::
Loggables
.
Value
,
name
,
i
;
group
=
(
)
)
name
=
_combinename
(
name
,
group
)
log_value
(
backend
.
logger
,
name
,
value
.
data
;
step
=
i
)
end
function
log_to
(
backend
::
TensorBoardBackend
,
image
::
Loggables
.
Image
,
name
,
i
;
group
=
(
)
)
name
=
_combinename
(
name
,
group
)
im
=
ImageCore
.
clamp01nan!
(
collect
(
image
.
data
)
)
log_image
(
backend
.
logger
,
name
,
im
;
step
=
i
)
end
function
log_to
(
backend
::
TensorBoardBackend
,
text
::
Loggables
.
Text
,
name
,
i
;
group
=
(
)
)
name
=
_combinename
(
name
,
group
)
log_text
(
backend
.
logger
,
name
,
text
.
data
;
step
=
i
)
end
function
log_to
(
backend
::
TensorBoardBackend
,
hist
::
Loggables
.
Histogram
,
name
,
i
;
group
=
(
)
)
name
=
_combinename
(
name
,
group
)
log_histogram
(
backend
.
logger
,
name
,
hist
.
data
,
step
=
i
)
endUtilities
_combinename
(
name
,
group
::
String
)
=
_combinename
(
(
group
,
name
)
)
_combinename
(
name
,
group
::
Tuple
)
=
_combinename
(
(
group
...
,
name
)
)
_combinename
(
strings
::
Tuple
)
=
join
(
strings
,
'/'
)