[
https://issues.apache.org/jira/browse/MADLIB-1393?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ekta Khanna updated MADLIB-1393:
--------------------------------
Description:
Single fit()
{code}
madlib_keras_fit(
source_table,
model,
model_arch_table,
model_arch_id,
compile_params,
fit_params,
num_iterations,
use_gpus, -- changed definition
validation_table,
metrics_compute_frequency,
warm_start,
name,
description
)
{code}
{{use_gpus}} (optional)
BOOLEAN, *default*: FALSE (i.e., CPU). Determines whether GPUs are to be used
for training the neural network. Set to TRUE to use GPUs.
*Note*
This parameter must not conflict with how the distribution rules are set in the
preprocessor function. For example, if you set a distribution rule to use
certain segments on hosts that do not have GPUs attached, you will get an error
if you set {{use_gpus}} to TRUE.
Also, we have seen some memory related issues when segments share GPU
resources. For example, if you have 4 segments sharing 1 GPU, you may get
memory related errors. The recommended configuration is to have 1 GPU per
segment.
*Multi model fit()*
Same idea as above ^^^ for single model fit.
*Evaluate*
Same idea as above ^^^ for single model fit..
was:
Single fit()
{code}
madlib_keras_fit(
source_table,
model,
model_arch_table,
model_arch_id,
compile_params,
fit_params,
num_iterations,
use_gpus, -- changed definition
validation_table,
metrics_compute_frequency,
warm_start,
name,
description
)
{code}
{{use_gpus}} (optional)
BOOLEAN, *default*: FALSE (i.e., CPU). Determines whether GPUs are to be used
for training the neural network. Set to TRUE to use GPUs.
*Note*
This parameter must not conflict with how the distribution rules are set in the
preprocessor function. For example, if you set a distribution rule to use
certain segments on hosts that do not have GPUs attached, you will get an error
if you set {{use_gpus}} to TRUE.
Also, we have seen some memory related issues when segments share GPU
resources. For example, if you have 4 segments sharing 1 GPU, you may get
memory related errors. The recommended configuration is to have 1 GPU per
segment.
Multi model fit()
Same idea as above ^^^ for single model fit.
Evaluate
Same idea as above ^^^ for single model fit..
> DL: Fit and evaluate changes for asymmetric cluster config
> ----------------------------------------------------------
>
> Key: MADLIB-1393
> URL: https://issues.apache.org/jira/browse/MADLIB-1393
> Project: Apache MADlib
> Issue Type: New Feature
> Components: Deep Learning
> Reporter: Ekta Khanna
> Priority: Major
> Fix For: v1.17
>
>
> Single fit()
> {code}
> madlib_keras_fit(
> source_table,
> model,
> model_arch_table,
> model_arch_id,
> compile_params,
> fit_params,
> num_iterations,
> use_gpus, -- changed definition
> validation_table,
> metrics_compute_frequency,
> warm_start,
> name,
> description
> )
> {code}
> {{use_gpus}} (optional)
> BOOLEAN, *default*: FALSE (i.e., CPU). Determines whether GPUs are to be used
> for training the neural network. Set to TRUE to use GPUs.
> *Note*
> This parameter must not conflict with how the distribution rules are set in
> the preprocessor function. For example, if you set a distribution rule to
> use certain segments on hosts that do not have GPUs attached, you will get an
> error if you set {{use_gpus}} to TRUE.
> Also, we have seen some memory related issues when segments share GPU
> resources. For example, if you have 4 segments sharing 1 GPU, you may get
> memory related errors. The recommended configuration is to have 1 GPU per
> segment.
> *Multi model fit()*
> Same idea as above ^^^ for single model fit.
> *Evaluate*
> Same idea as above ^^^ for single model fit..
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
