systemml-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "LI Guobao (JIRA)" <>
Subject [jira] [Commented] (SYSTEMML-2299) API design of the paramserv function
Date Wed, 11 Jul 2018 21:43:01 GMT


LI Guobao commented on SYSTEMML-2299:

[~mboehm7], in fact, until now, it seems that the arguments "val_features" and "val_labels"
have not been used inside paramserv function. For me, they are leveraged to calculate the
model precision and we use a UDF to do it but not inside our paramserv function. So do they
have some other utilities?

> API design of the paramserv function
> ------------------------------------
>                 Key: SYSTEMML-2299
>                 URL:
>             Project: SystemML
>          Issue Type: Sub-task
>            Reporter: LI Guobao
>            Assignee: LI Guobao
>            Priority: Major
>             Fix For: SystemML 1.2
> The objective of “paramserv” built-in function is to update an initial or existing
model with configuration. An initial function signature would be: 
> {code:java}
> model'=paramserv(model=paramsList, features=X, labels=Y, val_features=X_val, val_labels=Y_val,
upd="fun1", agg="fun2", mode="LOCAL", utype="BSP", freq="BATCH", epochs=100, batchsize=64,
k=7, scheme="disjoint_contiguous", hyperparams=params, checkpointing="NONE"){code}
> We are interested in providing the model (which will be a struct-like data structure
consisting of the weights, the biases and the hyperparameters), the training features and
labels, the validation features and labels, the batch update function (i.e., gradient calculation
func), the update strategy (e.g. sync, async, hogwild!, stale-synchronous), the update frequency
(e.g. epoch or mini-batch), the gradient aggregation function, the number of epoch, the batch
size, the degree of parallelism, the data partition scheme, a list of additional hyper parameters,
as well as the checkpointing strategy. And the function will return a trained model in struct
> *Inputs*:
>  * model <list>: a list consisting of the weight and bias matrices
>  * features <matrix>: training features matrix
>  * labels <matrix>: training label matrix
>  * val_features <matrix>: validation features matrix
>  * val_labels <matrix>: validation label matrix
>  * upd <string>: the name of gradient calculation function
>  * agg <string>: the name of gradient aggregation function
>  * mode <string> (options: LOCAL, REMOTE_SPARK): the execution backend where the
parameter is executed
>  * utype <string> (options: BSP, ASP, SSP): the updating mode
>  * freq <string> [optional] (default: BATCH) (options: EPOCH, BATCH) : the frequence
of updates
>  * epochs <integer>: the number of epoch
>  * batchsize <integer> [optional] (default: 64): the size of batch, if the update
frequence is "EPOCH", this argument will be ignored
>  * k <integer> [optional] (default: number of vcores, otherwise vcores / 2 if using
openblas): the degree of parallelism
>  * scheme <string> [optional] (default: disjoint_contiguous) (options: disjoint_contiguous,
disjoint_round_robin, disjoint_random, overlap_reshuffle): the scheme of data partition, i.e.,
how the data is distributed across workers
>  * hyperparams <list> [optional]: a list consisting of the additional hyper parameters,
e.g., learning rate, momentum
>  * checkpointing <string>[optional] (default: NONE) (options: NONE, EPOCH, EPOCH10)
: the checkpoint strategy, we could set a checkpoint for each epoch or each 10 epochs 
> *Output*:
>  * model' <list>: a list consisting of the updated weight and bias matrices

This message was sent by Atlassian JIRA

View raw message