Skip to content

Training dataset resource bug #86

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 9 commits into from
Apr 26, 2019
Merged

Training dataset resource bug #86

Show file tree
Hide file tree
Changes from 4 commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
193040a
Default spark compute resource to spark workloads
vishalbollu Apr 23, 2019
9122a03
Add spark compute to model config and remove default
vishalbollu Apr 25, 2019
5f93120
Merge branch 'master' into training-dataset-resource-bug
vishalbollu Apr 25, 2019
3089938
Update model config documentation
vishalbollu Apr 25, 2019
38b57b4
Refactoring variable and deleting unnecessary function
vishalbollu Apr 26, 2019
d023e84
Remove duplicate tag field in doc
vishalbollu Apr 26, 2019
758d88f
Merge branch 'master' into training-dataset-resource-bug
vishalbollu Apr 26, 2019
31b8750
Declare spark compute validation object const as ptr
vishalbollu Apr 26, 2019
087e2e4
Update models.md
deliahu Apr 26, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 14 additions & 1 deletion docs/applications/resources/models.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -41,11 +41,24 @@ Train custom TensorFlow models at scale.
start_delay_secs: <int> # start evaluating after waiting for this many seconds (default: 120)
throttle_secs: <int> # do not re-evaluate unless the last evaluation was started at least this many seconds ago (default: 600)

compute:
compute: # Resources for training and evaluations steps
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe mention it's the TensorFlow job? Like Resources for training and evaluations steps (TensorFlow)? Similarly # Resources for constructing training dataset (Spark)? We get Omer's stamp of approval on this

Copy link
Contributor Author

@vishalbollu vishalbollu Apr 26, 2019 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@ospillinger thoughts on the doc comments?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

cpu: <string> # CPU request (default: Null)
mem: <string> # memory request (default: Null)
gpu: <string> # GPU request (default: Null)

dataset_compute: # Resources for constructing training dataset
executors: <int> # number of spark executors (default: 1)
driver_cpu: <string> # CPU request for spark driver (default: 1)
driver_mem: <string> # memory request for spark driver (default: 500Mi)
driver_mem_overhead: <string> # off-heap (non-JVM) memory allocated to the driver (overrides mem_overhead_factor) (default: min[driver_mem * 0.4, 384Mi])
executor_cpu: <string> # CPU request for each spark executor (default: 1)
executor_mem: <string> # memory request for each spark executor (default: 500Mi)
executor_mem_overhead: <string> # off-heap (non-JVM) memory allocated to each executor (overrides mem_overhead_factor) (default: min[executor_mem * 0.4, 384Mi])
mem_overhead_factor: <float> # the proportion of driver_mem/executor_mem which will be additionally allocated for off-heap (non-JVM) memory (default: 0.4)
tags:
<string>: <scalar> # arbitrary key/value pairs to attach to the resource (optional)
...

tags:
<string>: <scalar> # arbitrary key/value pairs to attach to the resource (optional)
...
Expand Down
2 changes: 1 addition & 1 deletion pkg/operator/api/userconfig/aggregates.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,7 +51,7 @@ var aggregateValidation = &cr.StructValidation{
},
},
inputValuesFieldValidation,
sparkComputeFieldValidation,
sparkComputeFieldValidation("Compute"),
tagsFieldValidation,
typeFieldValidation,
},
Expand Down
134 changes: 68 additions & 66 deletions pkg/operator/api/userconfig/compute.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,82 +45,84 @@ type SparkCompute struct {
MemOverheadFactor *float64 `json:"mem_overhead_factor" yaml:"mem_overhead_factor"`
}

var sparkComputeFieldValidation = &cr.StructFieldValidation{
StructField: "Compute",
StructValidation: &cr.StructValidation{
StructFieldValidations: []*cr.StructFieldValidation{
{
StructField: "Executors",
Int32Validation: &cr.Int32Validation{
Default: 1,
GreaterThan: pointer.Int32(0),
func sparkComputeFieldValidation(fieldName string) *cr.StructFieldValidation {
return &cr.StructFieldValidation{
StructField: fieldName,
StructValidation: &cr.StructValidation{
StructFieldValidations: []*cr.StructFieldValidation{
{
StructField: "Executors",
Int32Validation: &cr.Int32Validation{
Default: 1,
GreaterThan: pointer.Int32(0),
},
},
},
{
StructField: "DriverCPU",
StringValidation: &cr.StringValidation{
Default: "1",
{
StructField: "DriverCPU",
StringValidation: &cr.StringValidation{
Default: "1",
},
Parser: QuantityParser(&QuantityValidation{
Min: k8sresource.MustParse("1"),
}),
},
Parser: QuantityParser(&QuantityValidation{
Min: k8sresource.MustParse("1"),
}),
},
{
StructField: "ExecutorCPU",
StringValidation: &cr.StringValidation{
Default: "1",
{
StructField: "ExecutorCPU",
StringValidation: &cr.StringValidation{
Default: "1",
},
Parser: QuantityParser(&QuantityValidation{
Min: k8sresource.MustParse("1"),
Int: true,
}),
},
Parser: QuantityParser(&QuantityValidation{
Min: k8sresource.MustParse("1"),
Int: true,
}),
},
{
StructField: "DriverMem",
StringValidation: &cr.StringValidation{
Default: "500Mi",
{
StructField: "DriverMem",
StringValidation: &cr.StringValidation{
Default: "500Mi",
},
Parser: QuantityParser(&QuantityValidation{
Min: k8sresource.MustParse("500Mi"),
}),
},
Parser: QuantityParser(&QuantityValidation{
Min: k8sresource.MustParse("500Mi"),
}),
},
{
StructField: "ExecutorMem",
StringValidation: &cr.StringValidation{
Default: "500Mi",
{
StructField: "ExecutorMem",
StringValidation: &cr.StringValidation{
Default: "500Mi",
},
Parser: QuantityParser(&QuantityValidation{
Min: k8sresource.MustParse("500Mi"),
}),
},
Parser: QuantityParser(&QuantityValidation{
Min: k8sresource.MustParse("500Mi"),
}),
},
{
StructField: "DriverMemOverhead",
StringPtrValidation: &cr.StringPtrValidation{
Default: nil, // min(DriverMem * 0.4, 384Mi)
{
StructField: "DriverMemOverhead",
StringPtrValidation: &cr.StringPtrValidation{
Default: nil, // min(DriverMem * 0.4, 384Mi)
},
Parser: QuantityParser(&QuantityValidation{
Min: k8sresource.MustParse("0"),
}),
},
Parser: QuantityParser(&QuantityValidation{
Min: k8sresource.MustParse("0"),
}),
},
{
StructField: "ExecutorMemOverhead",
StringPtrValidation: &cr.StringPtrValidation{
Default: nil, // min(ExecutorMem * 0.4, 384Mi)
{
StructField: "ExecutorMemOverhead",
StringPtrValidation: &cr.StringPtrValidation{
Default: nil, // min(ExecutorMem * 0.4, 384Mi)
},
Parser: QuantityParser(&QuantityValidation{
Min: k8sresource.MustParse("0"),
}),
},
Parser: QuantityParser(&QuantityValidation{
Min: k8sresource.MustParse("0"),
}),
},
{
StructField: "MemOverheadFactor",
Float64PtrValidation: &cr.Float64PtrValidation{
Default: nil, // set to 0.4 by Spark
GreaterThanOrEqualTo: pointer.Float64(0),
LessThan: pointer.Float64(1),
{
StructField: "MemOverheadFactor",
Float64PtrValidation: &cr.Float64PtrValidation{
Default: nil, // set to 0.4 by Spark
GreaterThanOrEqualTo: pointer.Float64(0),
LessThan: pointer.Float64(1),
},
},
},
},
},
}
}

func (sparkCompute *SparkCompute) ID() string {
Expand Down
2 changes: 2 additions & 0 deletions pkg/operator/api/userconfig/models.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -42,6 +42,7 @@ type Model struct {
Training *ModelTraining `json:"training" yaml:"training"`
Evaluation *ModelEvaluation `json:"evaluation" yaml:"evaluation"`
Compute *TFCompute `json:"compute" yaml:"compute"`
DatasetCompute *SparkCompute `json:"dataset_compute" yaml:"dataset_compute"`
Tags Tags `json:"tags" yaml:"tags"`
}

Expand Down Expand Up @@ -127,6 +128,7 @@ var modelValidation = &cr.StructValidation{
StructValidation: modelEvaluationValidation,
},
tfComputeFieldValidation,
sparkComputeFieldValidation("DatasetCompute"),
tagsFieldValidation,
typeFieldValidation,
},
Expand Down
9 changes: 9 additions & 0 deletions pkg/operator/api/userconfig/quantity.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,6 +30,15 @@ type Quantity struct {
UserString string
}

func MustNewQuantity(str string) Quantity {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this used?

k8sQuantity := k8sresource.MustParse(str)

return Quantity{
Quantity: k8sQuantity,
UserString: str,
}
}

type QuantityValidation struct {
Min k8sresource.Quantity
Int bool
Expand Down
6 changes: 3 additions & 3 deletions pkg/operator/api/userconfig/raw_columns.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -96,7 +96,7 @@ var rawIntColumnFieldValidations = []*cr.StructFieldValidation{
AllowNull: true,
},
},
sparkComputeFieldValidation,
sparkComputeFieldValidation("Compute"),
tagsFieldValidation,
typeFieldValidation,
}
Expand Down Expand Up @@ -145,7 +145,7 @@ var rawFloatColumnFieldValidations = []*cr.StructFieldValidation{
AllowNull: true,
},
},
sparkComputeFieldValidation,
sparkComputeFieldValidation("Compute"),
tagsFieldValidation,
typeFieldValidation,
}
Expand Down Expand Up @@ -182,7 +182,7 @@ var rawStringColumnFieldValidations = []*cr.StructFieldValidation{
AllowNull: true,
},
},
sparkComputeFieldValidation,
sparkComputeFieldValidation("Compute"),
tagsFieldValidation,
typeFieldValidation,
}
Expand Down
2 changes: 1 addition & 1 deletion pkg/operator/api/userconfig/transformed_columns.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,7 +51,7 @@ var transformedColumnValidation = &cr.StructValidation{
},
},
inputValuesFieldValidation,
sparkComputeFieldValidation,
sparkComputeFieldValidation("Compute"),
tagsFieldValidation,
typeFieldValidation,
},
Expand Down
7 changes: 1 addition & 6 deletions pkg/operator/workloads/data_job.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -251,12 +251,7 @@ func dataWorkloadSpecs(ctx *context.Context) ([]*WorkloadSpec, error) {
}
trainingDatasets = append(trainingDatasets, modelName)
trainingDatasetIDs.Add(dataset.GetID())
dependencyIDs := ctx.AllComputedResourceDependencies(dataset.GetID())
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think we still need to append the transformedColumn.Computes since transforming the data happens in the same step as preparing the dataset

for _, transformedColumn := range ctx.TransformedColumns {
if _, ok := dependencyIDs[transformedColumn.ID]; ok {
allComputes = append(allComputes, transformedColumn.Compute)
}
}
allComputes = append(allComputes, model.DatasetCompute)
}

resourceIDSet := strset.Union(rawColumnIDs, aggregateIDs, transformedColumnIDs, trainingDatasetIDs)
Expand Down