♻️ Add download original dataset URL option

Signed-off-by: vankichi <[email protected]>

charts/vald-benchmark-operator/crds/valdbenchmarkjob.yaml

@@ -252,6 +252,7 @@ spec:
                     name:
                       type: string
                       enum:
+                        - original
                         - fashion-mnist
                     range:
                       type: object
@@ -262,6 +263,11 @@ spec:
                         start:
                           type: integer
                           minimum: 1
+                      required:
+                        - start
+                        - end
+                    url:
+                      type: string
                   required:
                     - name
                     - indexes

charts/vald-benchmark-operator/crds/valdbenchmarkscenario.yaml

@@ -84,6 +84,11 @@ spec:
                         start:
                           type: integer
                           minimum: 1
+                      required:
+                        - start
+                        - end
+                    url:
+                      type: string
                   required:
                     - name
                     - indexes

charts/vald-benchmark-operator/job-values.schema.json

@@ -305,7 +305,7 @@
         "name": {
           "type": "string",
           "description": "the name of dataset",
-          "enum": ["fashion-mnist"]
+          "enum": ["original", "fashion-mnist"]
         },
         "range": {
           "type": "object",
@@ -321,7 +321,12 @@
               "description": "start index number",
               "minimum": 1
             }
-          }
+          },
+          "required": ["start", "end"]
+        },
+        "url": {
+          "type": "string",
+          "description": "the dataset url which is used for executing benchmark job with user defined hdf5 file"
         }
       },
       "required": ["name", "indexes", "group", "range"]

charts/vald-benchmark-operator/scenario-values.schema.json

@@ -36,7 +36,12 @@
               "description": "start index number",
               "minimum": 1
             }
-          }
+          },
+          "required": ["start", "end"]
+        },
+        "url": {
+          "type": "string",
+          "description": "the dataset url which is used for executing benchmark job with user defined hdf5 file"
         }
       },
       "required": ["name", "indexes", "group", "range"]

charts/vald-benchmark-operator/schemas/job-values.yaml

@@ -27,7 +27,7 @@ target:
 # @schema {"name": "dataset", "type": "object", "required": ["name", "indexes", "group", "range"]}
 # dataset -- dataset information
 dataset:
-  # @schema {"name": "dataset.name", "type": "string", "enum": ["fashion-mnist"] }
+  # @schema {"name": "dataset.name", "type": "string", "enum": ["original", "fashion-mnist"] }
   # dataset.name -- the name of dataset
   name: "fashion-mnist"
   # @schema {"name": "dataset.indexes", "type": "integer", "minimum": 0}
@@ -36,7 +36,7 @@ dataset:
   # @schema {"name": "dataset.group", "type": "string", "minLength": 1}
   # dataset.group -- the hdf5 group name of dataset
   group: "test"
-  # @schema {"name": "dataset.range", "type": "object", "range": ["start", "port"]}
+  # @schema {"name": "dataset.range", "type": "object", "required": ["start", "end"]}
   # dataset.range -- the data range of indexes
   range:
     # @schema {"name": "dataset.range.start", "type": "integer", "minimum": 1}
@@ -45,6 +45,9 @@ dataset:
     # @schema {"name": "dataset.range.end", "type": "integer", "minimum": 1}
     # dataset.range.end -- end index number
     end: 1000
+  # @schema {"name": "dataset.url", "type": "string"}
+  # dataset.url -- the dataset url which is used for executing benchmark job with user defined hdf5 file
+  url: ""
 # @schema {"name": "dimension", "type": "integer", "minimum": 1}
 # dimension -- vector dimension
 dimension: 784

charts/vald-benchmark-operator/schemas/scenario-values.yaml

@@ -37,7 +37,7 @@ dataset:
   # @schema {"name": "dataset.group", "type": "string", "minLength": 1}
   # dataset.group -- the hdf5 group name of dataset
   group: "test"
-  # @schema {"name": "dataset.range", "type": "object", "range": ["start", "port"]}
+  # @schema {"name": "dataset.range", "type": "object", "required": ["start", "end"]}
   # dataset.range -- the data range of indexes
   range:
     # @schema {"name": "dataset.range.start", "type": "integer", "minimum": 1}
@@ -46,6 +46,9 @@ dataset:
     # @schema {"name": "dataset.range.end", "type": "integer", "minimum": 1}
     # dataset.range.end -- end index number
     end: 1000
+  # @schema {"name": "dataset.url", "type": "string"}
+  # dataset.url -- the dataset url which is used for executing benchmark job with user defined hdf5 file
+  url: ""
 
 # @schema {"name": "jobs", "type": "array", "items": {"type": "object"}}
 jobs:

internal/config/benchmark.go

@@ -62,10 +62,13 @@ type BenchmarkDataset struct {
 	Group   string                 `json:"group,omitempty"`
 	Indexes int                    `json:"indexes,omitempty"`
 	Range   *BenchmarkDatasetRange `json:"range,omitempty"`
+	URL     string                 `json:"url,omitempty"`
 }
 
 func (d *BenchmarkDataset) Bind() *BenchmarkDataset {
 	d.Name = GetActualValue(d.Name)
+	d.Group = GetActualValue(d.Group)
+	d.URL = GetActualValue(d.URL)
 	return d
 }
 

internal/test/data/hdf5/hdf5.go

@@ -28,7 +28,7 @@ import (
 )
 
 type Data interface {
-	Download() error
+	Download(url string) error
 	Read() error
 	GetName() DatasetName
 	GetPath() string
@@ -41,11 +41,14 @@ type Data interface {
 type DatasetName int
 
 const (
-	FashionMNIST784Euclidean DatasetName = iota
+	Original DatasetName = iota
+	FashionMNIST784Euclidean
 )
 
 func (d DatasetName) String() string {
 	switch d {
+	case Original:
+		return "original"
 	case FashionMNIST784Euclidean:
 		return "fashion-mnist-784-euc"
 	default:
@@ -109,8 +112,10 @@ func New(opts ...Option) (Data, error) {
 
 // Get downloads the hdf5 file.
 // https://github.com/erikbern/ann-benchmarks/#data-sets
-func (d *data) Download() error {
+func (d *data) Download(url string) error {
 	switch d.name {
+	case Original:
+		return downloadFile(url, d.path)
 	case FashionMNIST784Euclidean:
 		return downloadFile(FashionMNIST784EuclideanUrl.String(), d.path)
 	default:

internal/test/data/hdf5/hdf5_test.go

@@ -353,12 +353,16 @@ func Test_data_Download(t *testing.T) {
 		test      [][]float32
 		neighbors [][]int
 	}
+	type args struct {
+		url string
+	}
 	type want struct {
 		err error
 	}
 	type test struct {
 		name       string
 		fields     fields
+		args       args
 		want       want
 		checkFunc  func(want, error) error
 		beforeFunc func(*testing.T)
@@ -440,7 +444,7 @@ func Test_data_Download(t *testing.T) {
 				neighbors: test.fields.neighbors,
 			}
 
-			err := d.Download()
+			err := d.Download(test.args.url)
 			if err := checkFunc(test.want, err); err != nil {
 				tt.Errorf("error = %v", err)
 			}

internal/test/data/hdf5/option.go

@@ -31,6 +31,8 @@ var defaultOptions = []Option{
 func WithNameByString(n string) Option {
 	var name DatasetName
 	switch n {
+	case Original.String():
+		name = Original
 	case FashionMNIST784Euclidean.String():
 		name = FashionMNIST784Euclidean
 	}
@@ -40,6 +42,8 @@ func WithNameByString(n string) Option {
 func WithName(dn DatasetName) Option {
 	return func(d *data) error {
 		switch dn {
+		case Original:
+			d.name = dn
 		case FashionMNIST784Euclidean:
 			d.name = dn
 		default:

k8s/tools/benchmark/operator/crds/valdbenchmarkjob.yaml

@@ -252,6 +252,7 @@ spec:
                     name:
                       type: string
                       enum:
+                        - original
                         - fashion-mnist
                     range:
                       type: object
@@ -262,6 +263,11 @@ spec:
                         start:
                           type: integer
                           minimum: 1
+                      required:
+                        - start
+                        - end
+                    url:
+                      type: string
                   required:
                     - name
                     - indexes

k8s/tools/benchmark/operator/crds/valdbenchmarkscenario.yaml

@@ -84,6 +84,11 @@ spec:
                         start:
                           type: integer
                           minimum: 1
+                      required:
+                        - start
+                        - end
+                    url:
+                      type: string
                   required:
                     - name
                     - indexes

pkg/tools/benchmark/job/service/job.go

@@ -191,7 +191,7 @@ func New(opts ...Option) (Job, error) {
 
 func (j *job) PreStart(ctx context.Context) error {
 	log.Infof("[benchmark job] start download dataset of %s", j.hdf5.GetName().String())
-	if err := j.hdf5.Download(); err != nil {
+	if err := j.hdf5.Download(j.dataset.URL); err != nil {
 		return err
 	}
 	log.Infof("[benchmark job] success download dataset of %s", j.hdf5.GetName().String())

pkg/tools/benchmark/job/service/option.go

@@ -141,12 +141,15 @@ func WithHdf5(d hdf5.Data) Option {
 	}
 }
 
-// WithDataset sets the config.BenchmarkDataset including benchmakr dataset name, group name of hdf5.Data, the number of index, start range and end range.
+// WithDataset sets the config.BenchmarkDataset including benchmark dataset name, group name of hdf5.Data, the number of index, start range and end range, and original URL which is used for download user defined hdf5.
 func WithDataset(d *config.BenchmarkDataset) Option {
 	return func(j *job) error {
 		if d == nil {
 			return errors.NewErrInvalidOption("dataset", d)
 		}
+		if d.Name == hdf5.Original.String() && len(d.URL) == 0 {
+			return errors.NewErrInvalidOption("dataset", d)
+		}
 		j.dataset = d
 		return nil
 	}