Update tf function imports

Author: risenW

danfojs/src/core/frame.js

@@ -12,7 +12,7 @@
 * limitations under the License.
 */
 
-import * as tf from "@tensorflow/tfjs";
+import { tensor } from "@tensorflow/tfjs";
 import Ndframe from "./generic";
 import { Series } from "./series";
 import { Utils } from "./utils";
@@ -183,7 +183,7 @@ export class DataFrame extends Ndframe {
           index: new_index
         });
       } else {
-        this.row_data_tensor = tf.tensor(new_data);
+        this.row_data_tensor = tensor(new_data);
         this.data = new_data;
         this.__set_index(new_index);
       }
@@ -916,7 +916,7 @@ export class DataFrame extends Ndframe {
       }
 
       values.map((arr) => {
-        let temp_sum = tf.tensor(arr).sum().arraySync();
+        let temp_sum = tensor(arr).sum().arraySync();
         val_sums.push(Number(temp_sum.toFixed(5)));
       });
 
@@ -940,7 +940,7 @@ export class DataFrame extends Ndframe {
   abs() {
     let data = this.values;
 
-    let tensor_data = tf.tensor(data);
+    let tensor_data = tensor(data);
     let abs_data = tensor_data.abs().arraySync();
     let df = new DataFrame(utils.__round(abs_data, 2, false), {
       columns: this.column_names,
@@ -1406,7 +1406,7 @@ export class DataFrame extends Ndframe {
       }
 
       for (let i = 0; i < df_data.length; i++) {
-        let value = tf.tensor(df_data[i]);
+        let value = tensor(df_data[i]);
         let callable_data;
         try {
           callable_data = callable(value).arraySync();
@@ -1659,18 +1659,18 @@ export class DataFrame extends Ndframe {
               `Shape Error: Operands could not be broadcast together with shapes ${this.shape} and ${val.values.length}.`
             );
           }
-          other = tf.tensor(val.values);
+          other = tensor(val.values);
         } else {
           if (val.values.length != this.shape[1]) {
             throw Error(
               `Shape Error: Operands could not be broadcast together with shapes ${this.shape} and ${val.values.length}.`
             );
           }
-          other = tf.tensor(val.values);
+          other = tensor(val.values);
         }
       } else if (Array.isArray(val)) {
         //Array of Array
-        other = tf.tensor(val);
+        other = tensor(val);
       } else {
         //DataFrame
         other = val.row_data_tensor;
@@ -1679,22 +1679,22 @@ export class DataFrame extends Ndframe {
 
     switch (logical_type) {
     case "lt":
-      int_vals = tf.tensor(this.values).less(other).arraySync();
+      int_vals = tensor(this.values).less(other).arraySync();
       break;
     case "gt":
-      int_vals = tf.tensor(this.values).greater(other).arraySync();
+      int_vals = tensor(this.values).greater(other).arraySync();
       break;
     case "le":
-      int_vals = tf.tensor(this.values).lessEqual(other).arraySync();
+      int_vals = tensor(this.values).lessEqual(other).arraySync();
       break;
     case "ge":
-      int_vals = tf.tensor(this.values).greaterEqual(other).arraySync();
+      int_vals = tensor(this.values).greaterEqual(other).arraySync();
       break;
     case "ne":
-      int_vals = tf.tensor(this.values).notEqual(other).arraySync();
+      int_vals = tensor(this.values).notEqual(other).arraySync();
       break;
     case "eq":
-      int_vals = tf.tensor(this.values).equal(other).arraySync();
+      int_vals = tensor(this.values).equal(other).arraySync();
       break;
     }
     let bool_vals = utils.__map_int_to_bool(int_vals, 2);
@@ -1754,7 +1754,7 @@ export class DataFrame extends Ndframe {
 
         this_tensor = tensors[0].row_data_tensor; //tensorflow uses 1 for rows axis and 0 for column axis
         if (tensors[1].series) {
-          other_tensor = tf.tensor(tensors[1].values, [
+          other_tensor = tensor(tensors[1].values, [
             1,
             tensors[1].values.length
           ]);
@@ -1771,7 +1771,7 @@ export class DataFrame extends Ndframe {
 
         this_tensor = tensors[0].row_data_tensor;
         if (tensors[1].series) {
-          other_tensor = tf.tensor(tensors[1].values, [
+          other_tensor = tensor(tensors[1].values, [
             tensors[1].values.length,
             1
           ]);
@@ -2153,7 +2153,7 @@ export class DataFrame extends Ndframe {
       let val = sorted_val[row_i];
       let index = null;
 
-      if (duplicate_obj.hasOwnProperty(val)) {
+      if (val in duplicate_obj) {
         index = duplicate_obj[val]["index"][0];
         duplicate_obj[val]["index"].splice(0, 1);
       } else {

danfojs/src/core/generic.js

@@ -13,7 +13,7 @@
  *
 */
 
-import * as tf from "@tensorflow/tfjs";
+import { tensor, Tensor } from "@tensorflow/tfjs";
 import { table } from "table";
 import { Utils } from "./utils";
 import { Configs } from "../config/config";
@@ -38,7 +38,7 @@ export default class NDframe {
   constructor(data, kwargs = {}) {
     this.kwargs = kwargs;
 
-    if (data instanceof tf.Tensor) {
+    if (data instanceof Tensor) {
       data = data.arraySync();
     }
 
@@ -71,15 +71,15 @@ export default class NDframe {
    */
   _read_array(data) {
     this.data = utils.__replace_undefined_with_NaN(data, this.series);
-    this.row_data_tensor = tf.tensor(this.data);
+    this.row_data_tensor = tensor(this.data);
 
     if (this.series) {
       this.col_data = [ this.values ];
     } else {
       this.col_data = utils.__get_col_values(this.data);
     }
 
-    this.col_data_tensor = tf.tensor(this.col_data); //data saved as 2D column tensors
+    this.col_data_tensor = tensor(this.col_data); //data saved as 2D column tensors
 
     if ("index" in this.kwargs) {
       this.__set_index(this.kwargs["index"]);
@@ -133,7 +133,7 @@ export default class NDframe {
       });
 
       this.data = utils.__replace_undefined_with_NaN(data_arr, this.series); //Defualt array data in row format
-      this.row_data_tensor = tf.tensor(this.data); //data saved as row tensors
+      this.row_data_tensor = tensor(this.data); //data saved as row tensors
       this.kwargs["columns"] = Object.keys(Object.values(data)[0]); //get names of the column from the first entry
 
       if (this.series) {
@@ -142,7 +142,7 @@ export default class NDframe {
         this.col_data = utils.__get_col_values(this.data);
       }
 
-      this.col_data_tensor = tf.tensor(this.col_data); //data saved as 2D column tensors
+      this.col_data_tensor = tensor(this.col_data); //data saved as 2D column tensors
 
       if ("index" in this.kwargs) {
         this.__set_index(this.kwargs["index"]);

danfojs/src/core/series.js

@@ -14,7 +14,7 @@
 */
 
 
-import * as tf from "@tensorflow/tfjs";
+import { tensor, round } from "@tensorflow/tfjs";
 import { variance, std } from 'mathjs';
 import { Utils } from "./utils";
 import { Str } from "./strings";
@@ -55,7 +55,7 @@ export class Series extends NDframe {
     * @returns {1D Tensor}
     */
   get tensor() {
-    return tf.tensor(this.values).asType(this.dtypes[0]);
+    return tensor(this.values).asType(this.dtypes[0]);
   }
 
 
@@ -250,7 +250,7 @@ export class Series extends NDframe {
   mean() {
     utils._throw_str_dtype_error(this, 'mean');
     let values = utils._remove_nans(this.values);
-    let mean = tf.tensor(values).mean().arraySync();
+    let mean = tensor(values).mean().arraySync();
     return mean;
   }
 
@@ -382,7 +382,7 @@ export class Series extends NDframe {
   round(dp) {
     if (utils.__is_undefined(dp)) {
       //use tensorflow round function to roound to the nearest whole number
-      let result = tf.round(this.row_data_tensor).arraySync();
+      let result = round(this.row_data_tensor).arraySync();
       return new Series(result, { columns: this.column_names, index: this.index });
 
     } else {

danfojs/src/core/utils.js

@@ -1,5 +1,4 @@
-import * as tf from "@tensorflow/tfjs";
-import '@tensorflow/tfjs-backend-cpu';
+import { tensor, linspace } from "@tensorflow/tfjs";
 import { Configs } from "../config/config";
 
 const config = new Configs();
@@ -113,7 +112,7 @@ export class Utils {
 
   //generate integers between two set of numbers
   __range(start, end) {
-    let value = tf.linspace(start, end, end - start + 1).arraySync();
+    let value = linspace(start, end, end - start + 1).arraySync();
     return value;
   }
 
@@ -689,7 +688,7 @@ export class Utils {
     let rslt_obj = {};
 
     arr.forEach((val, index) => {
-      if (temp_obj.hasOwnProperty(val)) {
+      if (val in temp_obj) {
         temp_obj[val]["count"] += 1;
         temp_obj[val]["index"].push(index);
       } else {

danfojs/src/index.js

@@ -15,4 +15,4 @@ export { NDframe };
 export { Str } from "./core/strings";
 export { Utils } from "./core/utils";
 
-export const _version = "0.2.1";
+export const _version = "0.2.3";

danfojs/src/io/reader.js

@@ -1,4 +1,4 @@
-import * as tf from "@tensorflow/tfjs";
+import { data } from "@tensorflow/tfjs";
 import fetch from "node-fetch";
 import XLSX from "xlsx";
 import { open, Dataset, isDataset } from "frictionless.js";
@@ -27,11 +27,12 @@ export const read_csv = async (source, chunk) => {
     )
   ) {
     //probabily a relative path, append file:// to it
+    // eslint-disable-next-line no-undef
     source = `file://${process.cwd()}/${source}`;
   }
 
   let data = [];
-  const csvDataset = tf.data.csv(source);
+  const csvDataset = data.csv(source);
   const column_names = await csvDataset.columnNames();
   const sample = csvDataset.take(chunk);
   await sample.forEachAsync((row) => data.push(Object.values(row)));

danfojs/src/preprocessing/scalers.js

@@ -1,4 +1,4 @@
-import * as tf from "@tensorflow/tfjs";
+import { tensor, moments } from "@tensorflow/tfjs";
 import { Series } from "../core/series";
 import { DataFrame } from "../core/frame";
 import { Utils } from "../core/utils";
@@ -14,12 +14,12 @@ export class MinMaxScaler {
   fit(data) {
     let tensor_data = null;
     if (Array.isArray(data)) {
-      tensor_data = tf.tensor(data);
+      tensor_data = tensor(data);
     } else if (data instanceof DataFrame || data instanceof Series) {
       if (data.dtypes.includes("string")) {
         throw Error("Dtype Error: Cannot perform operation on string dtypes");
       }
-      tensor_data = tf.tensor(data.values);
+      tensor_data = tensor(data.values);
     } else {
       throw new Error("data must either be an Array, DataFrame or Series");
     }
@@ -49,14 +49,14 @@ export class MinMaxScaler {
       if (data.dtypes.includes("string")) {
         throw Error("Dtype Error: Cannot perform operation on string dtypes");
       }
-      let tensor_data = tf.tensor(data.values);
+      let tensor_data = tensor(data.values);
       let output_data = tensor_data
         .sub(this.min)
         .div(this.max.sub(this.min))
         .arraySync();
       return new Series(output_data);
     } else if (Array.isArray(data)) {
-      let tensor_data = tf.tensor(data);
+      let tensor_data = tensor(data);
       let output_data = tensor_data
         .sub(this.min)
         .div(this.max.sub(this.min))
@@ -70,7 +70,7 @@ export class MinMaxScaler {
       if (data.dtypes.includes("string")) {
         throw Error("Dtype Error: Cannot perform operation on string dtypes");
       }
-      let tensor_data = tf.tensor(data.values);
+      let tensor_data = tensor(data.values);
       let output_data = tensor_data
         .sub(this.min)
         .div(this.max.sub(this.min))
@@ -91,17 +91,17 @@ export class StandardScaler {
   fit(data) {
     let tensor_data = null;
     if (Array.isArray(data)) {
-      tensor_data = tf.tensor(data);
+      tensor_data = tensor(data);
     } else if (data instanceof DataFrame || data instanceof Series) {
       if (data.dtypes.includes("string")) {
         throw Error("Dtype Error: Cannot perform operation on string dtypes");
       }
-      tensor_data = tf.tensor(data.values);
+      tensor_data = tensor(data.values);
     } else {
       throw new Error("data must either be an Array, DataFrame or Series");
     }
 
-    this.std = tf.moments(tensor_data, 0).variance.sqrt();
+    this.std = moments(tensor_data, 0).variance.sqrt();
     this.mean = tensor_data.mean(0);
     let output_data = tensor_data.sub(this.mean).div(this.std).arraySync();
 
@@ -121,11 +121,11 @@ export class StandardScaler {
       if (data.dtypes.includes("string")) {
         throw Error("Dtype Error: Cannot perform operation on string dtypes");
       }
-      let tensor_data = tf.tensor(data.values);
+      let tensor_data = tensor(data.values);
       let output_data = tensor_data.sub(this.mean).div(this.std).arraySync();
       return new Series(output_data);
     } else if (Array.isArray(data)) {
-      let tensor_data = tf.tensor(data);
+      let tensor_data = tensor(data);
       let output_data = tensor_data.sub(this.mean).div(this.std).arraySync();
       if (utils.__is_1D_array(data)) {
         return new Series(output_data);
@@ -136,7 +136,7 @@ export class StandardScaler {
       if (data.dtypes.includes("string")) {
         throw Error("Dtype Error: Cannot perform operation on string dtypes");
       }
-      let tensor_data = tf.tensor(data.values);
+      let tensor_data = tensor(data.values);
       let output_data = tensor_data.sub(this.mean).div(this.std).arraySync();
       return new DataFrame(output_data);
     } else {
@@ -245,21 +245,21 @@ export class StandardScaler {
 //         let tensor_data = null
 //         let isTensor = false;
 //         if(Array.isArray(data)){
-//             tensor_data = tf.tensor(data)
+//             tensor_data = tensor(data)
 //         }
 //         else if((data instanceof DataFrame)){
-//             tensor_data = tf.tensor(data.values)
+//             tensor_data = tensor(data.values)
 //             isTensor = true;
 //         }
 //         else if((data instanceof Series)){
-//             tensor_data = tf.tensor(data.values)
+//             tensor_data = tensor(data.values)
 //         }
 //         else{
 //             throw new Error("data must either be an Array, DataFrame or Series")
 //         }
 
 //         let [q1, q3, median] = this.quantile(data,isTensor)
-//         let q3_tensor = tf.tensor(q3)
+//         let q3_tensor = tensor(q3)
 //         let output_data =  tensor_data.sub(median).div(q3_tensor.sub(q1)).arraySync()
 
 //         return output_data;

package.json

@@ -88,5 +88,6 @@
       "lcov",
       "text"
     ]
-  }
+  },
+  "sideEffects": false
 }