fix Tensorflow -> TensorFlow (#2783)

Author: sneaxiy

doc/design/alps_submitter.md

@@ -158,7 +158,7 @@ LABEL class
 ```
 
 ### Semantic Analyze
-Feature Expressions except for Tensorflow Feature Column API should raise an error.
+Feature Expressions except for TensorFlow Feature Column API should raise an error.
 ```sql
 /* Not supported */
 select * from kaggle_credit_fraud_training_data

doc/design/codegen_couler_use_ir.md

@@ -10,12 +10,12 @@ In this design, we'll only have `codegen_couler.go` to generate programs to run
 
 For example, to run the training steps, we intend to use couler API call: `couler.{xgboost,tensorflow,elasticdl}.train(model_def, data)` to generate a training step. The `train` function must take lots of parameters to run the training job as the SQLFlow statement describes, see [here](https://github.com/sql-machine-learning/sqlflow/blob/develop/python/runtime/tensorflow/train.py#L52) as an example.
 
-To implement the single `codegen_couler.go` to support generate code that can run either Tensorflow/XGBoost/ElasticDL/ALPS programs, we have below two choices:
+To implement the single `codegen_couler.go` to support generate code that can run either TensorFlow/XGBoost/ElasticDL/ALPS programs, we have below two choices:
 
 1. `couler.{tensorflow/xgboost/elasticdl}.train` have different arguments defined, so we can do:
 
  ```go
- if ir.ModelType == "Tensorflow":
+ if ir.ModelType == "TensorFlow":
  tfFiller := TFFiller{
  Estimator: generateTFEstimatorCode(ir),
  FeatureColumns: generateFeatureColumnsCode(ir),
@@ -55,6 +55,6 @@ To implement the single `codegen_couler.go` to support generate code that can ru
 
 We intend to use the solution **No.2** described above for these reasons.
 
-1. If a data scientist needs to add a new type of engine (SVM, SKLearn, PyTorch, etc.) other than Tensorflow/XGBoost/ElasticDL, he/she can use python only to define a `couler.{new_engine}.train` function without modifying the Go code in SQLFlow.
+1. If a data scientist needs to add a new type of engine (SVM, SKLearn, PyTorch, etc.) other than TensorFlow/XGBoost/ElasticDL, he/she can use python only to define a `couler.{new_engine}.train` function without modifying the Go code in SQLFlow.
 1. `codegen_couler.go` have less code.
 1. All submitter unit tests can run in couler.

doc/design/couler_sqlflow.md

@@ -104,7 +104,7 @@ Note: You can check more details about the IR definition from [ir.go](/pkg/ir/ir
 
 ### SQLFLow Submitter Python Module
 
-An SQLFlow submitter Python module `runtime.{tensorflow,xgboost,elasticdl}.train` accepts an SQLFlow IR with protobuf text format, and then submit a Tensorflow, XGBoost or ElasticDL training job, we can call it like:
+An SQLFlow submitter Python module `runtime.{tensorflow,xgboost,elasticdl}.train` accepts an SQLFlow IR with protobuf text format, and then submit a TensorFlow, XGBoost or ElasticDL training job, we can call it like:
 
 ``` bash
 cat ir.proto_text | python -m runtime.xgboost.train

doc/design/diag_attribute_error.md

@@ -31,7 +31,7 @@ this can active three advantages at least:
 
 1. Early testing, we can do early testing before running the job; users can wait less time and cluster save resources.
 2. More accurate diagnostic message.
-3. Model developers do not have to involve dependencies other than Keras or Tensorflow.
+3. Model developers do not have to involve dependencies other than Keras or TensorFlow.
 
 ## Design
 

doc/design/model_zoo.md

@@ -101,7 +101,7 @@ $ sqlflow delete repo model_image:v0.1
 
 For a SELECT program using the SQLFlow syntax extension, the SQLFlow server converts it into a [workflow](workflow.md) and submit the workflow job to a workflow engine like Argo/Tekton on Kubernetes. Each step in the workflow is one SQL statement.
 
-By default, we use a default Docker image to run the training, predicting or explaining job. The default Docker image contains pre-made Tensorflow estimator models, Keras models defined in [sqlflow_models repo](https://github.com/sql-machine-learning/models) and XGBoost. To use a custom model repo Docker image, write SQL statements mentioned above:
+By default, we use a default Docker image to run the training, predicting or explaining job. The default Docker image contains pre-made TensorFlow estimator models, Keras models defined in [sqlflow_models repo](https://github.com/sql-machine-learning/models) and XGBoost. To use a custom model repo Docker image, write SQL statements mentioned above:
 
 ```sql
 SELECT * FROM employee WHERE onboard_year < 2019

doc/model_parameter.md

@@ -206,7 +206,7 @@ TBD
 
 TBD
 
-## Tensorflow Parameters
+## TensorFlow Parameters
 
 ### TRAIN
 

doc/sqlflow.org_cn/index.md

@@ -12,8 +12,8 @@
  1. [模型解释]()
 1. [模型列表](model_list.cn.md)
  1. [有监督学习]()
- 1. [Tensorflow Estimator 模型]()
- 1. [Tensorflow Keras 模型]()
+ 1. [TensorFlow Estimator 模型]()
+ 1. [TensorFlow Keras 模型]()
  1. [XGBoost 模型]()
  1. [无监督学习]()
 1. [部署]()

doc/talk/201906/sqlflow.slide

@@ -81,7 +81,7 @@ TensorFlow model inputs are dense tensors, which are normally high-dimensional m
 - Hash the name string into a 64-bit integer, then bucketizing hashed value into 100 dimension columns.
 .image column_fea2.png _ 600
 
-Generated code using Tensorflow "Feature Column" should look like:
+Generated code using TensorFlow "Feature Column" should look like:
 .code feature_column_code.pysample
 
 * Design: Feature Derivation

doc/talk/20190620/sqlflow.slide

@@ -63,7 +63,7 @@ TensorFlow model inputs are dense tensors, which usually are high-dimensional ma
 - Hash the name string into a 64-bit integer, then bucketing hashed value into 100 dimension columns.
 .image ../201906/column_fea2.png _ 600
 
-Generated code using Tensorflow "Feature Column" should look like:
+Generated code using TensorFlow "Feature Column" should look like:
 .code ../201906/feature_column_code.pysample
 
 * Design: Feature Derivation

docker/step/README.md

@@ -1,6 +1,6 @@
-# A Slim Tensorflow Step Image
+# A Slim TensorFlow Step Image
 
-This image is used when submitting Argo workflows to run Tensorflow/PAI Tensorflow jobs. To build this image, you should follow the below steps:
+This image is used when submitting Argo workflows to run TensorFlow/PAI TensorFlow jobs. To build this image, you should follow the below steps:
 
 1. Go to SQLFlow root directory
  ```bash