python client changes for new redisaipy

Author: hhsecond

python_client/sklearn_boston_house_price_prediction.py

@@ -1,23 +1,22 @@
-import sys
 import numpy as np
 
-from redisai import BlobTensor, Client, Backend, Device
+import redisai
 from ml2rt import load_model
 
 from cli import arguments
 
 
-tensor = BlobTensor.from_numpy(np.ones((1, 13), dtype=np.float32))
+tensor = np.ones((1, 13), dtype=np.float32)
 model = load_model('../models/sklearn/boston_house_price_prediction/boston.onnx')
 
 if arguments.gpu:
-    device = Device.gpu
+    device = 'GPU'
 else:
-    device = Device.cpu
+    device = 'CPU'
 
-con = Client(host=arguments.host, port=arguments.port)
+con = redisai.Client(host=arguments.host, port=arguments.port)
 con.tensorset('tensor', tensor)
-con.modelset('model', Backend.onnx, device, model)
+con.modelset('model', 'onnx', device, model)
 con.modelrun('model', inputs=['tensor'], outputs=['out'])
-out = con.tensorget('out', as_type=BlobTensor)
-print(out.to_numpy())
+out = con.tensorget('out')
+print(out)

python_client/sklearn_linear_regression.py

@@ -6,19 +6,17 @@
 model = load_model("../models/sklearn/linear_regression/linear_regression.onnx")
 
 if arguments.gpu:
-    device = rai.Device.gpu
+    device = 'gpu'
 else:
-    device = rai.Device.cpu
+    device = 'cpu'
 
 con = rai.Client(host=arguments.host, port=arguments.port)
-con.modelset("sklearn_model", rai.Backend.onnx, device, model)
+con.modelset("sklearn_model", 'onnx', device, model)
 
 # dummydata taken from sklearn.datasets.load_boston().data[0]
-dummydata = [
-    0.00632, 18.0, 2.31, 0.0, 0.538, 6.575, 65.2, 4.09, 1.0, 296.0, 15.3, 396.9, 4.98]
-tensor = rai.Tensor.scalar(rai.DType.float, *dummydata)
-con.tensorset("input", tensor)
+dummydata = [15.0]
+con.tensorset("input", dummydata, dtype='float32', shape=(1, 1))
 
 con.modelrun("sklearn_model", ["input"], ["output"])
-outtensor = con.tensorget("output", as_type=rai.BlobTensor)
-print(f"House cost predicted by model is ${outtensor.to_numpy().item() * 1000}")
+outtensor = con.tensorget("output")
+print(f"House cost predicted by model is ${outtensor.item() * 1000}")

python_client/sklearn_logistic_regression.py

@@ -7,20 +7,19 @@
 model = load_model("../models/sklearn/logistic_regression/logistic.onnx")
 
 if arguments.gpu:
-    device = rai.Device.gpu
+    device = 'gpu'
 else:
-    device = rai.Device.cpu
+    device = 'cpu'
 
 con = rai.Client(host=arguments.host, port=arguments.port)
-con.modelset("sklearn_model", rai.Backend.onnx, device, model)
+con.modelset("sklearn_model", 'onnx', device, model)
 
 dummydata = np.array([[6.9, 3.1, 5.4, 2.1]], dtype=np.float32)
-tensor = rai.BlobTensor.from_numpy(dummydata)
-con.tensorset("input", tensor)
+con.tensorset("input", dummydata)
 
 # dummy output because by default sklearn logistic regression outputs
 # value and probability. Since RedisAI doesn't support specifying required
 # outputs now, we need to keep placeholders for all the default outputs.
 con.modelrun("sklearn_model", ["input"], ["output", "dummy"])
-outtensor = con.tensorget("output", as_type=rai.BlobTensor)
-print(f" Output class: {outtensor.to_numpy().item()}")
+outtensor = con.tensorget("output")
+print(f" Output class: {outtensor.item()}")

python_client/spark_decisiontree.py

@@ -7,15 +7,14 @@
 model = load_model("../models/spark/decisiontree_with_pipeline/spark.onnx")
 
 if arguments.gpu:
-    device = rai.Device.gpu
+    device = 'gpu'
 else:
-    device = rai.Device.cpu
+    device = 'cpu'
 
 con = rai.Client(host=arguments.host, port=arguments.port)
-con.modelset("spark_model", rai.Backend.onnx, device, model)
-features = np.array([[0., 0., 0., 0., 0.], [0., 0., 0., 17., 206.]], dtype=np.float32)
-tensor = rai.BlobTensor.from_numpy(features)
-con.tensorset("input", tensor)
+con.modelset("spark_model", 'onnx', device, model)
+features = np.array([[0., 0., 0., 17., 206.]], dtype=np.float32)
+con.tensorset("input", features)
 con.modelrun("spark_model", ["input"], ["output"])
-outtensor = con.tensorget("output", as_type=rai.BlobTensor)
-print(outtensor.to_numpy())
+outtensor = con.tensorget("output")
+print(outtensor)

python_client/spark_linear_regression.py

@@ -6,15 +6,14 @@
 model = load_model("../models/spark/linear_regression/linear_regression.onnx")
 
 if arguments.gpu:
-    device = rai.Device.gpu
+    device = 'gpu'
 else:
-    device = rai.Device.cpu
+    device = 'cpu'
 
 con = rai.Client(host=arguments.host, port=arguments.port)
-con.modelset("spark_model", rai.Backend.onnx, device, model, inputs=['features'])
+con.modelset("spark_model", 'onnx', device, model, inputs=['features'])
 dummydata = [15.0]
-tensor = rai.Tensor.scalar(rai.DType.float, *dummydata)
-con.tensorset("input", tensor)
+con.tensorset("input", dummydata, shape=(1, 1), dtype='float32')
 con.modelrun("spark_model", ["input"], ["output"])
-outtensor = con.tensorget("output", as_type=rai.BlobTensor)
-print(outtensor.to_numpy())
+outtensor = con.tensorget("output")
+print(outtensor)

python_client/spark_one_vs_rest.py

@@ -7,16 +7,15 @@
 model = load_model("../models/spark/one_vs_rest/spark.onnx")
 
 if arguments.gpu:
-    device = rai.Device.gpu
+    device = 'gpu'
 else:
-    device = rai.Device.cpu
+    device = 'cpu'
 
 con = rai.Client(host=arguments.host, port=arguments.port)
-con.modelset("spark_model", rai.Backend.onnx, device, model)
+con.modelset("spark_model", 'onnx', device, model)
 dummydata = np.array(
     [[-0.222222, 0.5, -0.762712, -0.833333]], dtype=np.float32)
-tensor = rai.BlobTensor.from_numpy(dummydata)
-con.tensorset("input", tensor)
+con.tensorset("input", dummydata)
 con.modelrun("spark_model", ["input"], ["output"])
-outtensor = con.tensorget("output", as_type=rai.BlobTensor)
-print(outtensor.to_numpy())
+outtensor = con.tensorget("output")
+print(outtensor)

python_client/spark_pca.py

@@ -6,17 +6,16 @@
 model = load_model("../models/spark/pca/spark.onnx")
 
 if arguments.gpu:
-    device = rai.Device.gpu
+    device = 'gpu'
 else:
-    device = rai.Device.cpu
+    device = 'cpu'
 
 con = rai.Client(host=arguments.host, port=arguments.port)
-con.modelset("spark_model", rai.Backend.onnx, device, model)
+con.modelset("spark_model", 'onnx', device, model)
 dummydata = np.array(
-    [[2.0, 0.0, 3.0, 4.0, 5.0], [4.0, 0.0, 0.0, 6.0, 7.0]],
+    [[5, 1.0, 7.0, 7.0, 7.0], [2.0, 0.0, 3.0, 4.0, 5.0], [4.0, 0.0, 0.0, 6.0, 7.0]],
     dtype=np.float32)
-tensor = rai.BlobTensor.from_numpy(dummydata)
-con.tensorset("input", tensor)
+con.tensorset("input", dummydata)
 con.modelrun("spark_model", ["input"], ["output"])
-outtensor = con.tensorget("output", as_type=rai.BlobTensor)
-print(outtensor.to_numpy())
+outtensor = con.tensorget("output")
+print(outtensor)

python_client/tensorflow_imagenet.py

@@ -6,9 +6,9 @@
 from cli import arguments
 
 if arguments.gpu:
-    device = rai.Device.gpu
+    device = 'gpu'
 else:
-    device = rai.Device.cpu
+    device = 'cpu'
 
 con = rai.Client(host=arguments.host, port=arguments.port)
 
@@ -25,16 +25,15 @@
 
 
 out1 = con.modelset(
-    'imagenet_model', rai.Backend.tf, device,
+    'imagenet_model', 'tf', device,
     inputs=['images'], outputs=['output'], data=tf_model)
 out2 = con.scriptset('imagenet_script', device, script)
 a = time.time()
-tensor = rai.BlobTensor.from_numpy(image)
-con.tensorset('image', tensor)
+con.tensorset('image', image)
 out4 = con.scriptrun('imagenet_script', 'pre_process_3ch', 'image', 'temp1')
 out5 = con.modelrun('imagenet_model', 'temp1', 'temp2')
 out6 = con.scriptrun('imagenet_script', 'post_process', 'temp2', 'out')
-final = con.tensorget('out', as_type=rai.BlobTensor)
-ind = final.to_numpy().item()
+final = con.tensorget('out')
+ind = final.item()
 print(ind, class_idx[str(ind)])
 print(time.time() - a)

python_client/tensorflow_tinyyolo.py

@@ -31,15 +31,15 @@
 
 
 if arguments.gpu:
-    device = rai.Device.gpu
+    device = 'gpu'
 else:
-    device = rai.Device.cpu
+    device = 'cpu'
 
 con = rai.Client(host=arguments.host, port=arguments.port)
 model = ml2rt.load_model('../models/tensorflow/tinyyolo/tinyyolo.pb')
 script = ml2rt.load_script('../models/tensorflow/tinyyolo/yolo_boxes_script.py')
 
-con.modelset('yolo', rai.Backend.tf, device, model, ['input'], ['output'])
+con.modelset('yolo', 'tf', device, model, inputs=['input'], outputs=['output'])
 con.scriptset('yolo-post', device, script)
 
 img_jpg = Image.open('../data/sample_dog_416.jpg')
@@ -48,11 +48,10 @@
 img = np.expand_dims(img, axis=0)
 img /= 256.0
 
-tensor = rai.BlobTensor.from_numpy(img)
-con.tensorset('in', tensor)
+con.tensorset('in', img)
 con.modelrun('yolo', 'in', 'out')
 con.scriptrun('yolo-post', 'boxes_from_tf', inputs='out', outputs='boxes')
-boxes = con.tensorget('boxes', as_type=rai.BlobTensor).to_numpy()
+boxes = con.tensorget('boxes')
 
 n_boxes = 0
 for box in boxes[0]:

python_client/torch_charrnn.py

@@ -5,9 +5,9 @@
 from cli import arguments
 
 if arguments.gpu:
-    device = rai.Device.gpu
+    device = 'gpu'
 else:
-    device = rai.Device.cpu
+    device = 'cpu'
 
 con = rai.Client(host=arguments.host, port=arguments.port)
 all_characters = string.printable
@@ -25,11 +25,10 @@ def int2str(int_data):
 
 model = ml2rt.load_model(filepath)
 
-out1 = con.modelset('charRnn', rai.Backend.torch, device, model)
+out1 = con.modelset('charRnn', 'torch', device, model)
 hidden = np.zeros((n_layers, batch_size, hidden_size), dtype=np.float32)
-hidden_tensor = rai.BlobTensor.from_numpy(hidden)
-out2 = con.tensorset('hidden', hidden_tensor)
-prime_tensor = rai.Tensor(rai.DType.int64, shape=(1,), value=5)
+out2 = con.tensorset('hidden', hidden)
+prime_tensor = np.array([5], np.int64)
 out3 = con.tensorset('prime', prime_tensor)
 out4 = con.modelrun('charRnn', ['prime', 'hidden'], ['out'])
 out5 = con.tensorget('out')