version0.2.0

Author: refraction-ray

CHANGELOG.md

@@ -2,6 +2,8 @@
 
 ## Unreleased
 
+## 0.2.0
+
 ### Added
 
 - Add PyTorch nn Module wrapper in `torchnn`
@@ -10,6 +12,8 @@
 
 - Brand new `sample` API with batch support and sampling from state support
 
+- add more methods in global namespace, and add alias `KerasLayer`/`TorchLayer`
+
 ### Fixed
 
 - Fixed bug in merge single gates when all gates are single-qubit ones

tensorcircuit/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.1.3"
+__version__ = "0.2.0"
 __author__ = "TensorCircuit Authors"
 __creator__ = "refraction-ray"
 
@@ -25,14 +25,17 @@
 from . import interfaces
 from . import templates
 from . import quantum
+from .quantum import QuOperator, QuVector, QuAdjointVector, QuScalar
 
 try:
     from . import keras
+    from .keras import QuantumLayer as KerasLayer
 except ModuleNotFoundError:
     pass  # in case tf is not installed
 
 try:
     from . import torchnn
+    from .torchnn import QuantumNet as TorchLayer
 except ModuleNotFoundError:
     pass  # in case torch is not installed
 

tests/test_keras.py

@@ -10,7 +10,6 @@
 import numpy as np
 import tensorflow as tf
 import tensorcircuit as tc
-from tensorcircuit import keras as K
 
 
 dtype = np.complex128
@@ -55,13 +54,15 @@ def vqe_f2(inputs, xweights, zzweights, nlayers, n):
 
 def test_vqe_layer2(tfb, highp):
     vqe_fp = partial(vqe_f2, nlayers=3, n=6)
-    vqe_layer = K.QuantumLayer(vqe_fp, [(3, 6), (3, 6)])
+    vqe_layer = tc.KerasLayer(vqe_fp, [(3, 6), (3, 6)])
     inputs = np.zeros([1])
     with tf.GradientTape() as tape:
         e = vqe_layer(inputs)
     print(e, tape.gradient(e, vqe_layer.variables))
     model = tf.keras.Sequential([vqe_layer])
-    model.compile(loss=K.output_asis_loss, optimizer=tf.keras.optimizers.Adam(0.01))
+    model.compile(
+        loss=tc.keras.output_asis_loss, optimizer=tf.keras.optimizers.Adam(0.01)
+    )
     model.fit(np.zeros([1, 1]), np.zeros([1]), batch_size=1, epochs=300)
 
 
@@ -87,12 +88,14 @@ def vqe_f(inputs, weights, nlayers, n):
 
 def test_vqe_layer(tfb, highp):
     vqe_fp = partial(vqe_f, nlayers=6, n=6)
-    vqe_layer = K.QuantumLayer(vqe_fp, (6 * 2, 6))
+    vqe_layer = tc.keras.QuantumLayer(vqe_fp, (6 * 2, 6))
     inputs = np.zeros([1])
     inputs = tf.constant(inputs)
     model = tf.keras.Sequential([vqe_layer])
 
-    model.compile(loss=K.output_asis_loss, optimizer=tf.keras.optimizers.Adam(0.01))
+    model.compile(
+        loss=tc.keras.output_asis_loss, optimizer=tf.keras.optimizers.Adam(0.01)
+    )
 
     model.fit(np.zeros([2, 1]), np.zeros([2, 1]), batch_size=2, epochs=500)
 
@@ -104,7 +107,7 @@ def test_function_io(tfb, tmp_path, highp):
 
     vqe_f_p = tf.function(vqe_f_p)
     vqe_f_p(weights=tf.ones([6, 6], dtype=tf.float64), nlayers=3, n=6)
-    K.save_func(vqe_f_p, str(tmp_path))
-    loaded = K.load_func(str(tmp_path), fallback=vqe_f_p)
+    tc.keras.save_func(vqe_f_p, str(tmp_path))
+    loaded = tc.keras.load_func(str(tmp_path), fallback=vqe_f_p)
     print(loaded(weights=tf.ones([6, 6], dtype=tf.float64), nlayers=3, n=6))
     print(loaded(weights=tf.ones([6, 6], dtype=tf.float64), nlayers=3, n=6))

tests/test_torchnn.py

@@ -41,7 +41,7 @@ def qpred(x, weights):
     else:
         use_interface = True
 
-    ql = tc.torchnn.QuantumNet(
+    ql = tc.TorchLayer(
         qpred, weights_shape=[2 * nlayers, n], use_interface=use_interface
     )