diff --git a/binread4.py b/binread4.py
index eccd31a90..f120df92b 100644
--- a/binread4.py
+++ b/binread4.py
@@ -30,11 +30,12 @@ def main(mode):
     model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
 
     if mode == 'fit':
-        model.fit(x_train, y_train, batch_size=batch_size, epochs=1, validation_data=(x_test, y_test))
+        model.fit(x_train, y_train, batch_size=batch_size, epochs=4, validation_data=(x_test, y_test))
     else:
         train_steps, train_batches = batch_iter(x_train, y_train, batch_size)
         valid_steps, valid_batches = batch_iter(x_test, y_test, batch_size)
         model.fit_generator(train_batches, train_steps, epochs=1, validation_data=valid_batches, validation_steps=valid_steps)
+    model.save("v1.kersas")
 
 
 if __name__ == '__main__':
diff --git a/testonnx.py b/testonnx.py
new file mode 100644
index 000000000..e7d57e510
--- /dev/null
+++ b/testonnx.py
@@ -0,0 +1,68 @@
+
+# Certainly! To use ONNX Runtime to load and run a model in Python, you'll need to follow these steps:
+
+# 1. **Install ONNX Runtime:**
+#    You can install ONNX Runtime using pip:
+
+#    ```bash
+#    pip install onnxruntime
+#    ```
+
+# 2. **Load the ONNX Model:**
+#    Load your ONNX model using the `onnxruntime.InferenceSession` class:
+
+#    ```python
+#    import onnxruntime
+
+#    # Replace 'your_model.onnx' with the path to your ONNX model
+
+# 3. **Prepare Input Data:**
+#    Prepare input data as a dictionary where keys are the input names specified in your ONNX model and values are the corresponding input data:
+
+#    ```python
+#    import numpy as np
+
+#    # Replace 'input_name' with the actual input name from your model
+#    input_name = session.get_inputs()[0].name
+#    input_data = np.random.random((1, 3, 224, 224)).astype(np.float32)
+
+#    input_dict = {input_name: input_data}
+#    ```
+
+# 4. **Run Inference:**
+#    Run the inference using the `run` method of the `InferenceSession`:
+
+#    ```python
+#    output = session.run(None, input_dict)
+#    ```
+
+#    Replace `None` with the names of the output nodes in your model if you want to retrieve specific outputs.
+
+# Here's a condensed version of the above steps:
+
+# ```python
+import onnxruntime
+import numpy as np
+
+model_path = 'build/model.onnx'
+session = onnxruntime.InferenceSession(model_path)
+
+#{'name': 'embedding_input', 'type': {'tensorType': {'elemType': 1, 'shape': {'dim': [{'dimParam': 'unk__261'}, {'dimParam': 'unk__262'}]}}}}
+input_name = "embedding_input"
+#3session.get_inputs()[0].name
+
+input_data = np.zeros((4096,2),dtype=np.float32)
+for x in (range(2)):
+    output = session.run(None, {input_name: input_data})
+    #print(input_data.shape,output.shape)
+    print("in",input_data.size)
+    print("out",len(output[0]))
+    print("in","".join([ str(j) for j in input_data[0:16]]))
+    print("out","".join([ str(j) for j in output[0][0:16]]))
+
+    for i,x in enumerate(output[0]):
+        input_data[i] = x
+        #input_data[4096 + i] = x
+
+
+# Remember to replace `'your_model.onnx'` with the actual path to your ONNX model, and adjust input data accordingly based on your model's input requirements.