add: 2.16 task1 01~05

Project/ProjectSettings/BurstAotSettings_StandaloneWindows.json

@@ -0,0 +1,18 @@
+{
+  "MonoBehaviour": {
+    "Version": 4,
+    "EnableBurstCompilation": true,
+    "EnableOptimisations": true,
+    "EnableSafetyChecks": false,
+    "EnableDebugInAllBuilds": false,
+    "DebugDataKind": 1,
+    "EnableArmv9SecurityFeatures": false,
+    "CpuMinTargetX32": 0,
+    "CpuMaxTargetX32": 0,
+    "CpuMinTargetX64": 0,
+    "CpuMaxTargetX64": 0,
+    "CpuTargetsX32": 6,
+    "CpuTargetsX64": 72,
+    "OptimizeFor": 0
+  }
+}

tensorboard_data_01_copy.py

@@ -0,0 +1,31 @@
+import os
+import shutil
+
+# Source directory where files are located
+source_dir = os.path.join(os.getcwd(), "results")
+
+# Destination directory where files will be copied
+destination_dir = os.path.join(os.getcwd(), "tensorboard_data_copy_01_results")
+
+# Find all files in the source directory starting with "events"
+files = []
+for root, dirs, filenames in os.walk(source_dir):
+    for filename in filenames:
+        if filename.startswith('events'):
+            files.append(os.path.join(root, filename))
+
+# Loop through each file found
+for file_path in files:
+    # Get the directory where the file is located relative to the source directory
+    relative_dir = os.path.relpath(os.path.dirname(file_path), source_dir)
+
+    # Create the corresponding directory structure in the destination directory
+    os.makedirs(os.path.join(destination_dir, relative_dir), exist_ok=True)
+
+    # Copy the file to the destination directory, preserving the directory structure
+    shutil.copy(file_path, os.path.join(destination_dir, relative_dir))
+
+    # Print out the copied files
+    print(f"Copied {file_path} to {os.path.join(destination_dir, relative_dir)}")
+
+print("Copy operation completed.")

tensorboard_data_02_json.py

@@ -0,0 +1,64 @@
+import os
+import json
+from tensorboard.backend.event_processing.event_accumulator import EventAccumulator, ScalarEvent
+
+def scalar_event_list_to_dict_list(scalar_event_list):
+    dict_list = []
+    for scalar_event in scalar_event_list:
+        # remove wall_time
+        dict_list.append({'step': scalar_event.step, 'value': scalar_event.value})
+    return dict_list
+
+
+def convert_to_json(events_file_path, output_json_path):
+    accumulator = EventAccumulator(events_file_path)
+    accumulator.Reload()  # loads events from the file
+
+    # Get a List[ScalarEvents]
+    scalar_events = accumulator.Scalars('Environment/Cumulative Reward')
+
+    # Convert the List[ScalarEvents] to List[Dict]
+    dict_list = scalar_event_list_to_dict_list(scalar_events)
+
+    # Dump json from List[Dict], then write JSON to a file
+    with open(output_json_path, "w") as json_file:
+        json.dump(dict_list, json_file)
+
+
+if __name__ == "__main__":
+
+    # Source directory where files are located
+    source_dir = os.path.join(os.getcwd(), "tensorboard_data_copy_01_results")
+
+    # Destination directory where files will be copied
+    destination_dir = os.path.join(os.getcwd(), "tensorboard_data_json_02_results")
+
+    # Find all files in the source directory starting with "events"
+    files = []
+    for root, dirs, filenames in os.walk(source_dir):
+        for filename in filenames:
+            if filename.startswith('events'):
+                files.append(os.path.join(root, filename))
+
+    # Loop through each file found
+    for file_path in files:
+        # Get the directory where the file is located relative to the source directory
+        relative_dir = os.path.relpath(os.path.dirname(file_path), source_dir)
+
+        # Create the corresponding directory structure in the destination directory
+        os.makedirs(os.path.join(destination_dir, relative_dir), exist_ok=True)
+
+        # Define the new filename with '.json' extension
+        json_filename = os.path.splitext(os.path.basename(file_path))[0] + '.json'
+
+        # Define the new path of .json file
+        json_filepath = os.path.join(destination_dir, relative_dir, json_filename)
+
+        # Convert tensorboard events file to json
+        convert_to_json(file_path, json_filepath)
+
+        # Print out the coverted files
+        print(f"Covert {file_path} to {json_filepath}")
+
+    print("Convert operation completed.")
+

tensorboard_data_03_numpy.py

@@ -0,0 +1,58 @@
+import os
+import json
+import numpy as np
+
+
+def convert_to_npy(json_file_path, output_npy_path):
+    # Load the JSON file
+    with open(json_file_path, 'r') as f:
+        json_data = json.load(f)
+
+    # Extract 'step' and 'value' from the list of dictionaries
+    steps = np.array([d['step'] for d in json_data])
+    values = np.array([d['value'] for d in json_data])
+
+    # Combine 'steps' and 'values' arrays into a single 2D array
+    combined_array = np.array([steps, values])
+
+    # Save the arrays into separate files
+    np.save(output_npy_path, combined_array)
+
+
+if __name__ == "__main__":
+
+    # Source directory where files are located
+    source_dir = os.path.join(os.getcwd(), "tensorboard_data_json_02_results")
+
+    # Destination directory where files will be copied
+    destination_dir = os.path.join(os.getcwd(), "tensorboard_data_numpy_03_results")
+
+    # Find all files in the source directory starting with "events"
+    files = []
+    for root, dirs, filenames in os.walk(source_dir):
+        for filename in filenames:
+            if filename.startswith('events'):
+                files.append(os.path.join(root, filename))
+
+    # Loop through each file found
+    for file_path in files:
+        # Get the directory where the file is located relative to the source directory
+        relative_dir = os.path.relpath(os.path.dirname(file_path), source_dir)
+
+        # Create the corresponding directory structure in the destination directory
+        os.makedirs(os.path.join(destination_dir, relative_dir), exist_ok=True)
+
+        # Define the new filename with '.npy' extension
+        npy_filename = os.path.splitext(os.path.basename(file_path))[0] + '.npy'
+
+        # Define the new path of .npy file
+        npy_filepath = os.path.join(destination_dir, relative_dir, npy_filename)
+
+        # Convert json to step.npy and value.npy
+        convert_to_npy(file_path, npy_filepath)
+
+        # Print out the coverted files
+        print(f"Covert {file_path} to {npy_filepath}")
+
+    print("Convert operation completed.")
+

tensorboard_data_04_merged_numpy.py

@@ -0,0 +1,66 @@
+import os
+import numpy as np
+
+
+if __name__ == "__main__":
+
+    # Source directory where files are located
+    source_dir = os.path.join(os.getcwd(), "tensorboard_data_numpy_03_results")
+
+    # Destination directory where files will be copied
+    destination_dir = os.path.join(os.getcwd(), "tensorboard_data_merged_numpy_04_results")
+
+    # Find all files in the source directory starting with "events"
+    files = []
+    for root, dirs, filenames in os.walk(source_dir):
+        normal_root = os.path.normpath(root)
+        for filename in filenames:
+            version = os.path.split(normal_root)[-2]
+            if filename.startswith('events') and any(sub in version for sub in ('v5.0.1', 'v5.1.0', 'v5.2.0')) :
+                files.append(os.path.join(root, filename))
+
+    values = 0
+
+    # Loop through each file found
+    for num, file_path in enumerate(files):
+        # Get the directory where the file is located relative to the source directory
+        relative_dir = os.path.relpath(os.path.dirname(file_path), source_dir)
+
+        # Split the directory specifying the version number, va.b.c-d
+        first_level_dir = os.path.split(relative_dir.rstrip(os.sep))[0]
+
+        # Split the version number to get va.b.c
+        version_number = first_level_dir.split('-')[0]
+
+        # Load the data
+        data = np.load(file_path)
+
+        # Extract steps and values
+        steps = data[0]
+        values = values + data[1]
+
+        # Merge results from many seeds
+        if num % 3 == 2:
+            values = values / 3
+
+            # Create the corresponding directory structure in the destination directory
+            os.makedirs(os.path.join(destination_dir, relative_dir), exist_ok=True)
+
+            # Define the new filename with '.npy' extension
+            npy_filename = os.path.splitext(os.path.basename(file_path))[0] + '.npy'
+
+            # Define the new path of .npy file
+            npy_filepath = os.path.join(destination_dir, relative_dir, npy_filename)
+
+            # Convert 'steps' and 'values' arrays into a numpy array
+            combined_array = np.array([steps, values])
+
+            # Save the Dataframe into npy files
+            np.save(npy_filepath, combined_array)
+
+            print(f"Save merged pandas Dataframe {version_number}")
+
+        print(f"Add {file_path}")
+
+
+    print("Convert operation completed.")

tensorboard_data_05_v5.0.x_image.py

@@ -0,0 +1,81 @@
+import os
+import numpy as np
+import pandas as pd
+import seaborn as sns
+import matplotlib.pyplot as plt
+
+def smooth(scalars, weight):  # Weight between 0 and 1
+    last = scalars[0]  # First value in the plot (first timestep)
+    smoothed = list()
+    for point in scalars:
+        smoothed_val = last * weight + (1 - weight) * point  # Calculate smoothed value
+        smoothed.append(smoothed_val)                        # Save it
+        last = smoothed_val                                  # Anchor the last smoothed value
+
+    return smoothed
+
+if __name__ == "__main__":
+
+    # Source directory where files are located
+    source_dir = os.path.join(os.getcwd(), "tensorboard_data_merged_numpy_04_results")
+
+    # Destination directory where files will be copied
+    destination_dir = os.path.join(os.getcwd(), "tensorboard_data_seaborn_05_results")
+
+    # Find all files in the source directory starting with "events"
+    files = []
+    for root, dirs, filenames in os.walk(source_dir):
+        for filename in filenames:
+            if filename.startswith('events'):
+                files.append(os.path.join(root, filename))
+
+    # Legend labels
+    legend_labels = {'v5.0.1': 'Pos+CNN', 'v5.1.0': 'Pos', 'v5.2.0': 'CNN'}
+
+    # Create a line plot
+    plt.figure(figsize=(10, 8))
+
+    # Loop through each file found
+    for num, file_path in enumerate(files):
+        # Get the directory where the file is located relative to the source directory
+        relative_dir = os.path.relpath(os.path.dirname(file_path), source_dir)
+
+        # Split the directory specifying the version number, va.b.c-d
+        first_level_dir = os.path.split(relative_dir.rstrip(os.sep))[0]
+
+        # Split the version number to get va.b.c
+        version_number = first_level_dir.split('-')[0]
+
+        # Read npy file into array
+        data = np.load(file_path)
+
+        # Extract steps and values
+        steps = data[0]
+        values = data[1]
+
+        # Smooth values
+        smoothed_values = smooth(values, 0.95)
+
+
+
+        plt.plot(steps, smoothed_values, linestyle='-', label=legend_labels[version_number])
+        print(f"Add {file_path}")
+
+    # Add labels and title
+    plt.xlabel('Step')
+    plt.ylabel('Cumulative Reward')
+    plt.title('Line Plot of Step vs Cumulative Reward')
+
+    # Show grid
+    plt.grid(True)
+
+    # Add legend
+    plt.legend()
+
+    # Create the corresponding directory structure in the destination directory
+    os.makedirs(destination_dir, exist_ok=True)
+
+    # Save the plot as a JPG file
+    plt.savefig(os.path.join(destination_dir, 'line_plot.jpg'), format='jpg')
+
+    print("plot operation completed.")

test.py

@@ -0,0 +1,29 @@
+import matplotlib.pyplot as plt
+import numpy as np
+import seaborn as sns
+import pandas as pd
+# 生成数据
+x = np.arange(5)
+
+y_a = np.random.uniform(size=5)
+y_b = np.random.uniform(size=5)
+
+label_a = np.full(x.shape, fill_value='a')
+label_b = np.full(x.shape, fill_value='b')
+
+data_a = pd.DataFrame(np.concatenate((x[:, None], label_a[:, None], y_a[:, None]), axis=1),
+                      columns=['x', 'label', 'y'])
+data_b = pd.DataFrame(np.concatenate((x[:, None], label_b[:, None], y_b[:, None]), axis=1),
+                      columns=['x', 'label', 'y'])
+data = pd.concat([data_a, data_b], axis=0)
+
+# x、y轴数据需要为数字类型，但上面的操作后会变成object，所以要进行一下转换
+data[['x', 'y']] = data[['x', 'y']].apply(pd.to_numeric)
+print(data)
+# 设置样式
+sns.set_theme(context='paper', style='darkgrid')
+fig = plt.figure()
+plt.title('multi lines')
+# 绘图
+sns.lineplot(x="x", y="y", data=data, errorbar=('sd', 1))
+plt.show()