diff --git a/.gitignore b/.gitignore
index acbf51292469935bbd05cf70cc53c90b18eb6181..3f347a65cf8e2c12a45294705ce3cd6e51973f9f 100644
--- a/.gitignore
+++ b/.gitignore
@@ -2,6 +2,7 @@
**/test
**/wandb
**/Trajectory Data
+**/Trajectory_Data
**/Foundation_Model
**/models
**/analysis
diff --git a/coscine_watchdog/coscine_app_profile.ttl b/coscine_watchdog/coscine_app_profile-Trajectory-data.ttl
similarity index 100%
rename from coscine_watchdog/coscine_app_profile.ttl
rename to coscine_watchdog/coscine_app_profile-Trajectory-data.ttl
diff --git a/coscine_watchdog/requirements.txt b/coscine_watchdog/requirements.txt
index b8158ff26ddac1b773f6b0b2286f2d062ae9e4f1..0b79f27058e56dfbcf59d0927f60789811ae3750 100644
--- a/coscine_watchdog/requirements.txt
+++ b/coscine_watchdog/requirements.txt
@@ -12,3 +12,4 @@ furo==2024.5.6 # sphinx documentation theme
myst-parser==3.0.1 # sphinx plugin for markdown support
sphinx-copybutton==0.5.2 # inserts buttons to copy sourcecode in docs
flask==3.0.3
+coscine>=0.11.5 #for the coscine library. versions lower than 0.11.5 only download 1000 files.
\ No newline at end of file
diff --git a/dynamics_learning/Dockerfile.foundation_model b/dynamics_learning/Dockerfile.foundation_model
index fa2479404af6bb8221c4047a0c63233439f2d4d0..a4d2c6e571e0731f5da196ef3e60603dccd21d00 100644
--- a/dynamics_learning/Dockerfile.foundation_model
+++ b/dynamics_learning/Dockerfile.foundation_model
@@ -1,5 +1,25 @@
FROM nvcr.io/nvidia/tensorflow:23.05-tf2-py3
+# Install required packages for setting up the locale and timezone
+RUN apt-get update && apt-get install -y \
+ locales \
+ tzdata
+
+# Set the timezone to Europe/Berlin
+RUN ln -fs /usr/share/zoneinfo/Europe/Berlin /etc/localtime && \
+ dpkg-reconfigure -f noninteractive tzdata
+
+# Set the locale to en_US.UTF-8
+RUN echo "en_US.UTF-8 UTF-8" > /etc/locale.gen && \
+ locale-gen en_US.UTF-8 && \
+ update-locale LANG=en_US.UTF-8
+
+# Set environment variables for timezone and locale
+ENV LANG=en_US.UTF-8 \
+ LANGUAGE=en_US:en \
+ LC_ALL=en_US.UTF-8 \
+ TZ=Europe/Berlin
+
WORKDIR /app
COPY . .
diff --git a/dynamics_learning/Dockerfile.foundation_model_cron_job b/dynamics_learning/Dockerfile.foundation_model_cron_job
index 23236cf1960bd3ad3bcc61892777629b60fc1cef..0e5d7ff65288e895b220bd6ed44d3ff4681c0736 100644
--- a/dynamics_learning/Dockerfile.foundation_model_cron_job
+++ b/dynamics_learning/Dockerfile.foundation_model_cron_job
@@ -1,5 +1,25 @@
FROM nvcr.io/nvidia/tensorflow:23.05-tf2-py3
+# Install required packages for setting up the locale and timezone
+RUN apt-get update && apt-get install -y \
+ locales \
+ tzdata
+
+# Set the timezone to Europe/Berlin
+RUN ln -fs /usr/share/zoneinfo/Europe/Berlin /etc/localtime && \
+ dpkg-reconfigure -f noninteractive tzdata
+
+# Set the locale to en_US.UTF-8
+RUN echo "en_US.UTF-8 UTF-8" > /etc/locale.gen && \
+ locale-gen en_US.UTF-8 && \
+ update-locale LANG=en_US.UTF-8
+
+# Set environment variables for timezone and locale
+ENV LANG=en_US.UTF-8 \
+ LANGUAGE=en_US:en \
+ LC_ALL=en_US.UTF-8 \
+ TZ=Europe/Berlin
+
WORKDIR /app
COPY . .
diff --git a/dynamics_learning/benchmark_model_accuracy.py b/dynamics_learning/benchmark_model_accuracy.py
index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..d4e9016ea4837671541a6378e99ae3f4919e893c 100644
--- a/dynamics_learning/benchmark_model_accuracy.py
+++ b/dynamics_learning/benchmark_model_accuracy.py
@@ -0,0 +1,47 @@
+from pathlib import Path
+from dynamics_learning.data_retrieval import download_resource_content_into_uuid_folders
+
+
+def delete_files(num_trajectories_to_keep: int, robot_uuid: str) -> None:
+ """Delete files from the training data directory.
+
+ Files are sorted by date and the newest files are kept.
+
+ Args:
+ num_trajectories_to_keep (int): Number of trajectories to keep.
+ robot_uuid (str): Robot UUID.
+
+ Returns:
+ None: This function does not return anything.
+ """
+ files = [
+ str(file)
+ for file in Path(
+ f"/app/dynamics_learning/Trajectory Data/train/{robot_uuid}"
+ ).iterdir()
+ if file.is_file() and str(file).endswith("meas.csv")
+ ]
+ files.sort(reverse=True)
+ for file in files[num_trajectories_to_keep:]:
+ Path(file).unlink()
+ try:
+ file = file.replace("meas.csv", "com.csv")
+ Path(file).unlink()
+ except FileNotFoundError:
+ # logger.info("No com.csv file found.")
+ pass
+ try:
+ file = file.replace("com.csv", "interp_com.csv")
+ Path(file).unlink()
+ except FileNotFoundError:
+ # logger.info("No interp_com.csv file found.")
+ pass
+ return None
+
+
+if __name__ == "__main__":
+ download_resource_content_into_uuid_folders()
+ num_trajectories_to_keep = 10
+ LLT_ROBOT_UUID = "f2e72889-c140-4397-809f-fba1b892f17a"
+ robot_uuid = LLT_ROBOT_UUID
+ delete_files(num_trajectories_to_keep, robot_uuid)
diff --git a/dynamics_learning/benchmark_number_of_runs.py b/dynamics_learning/benchmark_number_of_runs.py
index 950a7ac41309396d95b0397206f7e753aefd9144..8fd86a35d11336cf4038fa68f61d42fe735faad5 100644
--- a/dynamics_learning/benchmark_number_of_runs.py
+++ b/dynamics_learning/benchmark_number_of_runs.py
@@ -3,15 +3,15 @@ from dynamics_learning.preprocessing.dataset_analysis import analyze
from dynamics_learning.preprocessing.trajectory_interpolation import interpolate
from dynamics_learning.environment import (
WANDB_API_TOKEN,
- SWEEP_ID,
WANDB_PROJECT,
WANDB_ENTITY,
)
-from dynamics_learning.sweep.setup import setup_sweep
+from dynamics_learning.sweep.setup import setup_sweep, setup_sweep_from_hyperparameters
from dynamics_learning.training import train
from dynamics_learning.model_io import (
save_model_to_binary_file,
load_model_from_binary_file,
+ load_config,
)
from pathlib import Path
from pritty_logger import RichLogger
@@ -25,8 +25,7 @@ from keras.models import Sequential
logger = RichLogger("dynamics_learning-benchmark_number_of_runs")
-THRESHOLD = 160
-NUMBER_OF_TRAJECTORIES = 100
+THRESHOLD = 50
LLT_ROBOT_UUID = "f2e72889-c140-4397-809f-fba1b892f17a"
ITA_ROBOT_UUID = "c9ff52e1-1733-4829-a209-ebd1586a8697"
WZL_ROBOT_UUID = "2e60a671-dcc3-4a36-9734-a239c899b57d"
@@ -34,9 +33,9 @@ runs = 0
val_loss = 1000
download_file = False
model1 = False
-model2 = True
-model3 = False
-model4 = False
+model2 = False
+model3 = True
+model4 = True
def prepare_data(directory: Path) -> tuple:
@@ -107,6 +106,7 @@ def train_until_threshold_val_loss(
q_qd_qdd_interpolated_command_input: tf.Tensor,
tau_attained_input: tf.Tensor,
model: Sequential = None,
+ notes: str = "",
):
training = partial(
training_loop,
@@ -117,10 +117,7 @@ def train_until_threshold_val_loss(
model=model,
)
wandb.agent(
- sweep_id,
- training,
- project=WANDB_PROJECT,
- entity=WANDB_ENTITY,
+ sweep_id, training, project=WANDB_PROJECT, entity=WANDB_ENTITY, notes=notes
)
logger.info(f"""Training concluded.
Number of runs: {runs}
@@ -141,6 +138,7 @@ if __name__ == "__main__":
# Download Training Data from the server
if download_file:
download_resource_content_into_uuid_folders()
+ # TODO implement max number of trajectories used for benchmark training
wandb.login(key=WANDB_API_TOKEN, relogin=True)
@@ -177,10 +175,12 @@ if __name__ == "__main__":
q_qd_qdd_interpolated_command_input=q_qd_qdd_interpolated_command_input,
tau_attained_input=tau_attained_input,
model=None,
+ notes="Sweep to train model from scratch. 50 Trajectories are avaiulable for training. Training ist stoped when the validation loss is below 50.",
)
runs_model1 = runs
val_loss_model1 = val_loss
+ sweep_id1 = sweep_id
###############################################
####################Model 2####################
@@ -220,39 +220,132 @@ if __name__ == "__main__":
q_qd_qdd_interpolated_command_input=q_qd_qdd_interpolated_command_input,
tau_attained_input=tau_attained_input,
model=model,
+ notes="Sweep to train model based on ITA model. 50 Trajectories are avaiulable for training. Training is stoped when the validation loss is below 50.",
)
runs_model2 = runs
val_loss_model2 = val_loss
+ sweep_id2 = sweep_id
###############################################
####################Model 3####################
###############################################
if model3:
# LLT model based on ITA model with known hyperparameters
- assert (
- SWEEP_ID == "fe3gjovo"
- ), "Sweep ID is not set correctly. Ensure that the sweep id is set to fe3gjovo"
+ robot_uuid = LLT_ROBOT_UUID
+ directory = Path(f"/app/dynamics_learning/Trajectory Data/train/{robot_uuid}")
+ # Interpolate Training Data in UUID folders
+ (
+ attained_data,
+ command_data,
+ interpolated_command_data,
+ q_qd_qdd_interpolated_command_input,
+ tau_attained_input,
+ ) = prepare_data(directory)
+ # assert SWEEP_ID == "42d8t40t", "Sweep ID is not set correctly. Ensure that the sweep id is set to 42d8t40t"
assert (
robot_uuid == LLT_ROBOT_UUID
- ), "Robot UUID is not set correctly. Ensure that the robot uuid is set to LLT_ROBOT"
+ ), "Robot UUID is not set correctly. Ensure that the robot uuid is set to LLT_ROBOT_UUID"
+
+ config_data = load_config(
+ "/app/dynamics_learning/Foundation_Model/models/hyperparameters.json"
+ )
+
+ sweep_id, sweep_config = setup_sweep_from_hyperparameters(
+ config_data=config_data, create_sweep=True
+ )
+
+ # reset runs counter
+ runs = 0
+ val_loss = 1000
+
+ # TODO load ITA model instead of dummy model
+ model = load_model_from_binary_file(
+ "/app/dynamics_learning/models/99.99706268310547.h5"
+ )
+
+ # Train the model until the threshold validation loss is reached
+ train_until_threshold_val_loss(
+ sweep_id=sweep_id,
+ robot_uuid=robot_uuid,
+ q_qd_qdd_interpolated_command_input=q_qd_qdd_interpolated_command_input,
+ tau_attained_input=tau_attained_input,
+ model=model,
+ notes="Sweep to train model based on ITA model with known hyperparameters. 50 Trajectories are avaiulable for training. Training ist stoped when the validation loss is below 50.",
+ )
+
+ runs_model3 = runs
+ val_loss_model3 = val_loss
+ sweep_id3 = sweep_id
- # TODO load LLT Data
- # TODO load ITA model
- # TODO set hyperparameters for LLT model based on ITA model with known hyperparameters
+ # assert (
+ # SWEEP_ID == "fe3gjovo"
+ # ), "Sweep ID is not set correctly. Ensure that the sweep id is set to fe3gjovo"
+ # assert (
+ # robot_uuid == LLT_ROBOT_UUID
+ # ), "Robot UUID is not set correctly. Ensure that the robot uuid is set to LLT_ROBOT"
###############################################
####################Model 4####################
###############################################
if model4:
# LLT model based on foundation model
- assert (
- SWEEP_ID == "7tglijx8"
- ), "Sweep ID is not set correctly. Ensure that the sweep id is set to 7tglijx8"
+ # assert (
+ # SWEEP_ID == "7tglijx8"
+ # ), "Sweep ID is not set correctly. Ensure that the sweep id is set to 7tglijx8"
+ # assert (
+ # robot_uuid == LLT_ROBOT_UUID
+ # ), "Robot UUID is not set correctly. Ensure that the robot uuid is set to LLT_ROBOT"
+
+ robot_uuid = LLT_ROBOT_UUID
+ directory = Path(f"/app/dynamics_learning/Trajectory Data/train/{robot_uuid}")
+ # Interpolate Training Data in UUID folders
+ (
+ attained_data,
+ command_data,
+ interpolated_command_data,
+ q_qd_qdd_interpolated_command_input,
+ tau_attained_input,
+ ) = prepare_data(directory)
+ # assert SWEEP_ID == "42d8t40t", "Sweep ID is not set correctly. Ensure that the sweep id is set to 42d8t40t"
assert (
robot_uuid == LLT_ROBOT_UUID
- ), "Robot UUID is not set correctly. Ensure that the robot uuid is set to LLT_ROBOT"
+ ), "Robot UUID is not set correctly. Ensure that the robot uuid is set to LLT_ROBOT_UUID"
+
+ config_data = load_config(
+ "/app/dynamics_learning/Foundation_Model/models/hyperparameters.json"
+ )
+
+ sweep_id, sweep_config = setup_sweep_from_hyperparameters(
+ config_data=config_data, create_sweep=True
+ )
- # TODO load LLT Data
- # TODO load foundation model
- # TODO set hyperparameters for LLT model based on foundation model
+ # reset runs counter
+ runs = 0
+ val_loss = 1000
+
+ model = load_model_from_binary_file(
+ "/app/dynamics_learning/Foundation_Model/models/Foundation_model.h5"
+ )
+
+ # Train the model until the threshold validation loss is reached
+ train_until_threshold_val_loss(
+ sweep_id=sweep_id,
+ robot_uuid=robot_uuid,
+ q_qd_qdd_interpolated_command_input=q_qd_qdd_interpolated_command_input,
+ tau_attained_input=tau_attained_input,
+ model=model,
+ notes="Sweep to train model based on foundation model with known hyperparameters. 50 Trajectories are avaiulable for training. Training ist stoped when the validation loss is below 50.",
+ )
+
+ runs_model4 = runs
+ val_loss_model4 = val_loss
+ sweep_id4 = sweep_id
+
+ logger.info(f"""Training concluded.
+The first model using sweep {sweep_id1} was trained for {runs_model1} runs and reached a validation loss of {val_loss_model1}.
+The second model using sweep {sweep_id2} was trained for {runs_model2} runs and reached a validation loss of {val_loss_model2}.
+The third model using sweep {sweep_id3} was trained for {runs_model3} runs and reached a validation loss of {val_loss_model3}.
+The fourth model using sweep {sweep_id4} was trained for {runs_model4} runs and reached a validation loss of {val_loss_model4}.
+""")
+ wandb.finish()
diff --git a/dynamics_learning/dynamics_learning/data_retrieval/__init__.py b/dynamics_learning/dynamics_learning/data_retrieval/__init__.py
index 3cf38bfdff501772e99815bb6955f882c7f3a597..6e4571d6e2d0d07bb86f323020e0818cddb66cfb 100644
--- a/dynamics_learning/dynamics_learning/data_retrieval/__init__.py
+++ b/dynamics_learning/dynamics_learning/data_retrieval/__init__.py
@@ -4,7 +4,7 @@
# Released under MIT License
# Contact us for other licensing options.
-import sys
+
from pathlib import Path
from typing import List, Tuple
@@ -13,11 +13,14 @@ import coscine.resource
from pritty_logger import RichLogger
from rich.progress import track
+
from dynamics_learning.environment import COSCINE_API_TOKEN
logger = RichLogger("dynamics_learning-coscine_data_retrieval")
-CLIENT = coscine.ApiClient(COSCINE_API_TOKEN, timeout=120, retries=5)
+CLIENT = coscine.ApiClient(
+ COSCINE_API_TOKEN, timeout=120, retries=5, verbose=False, enable_caching=True
+)
PROJECT = CLIENT.project(
"IoP Ws A.III FER WWL Demo"
) # Connect to the specified Coscine project
@@ -29,8 +32,45 @@ RESOURCE = PROJECT.resource(
) # Address the specific resource in the project
+def delete_files(num_trajectories_to_keep: int, robot_uuid: str) -> None:
+ """Delete files from the training data directory.
+
+ Files are sorted by date and the newest files are kept.
+
+ Args:
+ num_trajectories_to_keep (int): Number of trajectories to keep.
+ robot_uuid (str): Robot UUID.
+
+ Returns:
+ None: This function does not return anything.
+ """
+ files = [
+ str(file)
+ for file in Path(
+ f"/app/dynamics_learning/Trajectory Data/train/{robot_uuid}"
+ ).iterdir()
+ if file.is_file() and str(file).endswith("meas.csv")
+ ]
+ files.sort(reverse=True)
+ for file in files[num_trajectories_to_keep:]:
+ Path(file).unlink()
+ try:
+ file = file.replace("meas.csv", "com.csv")
+ Path(file).unlink()
+ except FileNotFoundError:
+ logger.info("No com.csv file found.")
+ try:
+ file = file.replace("com.csv", "interp_com.csv")
+ Path(file).unlink()
+ except FileNotFoundError:
+ logger.info("No interp_com.csv file found.")
+ return None
+
+
def download_resource_content_into_uuid_folders():
+ """Download the resource content into folders named after the robot UUID. Keeps only 50 newest trajectories per robot."""
files = RESOURCE.files(path="train", recursive=True, with_metadata=True)
+ logger.info(f"Attempting to download {len(files)} files:") # \n{files}
for file in track(files):
if file.is_folder:
continue
@@ -38,45 +78,86 @@ def download_resource_content_into_uuid_folders():
robot_uuid = file.metadata_form()["Robot UUID"][0]
Path(f"./Trajectory Data/train/{robot_uuid}").mkdir(parents=True, exist_ok=True)
file.download(f"./Trajectory Data/train/{robot_uuid}/{file.name}")
+ # logger.info(f"Keeping only 50 trajectories per robot.")
+ # delete_files(50, robot_uuid)
def download_resource_content(resource: coscine.resource.Resource = RESOURCE) -> Path:
- """Downloads none-existend contents of the resource.
+ logger.info(f"Downloading files from resource. This may take a while.\n{resource}")
+ resource.download()
+ # logger.info(f"Downloading files from resource\n{resource}")
+ # files = resource.files(path="train",recursive=True, with_metadata=True)
+ # logger.info(f"Attempting to download {len(files)} files.")
+ # current_iteration = 0
+ # num_currently_downloaded_files = 0
+ # robots = {}
+ # for file in track(files):
+ # logger.info(f"File: {file.name}")
+ # logger.info(f"Robot: {file.metadata_form()['Robot UUID'][0]}")
+ # try:
+ # robots[f"{file.metadata_form()['Robot UUID'][0]}"] = robots[f"{file.metadata_form()['Robot UUID'][0]}"] + 1
+ # except KeyError:
+ # robots[f"{file.metadata_form()['Robot UUID'][0]}"] = 1
+ # if file.is_folder:
+ # logger.info("Skipping folder.")
+ # current_iteration += 1
+ # continue
+ # file.download(f"./{resource.name}/{file.directory}{file.name}")
+ # logger.info(f"Downloaded file to: ./{resource.name}/{file.directory}{file.name}")
+ # current_iteration += 1
+ # if current_iteration % 490 == 0:
+ # logger.info("Sleeping for 60 seconds to not interfere with rate limit.")
+ # logger.info(f"Current iteration: {current_iteration}")
+ # time.sleep(60)
+ # logger.info((robots))
+ # return Path(f"./{resource.name}")
- Args:
- resource (coscine.resource.Resource, optional): The resource to download from. Defaults to RESOURCE.
- """
- path = Path(f"/app/dynamics_learning/{RESOURCE.name}")
- resource_files, resource_objects = get_resource_content(resource)
- resource_files = set(resource_files)
- existing_files = set(get_downloaded_files())
- unique_elements = resource_files.symmetric_difference(existing_files)
- missing_files = list(unique_elements)
- if missing_files:
- logger.log(
- f"Files not yet downloaded are {missing_files}. These files will now be downloaded."
- )
- for file in track(missing_files):
- file_path = path / file
- try:
- resource.file(file).download(
- path=f"/app/dynamics_learning/{RESOURCE.name}/{file}"
- )
- except Exception as e:
- logger.log(f"{e}", "warning")
- logger.log(
- "Likely you have a local file that is not in coscine resource."
- )
- logger.log("Checking if file exists in local directory.")
- logger.log(f"File: {file_path}")
- logger.log(f"Exists: {file_path.exists()}")
- if file_path.exists():
- continue
- else:
- sys.exit(1)
- else:
- logger.log("All files are already downloaded.")
- return path
+
+# def download_resource_content(resource: coscine.resource.Resource = RESOURCE) -> Path:
+# """Downloads none-existent contents of the resource.
+
+# Args:
+# resource (coscine.resource.Resource, optional): The resource to download from. Defaults to RESOURCE.
+# """
+# path = Path(f"/app/dynamics_learning/{RESOURCE.name}")
+# path.mkdir(parents=True, exist_ok=True)
+# path = path / "train"
+# path.mkdir(parents=True, exist_ok=True)
+# path = path / "test"
+# path.mkdir(parents=True, exist_ok=True)
+# path = Path(f"/app/dynamics_learning/{RESOURCE.name}")
+# resource_files, resource_objects = get_resource_content(resource)
+# resource_files = set(resource_files)
+# existing_files = set(get_downloaded_files())
+# unique_elements = resource_files.symmetric_difference(existing_files)
+# missing_files = list(unique_elements)
+# if missing_files:
+# logger.log(
+# f"Files not yet downloaded are {missing_files}. These files will now be downloaded."
+# )
+# for file in track(missing_files):
+# file_path = path / file
+# try:
+# if resource.file(file).is_folder():
+# continue
+# resource.file(file).download(
+# path=f"/app/dynamics_learning/{RESOURCE.name}/{file}"
+# )
+# except Exception as e:
+# logger.log(f"{e}", "warning")
+# logger.log(
+# "Likely you have a local file that is not in coscine resource."
+# )
+# logger.log("Checking if file exists in local directory.")
+# logger.log(f"File: {file_path}")
+# logger.log(f"Exists: {file_path.exists()}")
+# if file_path.exists():
+# continue
+# else:
+# sys.exit(1)
+# else:
+# logger.log("All files are already downloaded.")
+# return path
def get_resource_content(
diff --git a/dynamics_learning/dynamics_learning/sweep/setup.py b/dynamics_learning/dynamics_learning/sweep/setup.py
index a72c7ad8a4cb78793319bcefc797780ff4774a69..6db09d052d7321f4855623ad63e3c177b950381b 100644
--- a/dynamics_learning/dynamics_learning/sweep/setup.py
+++ b/dynamics_learning/dynamics_learning/sweep/setup.py
@@ -48,7 +48,7 @@ def setup_sweep(create_sweep: bool = False) -> tuple[str, dict[str, Any]]:
def setup_sweep_from_hyperparameters(
- config_data: Any, create_sweep: bool = False
+ config_data: dict, create_sweep: bool = False
) -> tuple[str, dict[str, Any]]:
"""Sets up a Sweep from given hyperparameters within a sain range.
@@ -61,35 +61,35 @@ def setup_sweep_from_hyperparameters(
"""
global SWEEP_ID
parameters_dict_from_hyperparameters = {
- "optimizer": config_data.optimizer,
+ "optimizer": {"value": config_data["optimizer"]},
"clipnorm": {
"distribution": "log_uniform_values",
- "min": config_data.clipnorm * 0.5,
- "max": config_data.clipnorm * 1.5,
+ "min": config_data["clipnorm"] * 0.5,
+ "max": config_data["clipnorm"] * 1.5,
},
"learning_rate": {
"distribution": "log_uniform_values",
- "min": config_data.learning_rate * 0.5,
- "max": min(config_data.learning_rate * 1.5, 0.9),
+ "min": config_data["learning_rate"] * 0.5,
+ "max": min(config_data["learning_rate"] * 1.5, 0.9),
},
- "window_size": {"value": config_data.window_size},
- "batch_size": {"value": config_data.batch_size},
+ "window_size": {"value": config_data["window_size"]},
+ "batch_size": {"value": config_data["batch_size"]},
"units": {
"distribution": "int_uniform",
- "min": max(1, config_data.units - 10),
- "max": min(config_data.units + 10, 100),
+ "min": max(1, config_data["units"] - 10),
+ "max": min(config_data["units"] + 10, 100),
},
"dropout": {
"distribution": "log_uniform_values",
- "min": config_data.dropout * 0.5,
- "max": min(config_data.dropout * 1.5, 1),
+ "min": config_data["dropout"] * 0.5,
+ "max": min(config_data["dropout"] * 1.5, 1),
},
"layers": {
"distribution": "int_uniform",
- "min": max(1, config_data.layers - 10),
- "max": min(config_data.layers + 10, 100),
+ "min": max(1, config_data["layers"] - 10),
+ "max": min(config_data["layers"] + 10, 100),
}, # {"value": 10},
- "epochs": {"value": config_data.epochs},
+ "epochs": {"value": config_data["epochs"]},
}
sweep_config_from_hyperparameters["parameters"] = (
parameters_dict_from_hyperparameters
diff --git a/dynamics_learning/dynamics_learning/training/__init__.py b/dynamics_learning/dynamics_learning/training/__init__.py
index 18505eb88b520125e82889840175e30b46a50d02..335551b39bd8ac21c8f9d2d49c0ecb6a7a3e2058 100644
--- a/dynamics_learning/dynamics_learning/training/__init__.py
+++ b/dynamics_learning/dynamics_learning/training/__init__.py
@@ -164,10 +164,10 @@ def train(
logger.info(f"Total layers in the model: {total_layers}")
# Define how many layers you want to remove
- layers_to_remove = 5
+ LAYERS_TO_REMOVE = 5
# Check if you can remove five layers; if not, adjust to delete fewer layers
- layers_to_remove = min(layers_to_remove, total_layers - 2)
+ layers_to_remove = min(LAYERS_TO_REMOVE, total_layers - 2)
# If it's a Sequential model
if isinstance(model, Sequential):
diff --git a/dynamics_learning/requirements.txt b/dynamics_learning/requirements.txt
index 201d0c1505efc2c32b1f17cedb3e1fbc1fb2f860..6454a395d31bb40fea5d70f6d379f7ff79cbfc5e 100644
--- a/dynamics_learning/requirements.txt
+++ b/dynamics_learning/requirements.txt
@@ -1,4 +1,4 @@
-coscine==0.10.7
+coscine>=0.11.5
wandb==0.17.3
tensorflow_text==2.12.0
tf_keras==2.14.1
diff --git a/dynamics_learning/train_instance.py b/dynamics_learning/train_instance.py
index 38fb8ccaa481e7019ed1fa082a353ffe8caa2a39..32cf2883c736c5ff9d27dc863ab000212df27502 100644
--- a/dynamics_learning/train_instance.py
+++ b/dynamics_learning/train_instance.py
@@ -1 +1,101 @@
-# 5vlv6m3t: ITA instance from scratch (used to train LLT instance)
+# download data from ita
+# train model like foundation model, but with other data
+# save model locally
+
+
+#!/usr/bin/env python3
+# -*- coding:utf-8 -*-
+# Copyright Leon Gorissen
+# Released under MIT License
+import warnings
+from functools import partial
+from pathlib import Path
+from dynamics_learning.environment import WANDB_API_TOKEN, WANDB_ENTITY, WANDB_PROJECT
+
+import numpy as np
+import tensorflow as tf
+from pritty_logger import RichLogger
+
+# import env variables and set tensorflow variables
+import wandb
+
+from dynamics_learning.preprocessing.dataset_analysis import analyze
+from dynamics_learning.preprocessing.trajectory_interpolation import interpolate
+from dynamics_learning.sweep.setup import setup_sweep
+
+# from dynamics_learning.data_retrieval import download_resource_content
+from dynamics_learning.training import train
+
+# Suppress FutureWarning for the specific deprecation warning in pandas
+warnings.simplefilter(action="ignore", category=FutureWarning)
+
+logger = RichLogger("dynamics_learning-instance_model")
+
+
+if __name__ == "__main__":
+ # download not existing data
+ # local_resource_path = download_resource_content()
+ # TODO download data from ita / use data from ita
+ local_resource_path = Path("/app/dynamics_learning/Trajectory Data")
+
+ # preprocess data
+ attained_data, command_data = analyze(local_resource_path / "train")
+ interpolated_command_data = interpolate(local_resource_path / "train")
+
+ # build input and cross validation tensors
+ interpolated_command_input = np.column_stack(
+ (
+ interpolated_command_data["q_interpolated_command"],
+ interpolated_command_data["qd_interpolated_command"],
+ interpolated_command_data["qdd_interpolated_command"],
+ )
+ )
+ q_qd_qdd_interpolated_command_input = tf.convert_to_tensor(
+ interpolated_command_input
+ )
+ tau_attained_input = tf.convert_to_tensor(attained_data["tau_attained"])
+
+ wandb.login(key=WANDB_API_TOKEN, relogin=True)
+ # check if sweep_id is set, if not create a new sweep
+ sweep_id, sweep_config = setup_sweep()
+
+ def train_save_upload(
+ q_qd_qdd_interpolated_command_input: tf.Tensor, tau_attained_input: tf.Tensor
+ ):
+ model, history, _run, config = train(
+ q_qd_qdd_interpolated_command_input, tau_attained_input
+ )
+ logger.info(
+ "\n=====================================\nModel trained\n=====================================\n"
+ )
+ # TODO save model and hyperparameters locally
+ # model = upload(model, history, config)
+ # logger.info(
+ # "\n=====================================\nModel uploaded\n=====================================\n"
+ # )
+ # # if a model was return it is the best performing model
+ # if model:
+ # logger.info("Evaluating current model with test data.")
+ # test_dataset = Dataset(type="test")
+ # test_dataset.download()
+ # interpolate("/app/dynamics_learning/Trajectory Data/test")
+ # model_analysis(test_dataset, model)
+ # "\n=====================================\nModel evaluated\n=====================================\n"
+ # else:
+ # logger.info("Model is not being evaluated.")
+ return None
+
+ train_save_upload_with_args = partial(
+ train_save_upload,
+ q_qd_qdd_interpolated_command_input=q_qd_qdd_interpolated_command_input,
+ tau_attained_input=tau_attained_input,
+ )
+
+ wandb.agent(
+ sweep_id,
+ train_save_upload_with_args,
+ project=WANDB_PROJECT,
+ entity=WANDB_ENTITY,
+ # count=NUM_MODELS,
+ notes="Instance model training using ita data.",
+ )