Returning error - ValueError: array of sample points is empty after removing values from pose.csv files

[v-moore]

I'm being returned the error above in VAME after running the vame.egocentric_alignment() command. In the .csv files moved to VAME, I cleared two columns of DLC tracked points, 6 total columns including the x and y points of the tracked points. The num_features variable and the pose_ref_index=[] were changed in the config file to reflect the change in the tracked points. Previously, running VAME on these datasets with no columns removed from the .csv files returned no error and the entire workflow was able to be run with no problem. Are there additional areas that should be updated to reflect the change in the .csv file?
I've attached a screenshot of the error and the config file.
I've tried reducing the pose_confidence as suggested in similar issues but the error has persisted.
Thanks for any help on sorting this out!

Project configurations

Project: CM-Flir-All-No-Tail
model_name: VAME
n_cluster: 15
pose_confidence: 0.99

Project path and videos

project_path: C:\Users\Sylwestrak Lab\Desktop\CM_FLIR_test\CM_FLIR-VM-2023-07-05\Ouput2\CM-Flir-All-No-Tail-Aug22-2023
video_sets:

• 20230619_MA_m440-03
• 20230619_MA_m440-05
• 20230619_MA_m444-02
• 20230619_MA_m444-03
• 20230619_MM_m440-03
• 20230619_MM_m440-05
• 20230619_MM_m444-02
• 20230619_MM_m444-03
• 20230619_SA_m440-04
• 20230619_SA_m440-06
• 20230619_SA_m444-01
• 20230619_SA_m444-04
• 20230619_SS_m440-04
• 20230619_SS_m440-06
• 20230619_SS_m444-01
• 20230619_SS_m444-04

Data

Data

all_data: yes

Creation of train set:

robust: true
iqr_factor: 4
axis:
savgol_filter: true
savgol_length: 5
savgol_order: 2
test_fraction: 0.1

RNN model general hyperparameter:

pretrained_model: None
pretrained_weights: false
num_features: 28
batch_size: 256
max_epochs: 500
model_snapshot: 50
model_convergence: 50
transition_function: GRU
beta: 1
beta_norm: false
zdims: 30
learning_rate: 0.0005
time_window: 30
prediction_decoder: 1
prediction_steps: 15
noise: false
scheduler: 1
scheduler_step_size: 100
scheduler_gamma: 0.2
softplus: false

Segmentation:

load_data: -PE-seq-clean
individual_parameterization: false
random_state_kmeans: 42
n_init_kmeans: 15

Video writer:

length_of_motif_video: 1000

UMAP parameter:

min_dist: 0.1
n_neighbors: 200
random_state: 42
num_points: 90000

ONLY CHANGE ANYTHING BELOW IF YOU ARE FAMILIAR WITH RNN MODELS

RNN encoder hyperparamter:

hidden_size_layer_1: 256
hidden_size_layer_2: 256
dropout_encoder: 0

RNN reconstruction hyperparameter:

hidden_size_rec: 256
dropout_rec: 0
n_layers: 1

RNN prediction hyperparamter:

hidden_size_pred: 256
dropout_pred: 0

RNN loss hyperparameter:

mse_reconstruction_reduction: sum
mse_prediction_reduction: sum
kmeans_loss: 30
kmeans_lambda: 0.1
anneal_function: linear
kl_start: 2
annealtime: 4

Legacy mode

legacy: false