abstractor

Awni00's group workspace

Group: Pre-training Task (reshuffled attr); Transformer

1-1

of 1

Tags

Notes

Author

awni00

State

Finished

Start time

April 28th, 2023 10:40:17 PM

Runtime

1m 50s

Tracked hours

1m 43s

Run path

abstractor/object_argsort_autoregressive/wt4ylpbt

Linux-3.10.0-1160.80.1.el7.x86_64-x86_64-with-glibc2.10

Python version

3.8.16

Git repository

git clone https://github.com/jdlafferty/relational

Git state

git checkout -b "pretraining_mode=pretraining" 423329254a678a07b91af59cf8570e48942bbf54

Command

evaluate_argsort_model_learning_curves.py --model transformer --pretraining_mode pretraining --init_trainable True --pretraining_task_type "reshuffled attr" --pretraining_task_data_path object_sorting_datasets/product_structure_reshuffled_object_sort_dataset.npy --eval_task_data_path object_sorting_datasets/product_structure_object_sort_dataset.npy --n_epochs 200 --early_stopping True --min_train_size 100 --max_train_size 3000 --train_size_step 100 --num_trials 10 --start_trial 0 --pretraining_train_size 1000 --wandb_project_name object_argsort_autoregressive

System Hardware

CPU count	36
Logical CPU count	36
GPU count	1
GPU type	NVIDIA GeForce RTX 3090

W&B CLI Version

0.13.9

Group

Pre-training Task (reshuffled attr); Transformer

Config parameters are your model's inputs. Learn more

▶
Config parameters:{} 3 keys
- group:
  "Pre-training Task (reshuffled attr); Transformer"
- pretraining_mode:
  "pretraining"
- train size:
  87,500

Summary metrics are your model's outputs. Learn more

▶
Summary metrics:{} 10 keys
- acc_by_position_table:
  "table-file"
- elementwise_accuracy:
  0.122345
- epoch/epoch:
  199
- epoch/learning_rate:
  0.0010000000474974513
- epoch/loss:
  0.5213602185249329
- epoch/teacher_forcing_accuracy:
  0.8123999834060669
- epoch/val_loss:
  2.6465744972229004
- epoch/val_teacher_forcing_accuracy:
  0.3239000141620636
- full_sequence_accuracy:
  0.0077
- teacher_forcing_accuracy:
  0.32811427116394043

This run consumed these artifacts as inputs. Learn more

job

job-https___github.com_jdlafferty_relational_experiments_object_argsort_autoregressive_evaluate_argsort_model_learning_curves.py:v1

This run produced these artifacts as outputs. Total: 2. Learn more

run_table

run-wt4ylpbt-acc_by_position_table:v0

wandb-history

run-wt4ylpbt-history:v0