Challenges in Migrating Imperative Deep Learning Programs to Graph Execution: An Empirical Study

File Descriptions

File | Description --- | --- commit_categorizations.csv | Categorizations for the commits in our dataset. commits.csv | Information for the commits in our dataset datasets.csv | Contains the names and descriptions of our datasets. issue_categorizations.csv | Categorizations for the chosen issues from our dataset. issues.csv | Information for the issues in our dataset. pipeline_stages.csv | DL pipeline stages and their respective descriptions. problem_categories.csv | Problem categories and their respective descriptions. problem_causes.csv | Problem causes and their respective descriptions. problem_fixes.csv | Problem fixes and their respective descriptions. problem_symptoms.csv | Problem symptoms and their respective descriptions. studied_subjects_commits.csv | Project data for commits. studied_subjects_issues.csv | Project data for issues.

Column Descriptions

commit_categorizations.csv

Column | Description --- | --- tf.function related fix? | TRUE when a bug fix related to tf.function was found and FALSE otherwise. If FALSE, subsequent column values will be blank. stage | DL pipeline stage where the problem fix was found.

issue_categorizations.csv

Column | Description --- | --- tf.function related problem? | TRUE when a bug related to tf.function was found and FALSE otherwise. If FALSE, subsequent column values will be blank. stage | DL pipeline stage where the problem was found. GH_id | GitHub issue unique identifier.

issues.csv

Column | Description --- | --- GH_id | GitHub issue unique identifier.