added notebooks and scripts to reranking using graph features (catboo… #146
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
DmiitriyJarosh
approved these changes
Mar 16, 2024
| for _, row in tqdm(dataframe.iterrows()): | ||
| # convert from json dict to networkx graph | ||
| graph_obj = json_graph.node_link_graph(try_literal_eval(row["graph"])) | ||
| graph_node_names = get_node_names(graph_obj) |
There was a problem hiding this comment.
is it used anywhere except graph_to_sequence? maybe call it there?
| def get_distance_ans_cand(graph, ans_cand_id): | ||
| """get avg distance from ans entity to answer candidate""" | ||
| graph = graph.to_undirected() # for ssp both ways | ||
| ssp_dict = nx.shortest_path(graph, target=ans_cand_id) |
There was a problem hiding this comment.
can our graph contain intermediate nodes? or only 1-hop?
| "updated_graph_sequence_embedding", | ||
| ] | ||
| if args.sequence_type != "both": | ||
| DROP_EM_FEAT = ( |
There was a problem hiding this comment.
looks like you do the same thing as in catboost but in other way -- it is better to use the same approach for better readability
highly0
force-pushed
the
features/graph_features_reranking
branch
from
5cd76d1 to
0f484b1
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
I'm adding scripts to run the reranking pipeline using graph features (numerical, textual, and embedding features). There are also jupyter notebooks alternatives. Note: the scripts only train the model without reranking for now. Thus, converting the
.ipynbfiles and running from top to bottom will do all training, reranking, and features importance.Remember to change configurations before running the
.ipynbfiles. Or alternatively, you can train the models with the scripts, then load it in to theipynbfile and gather reranking results.In addition to the reranking pipeline, I added
graph_features_preparation.pywhich prepares the dataframe with graph features and publish it to HuggingFace (T5-large-ssm, T5-xl-ssm).The added notebooks and scripts and its functionality are as below:
graph_features_preparation.py: prepare the dataset with graph features and publish to HFlinear_regression.ipynb: notebooks to train linear regression and rerank with dataset above (can be ran from top to bottom)train_linear_regression.py: script to train linear regression using dataset abovecatboost_features.ipynb: notebooks to train catboost and rerank with dataset above (can be ran from top to bottom)train_catboost_regressor.py: script to train catboost using dataset above