YoshikiKubotani/slide_generation

Message	Author	SHA1	Date
:wrench: Rename the top-level config name	botany	793b3dc82e	1 year ago
:wrench: Update all the model configs to work with the new "model" group	botany	49feb8849c	1 year ago
:wrench: Make a new config group to manage all the training settings separately	botany	f40c062a52	1 year ago
:wrench: Change to manage the model configs in the same group	botany	0ddc9d69ad	1 year ago
:wrench: Rename the config name of models	botany	329d5b8cab	1 year ago
:wrench: Remove unused part in a schema	botany	c9981b0e75	1 year ago
:wrench: Update hydra configs to work with optax.inject_hyperparams	botany	6f1a6e3c6d	1 year ago
:sparkles: Reimplement the layout generator trainer	botany	59bdd3f4e5	1 year ago
:art: Remove an unused embedding and a parameter to switch whether to use it	botany	38549b68a7	1 year ago
:wrench: Update the default paramters for BLT training	botany	dac8f8470c	1 year ago
:wrench: Remove the unused parameter from the hydra config	botany	c59502e07f	1 year ago
:bug: Update the model initialization due to the modification of its structure through the bart training code development	botany	6941f5a9c9	1 year ago
:art: Use the same tokenize method for layout information as used in the bart training	botany	7ddb1b6e47	1 year ago
:bug: Use the connected one sentence as an input	botany	3bb4f2352c	1 year ago
:art: Use the same variable names	botany	2f1f9d7c9a	1 year ago
:sparkles: Update the bart model initialization to make manually added modules in its decoder recognizable from the flax init system	botany	277abea95c	1 year ago
:sparkles: Update the bart model to receive layout embedding and use it for attention calculation in the decoder	botany	4dfb445ee0	1 year ago
:sparkles: Change the bart input from the whole sentence to slide title related snippets	botany	0c959c09ae	1 year ago
:art: Change to use 4 spaces instead of 2 as tabs	botany	cfb59c6c88	1 year ago
:bug: Fix bug when padding the BART encoder inputs	botany	44eb6b2003	1 year ago
:wrench: Change qkv dimension of BLT training from 512 to 1024	botany	cf53c09b74	1 year ago
:bug: Fix bugs	botany	77f239843d	1 year ago
:bug: Make sure not to try to overwrite an immutable tuple	botany	4c64e30a0d	1 year ago
:art: Rename batches used for training to train_batch	botany	012d5ee48c	1 year ago
:art: Remove unnecessary variables	botany	82cc8d3fe9	1 year ago
:wrench: Change the dataset directory path	botany	575c0b9563	1 year ago
:art: Remove the initial_step variable	botany	fceab01c36	1 year ago
:sparkles: Allow BLT to receive BART encoder inputs	botany	14de902ad8	1 year ago
:sparkles: Add the feature to search snippets related to slide titles	botany	6b559c891c	1 year ago
:art: Clean the code of kl divergence for dataset generation	botany	d24b82ed0c	1 year ago

Newer Older

YoshikiKubotani / slide_generation connected to https://github.com/YoshikiKubotani/slide_generation.git

YoshikiKubotani
/
slide_generation
connected to https://github.com/YoshikiKubotani/slide_generation.git