README - long-t5-tglobal-base-16384-booksum-V11-big_patent-V2

this README was added because there wasn't one
created 2022-07-31_12-14-50

about

An experiment testing some transfer learning with pszemraj/long-t5-tglobal-base-16384-book-summary to evaluate the ability to learn some technical documentation through the big_patent dataset on huggingface.

This checkpoint has been trained on dataset subsection y of big_patent for approx 400 steps of functional batch size 128.

Downloads last month: 29

Safetensors

Model size

0.2B params

Tensor type

F32

Datasets used to train pszemraj/long-t5-tglobal-base-16384-booksum-V11-big_patent-V2

Space using pszemraj/long-t5-tglobal-base-16384-booksum-V11-big_patent-V2 1

Evaluation results

ROUGE-1 on kmfoda/booksum
test set verified

23.144
ROUGE-2 on kmfoda/booksum
test set verified

3.239
ROUGE-L on kmfoda/booksum
test set verified

12.704
ROUGE-LSUM on kmfoda/booksum
test set verified

19.810
loss on kmfoda/booksum
test set verified

2.766
gen_len on kmfoda/booksum
test set verified

63.449
ROUGE-1 on samsum
test set verified

26.803
ROUGE-2 on samsum
test set verified

6.066