The model was fine tuned on a dataset of 1000+ quarterly 13F reports. It is intended for use with automating the generation of summaries of articles before they are published. This allows you to put in a TL;DR summary without having to write one on your own.

NOTE: The HuggingFace hosted Inference API interface takes the default parameters and so only outputs about 20 words of text. To get a full summary, use the Inference API directly and pass in max_length=120 or so.

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5.6e-05
train_batch_size: 8
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 8

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
11.4662	1.0	126	2.9329	0.2023	0.0998	0.1717	0.1792
3.4401	2.0	252	1.9914	0.3142	0.2573	0.3015	0.3036
2.5139	3.0	378	1.7493	0.3131	0.2576	0.3022	0.3039
2.152	4.0	504	1.6465	0.3114	0.2564	0.3009	0.3024
1.9624	5.0	630	1.5607	0.3202	0.2695	0.3114	0.3127
1.851	6.0	756	1.5163	0.3205	0.2704	0.3101	0.311
1.8002	7.0	882	1.4848	0.3225	0.2718	0.3148	0.3161
1.7864	8.0	1008	1.4818	0.3235	0.2725	0.3146	0.3161

Framework versions

Transformers 4.34.0
Pytorch 2.0.1+cu118
Datasets 2.14.5
Tokenizers 0.14.0

jkv53/mt5-small-finetuned-13f-reports

mt5-small-finetuned-13f-reports

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

jkv53/
mt5-small-finetuned-13f-reports