electra-finetuned

ucla-nb-project/electra-finetuned

electra-finetuned is a fill-mask model from ucla-nb-project. Use it when you need the model to fill a missing word. It is set up for transformers. The card lists the license as apache-2.0.

should probably proofread and complete it, then remove this comment. --

Downloads · 30 days

30% of all-time downloads

All-time downloads

Public

Parameters

33.7M

5.5 GB on disk

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.safetensors135 MB · 99%

At a glance

Task: Fill-Mask
Library: transformers
License: apache-2.0
Model type: electra
Access: Public
Created: May 18, 2024
Updated: May 18, 2024
SHA: 97121e7a

Try a prompt

Base models

google/electra-base-generator

Datasets

datasets/all_binary_and_xe_ey_fae_counterfactual

Model card

From the Hugging Face model README

electra-base-finetuned-xe_ey_fae

This model is a fine-tuned version of google/electra-base-generator on the datasets/all_binary_and_xe_ey_fae_counterfactual dataset. It achieves the following results on the evaluation set:

Loss: 1.7211
Accuracy: 0.6673

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 1e-05
train_batch_size: 8
eval_batch_size: 8
seed: 100
gradient_accumulation_steps: 2
total_train_batch_size: 16
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 3.0
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.5359	0.06	500	2.0696	0.6228
2.1807	0.13	1000	1.9677	0.6352
2.1028	0.19	1500	1.9192	0.6415
2.0658	0.26	2000	1.8923	0.6451
2.0426	0.32	2500	1.8699	0.6478
2.0133	0.39	3000	1.8580	0.6490
1.9978	0.45	3500	1.8411	0.6507
1.9862	0.52	4000	1.8297	0.6524
1.9745	0.58	4500	1.8154	0.6545
1.9606	0.64	5000	1.8056	0.6557
1.9486	0.71	5500	1.8033	0.6560
1.9416	0.77	6000	1.7894	0.6581
1.9279	0.84	6500	1.7848	0.6582
1.9196	0.9	7000	1.7786	0.6593
1.9168	0.97	7500	1.7762	0.6592
1.9123	1.03	8000	1.7744	0.6597
1.8942	1.1	8500	1.7625	0.6611
1.9053	1.16	9000	1.7576	0.6623
1.898	1.22	9500	1.7588	0.6620
1.8896	1.29	10000	1.7518	0.6625
1.8796	1.35	10500	1.7557	0.6619
1.8838	1.42	11000	1.7511	0.6628
1.8869	1.48	11500	1.7437	0.6640
1.8756	1.55	12000	1.7425	0.6641
1.8775	1.61	12500	1.7409	0.6641
1.8757	1.68	13000	1.7372	0.6649
1.8616	1.74	13500	1.7387	0.6646
1.8675	1.8	14000	1.7335	0.6648
1.8725	1.87	14500	1.7288	0.6660
1.8678	1.93	15000	1.7305	0.6659
1.8611	2.0	15500	1.7256	0.6666
1.853	2.06	16000	1.7286	0.6661
1.8487	2.13	16500	1.7285	0.6659
1.8543	2.19	17000	1.7229	0.6668
1.8519	2.26	17500	1.7240	0.6670
1.851	2.32	18000	1.7275	0.6662
1.8547	2.38	18500	1.7197	0.6673
1.8476	2.45	19000	1.7164	0.6675
1.8444	2.51	19500	1.7214	0.6676
1.8544	2.58	20000	1.7217	0.6668
1.8491	2.64	20500	1.7175	0.6678

Framework versions

Transformers 4.36.2
Pytorch 2.2.0+cu121
Datasets 2.17.0
Tokenizers 0.15.2

Task: Fill-Mask
Library: transformers
Type: electra
License: apache-2.0
Created: May 18, 2024
Updated: May 18, 2024