CeLLaTe-ner-3class-pubmedbert-tapt-tokenizer-original-spanmask

Mardiyyah/CeLLaTe-ner-3class-pubmedbert-tapt-tokenizer-original-spanmask

CeLLaTe-ner-3class-pubmedbert-tapt-tokenizer-original-spanmask is a token classification model from Mardiyyah. Use it when you need labels on individual words, such as names. 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

26% of all-time downloads

All-time downloads

Public

Parameters

109M

436 MB on disk

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.safetensors436 MB · 100%

At a glance

Task: Token Classification
Library: transformers
License: apache-2.0
Model type: bert
Access: Public
Created: Jul 27, 2026
Updated: Jul 27, 2026
SHA: f629260c

Try a prompt

Base models

Mardiyyah/CeLLaTe-tapt-pubmedbert-tokenizer-original-spanmask

Spaces

OTAR3088/CeLLaTe

Model card

From the Hugging Face model README

CeLLaTe-ner-3class-pubmedbert-tapt-tokenizer-original-spanmask

This model is a fine-tuned version of Mardiyyah/CeLLaTe-tapt-pubmedbert-tokenizer-original-spanmask on the OTAR3088/CeLLaTe-ner-3class-iob_final dataset. It achieves the following results on the evaluation set:

Loss: 0.1137
Precision: 0.7609
Recall: 0.7553
Micro F1: 0.7581
Weighted F1: 0.7577
Macro F1: 0.7544
Accuracy: 0.9823

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: 2e-05
train_batch_size: 16
eval_batch_size: 16
seed: 3407
gradient_accumulation_steps: 2
total_train_batch_size: 32
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.01
num_epochs: 20
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	Micro F1	Weighted F1	Macro F1	Accuracy
0.3682	1.0	263	0.0777	0.5310	0.6283	0.5756	0.5686	0.5479	0.9754
0.056	2.0	526	0.0636	0.7462	0.7100	0.7277	0.7276	0.7262	0.9812
0.0351	3.0	789	0.0682	0.7267	0.7309	0.7288	0.7294	0.7271	0.9809
0.0242	4.0	1052	0.0756	0.7670	0.7163	0.7408	0.7400	0.7365	0.9819
0.0185	5.0	1315	0.0741	0.7752	0.7179	0.7454	0.7455	0.7425	0.9820
0.0144	6.0	1578	0.0816	0.7153	0.7350	0.7250	0.7250	0.7224	0.9808
0.0112	7.0	1841	0.0887	0.7523	0.7288	0.7403	0.7389	0.7359	0.9818
0.0091	8.0	2104	0.0913	0.7351	0.7236	0.7293	0.7299	0.7290	0.9812
0.0078	9.0	2367	0.0939	0.7615	0.7397	0.7505	0.7497	0.7463	0.9820
0.0066	10.0	2630	0.1002	0.7700	0.7163	0.7422	0.7414	0.7388	0.9818
0.0056	11.0	2893	0.1002	0.7560	0.7356	0.7456	0.7450	0.7413	0.9818
0.0049	12.0	3156	0.1047	0.7563	0.7543	0.7553	0.7554	0.7513	0.9822
0.0042	13.0	3419	0.1077	0.7561	0.7522	0.7542	0.7532	0.7495	0.9822
0.0037	14.0	3682	0.1076	0.7651	0.7408	0.7527	0.7522	0.7482	0.9821
0.0032	15.0	3945	0.1100	0.7404	0.7616	0.7508	0.7504	0.7470	0.9818
0.0027	16.0	4208	0.1098	0.7575	0.7527	0.7551	0.7548	0.7516	0.9823
0.0025	17.0	4471	0.1143	0.7635	0.7376	0.7503	0.7500	0.7474	0.9819
0.0024	18.0	4734	0.1145	0.7444	0.7548	0.7495	0.7493	0.7468	0.9818
0.002	19.0	4997	0.1141	0.7609	0.7553	0.7581	0.7577	0.7544	0.9823
0.002	20.0	5260	0.1137	0.7578	0.7574	0.7576	0.7573	0.7538	0.9821

Framework versions

Transformers 4.48.2
Pytorch 2.4.1+cu121
Datasets 3.0.2
Tokenizers 0.21.0

Task: Token Classification
Library: transformers
Type: bert
License: apache-2.0
Languages: en
Created: Jul 27, 2026
Updated: Jul 27, 2026