SmolVLM-Base-ocr-isl-with-isl-backbone

Sigurdur/SmolVLM-Base-ocr-isl-with-isl-backbone

SmolVLM-Base-ocr-isl-with-isl-backbone is a machine learning model from Sigurdur. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for peft. The card lists the license as apache-2.0.

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

Downloads · 30 days

7% of all-time downloads

All-time downloads

150

Public

Repo size

343 MB

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.safetensors172 MB · 100%

At a glance

Library: peft
License: apache-2.0
Access: Public
Created: Dec 15, 2025
Updated: Dec 17, 2025
SHA: e122083a

Base models

HuggingFaceTB/SmolVLM-Base

Model card

From the Hugging Face model README

SmolVLM-Base-ocr-isl-with-isl-backbone

This model is a fine-tuned version of HuggingFaceTB/SmolVLM-Base on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.0147
Wer: 0.2907
Cer: 0.5314
Exact Match: 0.0
Special Char Acc: 1.0
Seq Acc 5: 0.0
Seq Acc 10: 0.0

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: 0.0003
train_batch_size: 4
eval_batch_size: 4
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 16
optimizer: Use OptimizerNames.PAGED_ADAMW_8BIT with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 500
num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer	Special Char Acc
0.3666	0.0325	125	0.2243	0.8704	0.8629	0.9622
0.168	0.0649	250	0.1485	0.7556	0.9306	0.9680
0.1282	0.0974	375	0.1187	0.4481	0.6394	0.9797
0.0984	0.1299	500	0.0965	0.5056	0.7014	0.9826
0.0891	0.1624	625	0.0755	0.4611	0.6485	0.9913
0.0744	0.1948	750	0.0638	0.4963	0.7116	0.9913
0.0708	0.2273	875	0.0518	0.3944	0.5805	0.9942
0.0647	0.2598	1000	0.0611	0.5389	0.8122	0.9855
0.0572	0.2922	1125	0.0454	0.4796	0.7158	0.9913
0.0555	0.3247	1250	0.0320	0.5685	0.7432	0.9884
0.0445	0.3572	1375	0.0386	0.4611	0.6404	0.9971
0.0455	0.3897	1500	0.0392	0.4259	0.6783	0.9913
0.0469	0.4221	1625	0.0319	0.2944	0.6415	0.9971
0.0386	0.4546	1750	0.0305	0.3574	0.5656	0.9971
0.0393	0.4871	1875	0.0327	0.2889	0.5405	0.9971
0.0364	0.5195	2000	0.0254	0.2685	0.4360	0.9971
0.0338	0.5520	2125	0.0254	0.2556	0.4616	1.0
0.0332	0.5845	2250	0.0217	0.3111	0.5105	1.0
0.0285	0.6170	2375	0.0257	0.3167	0.4898	1.0
0.0291	0.6494	2500	0.0230	0.4481	0.6054	1.0
0.028	0.6819	2625	0.0204	0.3741	0.5687	1.0
0.024	0.7144	2750	0.0204	0.3352	0.5242	0.9971
0.0262	0.7469	2875	0.0170	0.3796	0.6020	1.0
0.0237	0.7793	3000	0.0155	0.4222	0.6574	1.0
0.0246	0.8118	3125	0.0164	0.2796	0.5151	1.0
0.023	0.8443	3250	0.0152	0.2815	0.5054	1.0
0.0225	0.8767	3375	0.0150	0.2722	0.5151	1.0
0.0216	0.9092	3500	0.0153	0.2815	0.5268	1.0
0.0213	0.9417	3625	0.0149	0.2852	0.5300	1.0
0.0204	0.9742	3750	0.0147	0.2907	0.5314	1.0

Framework versions

PEFT 0.18.0
Transformers 4.57.3
Pytorch 2.9.1+cu128
Datasets 4.4.1
Tokenizers 0.22.1

Library: peft
License: apache-2.0
Created: Dec 15, 2025
Updated: Dec 17, 2025