Skip to content

Segmentation runtime #13

@afilt

Description

@afilt

Hi @PierreMarza,

I'm running a segmentation task on segpath_epithelial with pre-computed embeddings:

Running dataset: segpath_epithelial
2026-04-13 23:06:52,459 - INFO -
🚀  Experiment Info
------------------------------------------------------------
Task   : segmentation
Dataset: segpath_epithelial
Model  : my_custom_vit_base_model
------------------------------------------------------------

Hyper-parameters
------------------------------------------------------------
type: frozen
batch_size: 64
num_workers: 40
lr: [0.001, 0.0001, 1e-05]
weight_decay: [0.0, 0.001, 0.0001]
epochs: 200
------------------------------------------------------------
2026-04-13 23:09:00,420 - INFO - Not found already trained best model. Training a model.

on 1 V100 GPUs (32 Gb) and 40 CPUs and have this terrible ETA:

  4%|▎         | 7/200 [9:28:22<260:53:27, 4866.36s/it]

How can I speed things up ? Thanks !

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions