Search results

[BUG] Pipeline model parallel > 1 does fails Evo2 training

bugSomething isn't workingSomething isn't working

Evo2

Status: Open.

#815 In NVIDIA/bionemo-framework;

· dorotat-nv opened on Apr 9, 2025

[BUG] Evo2 training does not work with tp>1 when num_attention_heads is not divisible by tp

bugSomething isn't workingSomething isn't working

Evo2

Status: Open.

#814 In NVIDIA/bionemo-framework;

· dorotat-nv opened on Apr 9, 2025

[BUG] failing test test_train_evo2_stop_at_max_steps_and_continue for Evo2

bugSomething isn't workingSomething isn't working

Evo2

Status: Open.

#769 In NVIDIA/bionemo-framework;

· dorotat-nv opened on Mar 19, 2025

[BUG] Failing test_train_evo2_stops unit test

bugSomething isn't workingSomething isn't working

Evo2

Status: Open.

#760 In NVIDIA/bionemo-framework;

· dorotat-nv opened on Mar 17, 2025

Move unit test test_hyena_operators to NeMo

enhancementNew feature or requestNew feature or request

Evo2

Status: Open.

#750 In NVIDIA/bionemo-framework;

· dorotat-nv opened on Mar 12, 2025

[Feature] Allow early training termination at specific step using Trainer.max_steps without modifying LR schedule ie from full convergence run

Evo2

Status: Open.

#749 In NVIDIA/bionemo-framework;

· dorotat-nv opened on Mar 12, 2025