Ethanfel 8fade1b0e3 fix: initialize LoRA params on same device as wrapped linear ...

apply_lora() is called after generator.to(device), so lora_A/lora_B were
being created on CPU while the rest of the model was on CUDA.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

2026-04-05 22:17:29 +02:00

..

utils

fix: weights_only=False for SelVA checkpoints (PyTorch 2.6 compat)

2026-04-04 16:38:31 +02:00

__init__.py

chore: vendor selva_core from jnwnlee/selva@d7d40a9

2026-04-04 15:18:09 +02:00

embeddings.py

chore: vendor selva_core from jnwnlee/selva@d7d40a9

2026-04-04 15:18:09 +02:00

flow_matching.py

chore: vendor selva_core from jnwnlee/selva@d7d40a9

2026-04-04 15:18:09 +02:00

lora.py

fix: initialize LoRA params on same device as wrapped linear

2026-04-05 22:17:29 +02:00

low_level.py

chore: vendor selva_core from jnwnlee/selva@d7d40a9

2026-04-04 15:18:09 +02:00

networks_generator.py

chore: vendor selva_core from jnwnlee/selva@d7d40a9

2026-04-04 15:18:09 +02:00

networks_video_enc.py

chore: vendor selva_core from jnwnlee/selva@d7d40a9

2026-04-04 15:18:09 +02:00

sequence_config.py

chore: vendor selva_core from jnwnlee/selva@d7d40a9

2026-04-04 15:18:09 +02:00

text_synchformer.py

chore: vendor selva_core from jnwnlee/selva@d7d40a9

2026-04-04 15:18:09 +02:00

transformer_layers.py

chore: vendor selva_core from jnwnlee/selva@d7d40a9

2026-04-04 15:18:09 +02:00