bymyself 49b6313322 test: cover Gemma4 channel-aware close marker decoding ...

The LTX2 prompt node primes a 'final' channel, so only its close is generated. Decoding that close as </think> made the whole answer read as reasoning and the node returned an empty string.

2026-08-16 18:30:34 -07:00

..

folder_path_test.py

add models-directory launch argument (#9113)

2026-07-08 22:20:47 +08:00

gemma4_decode_test.py

test: cover Gemma4 channel-aware close marker decoding

2026-08-16 18:30:34 -07:00

gemma4_template_test.py

Fix Generate Text ignoring thinking=false on Gemma4 E2B/E4B (#15278)

2026-08-12 22:44:14 -04:00

model_detection_test.py

[feat]Add JoyImageEdit native model support (#14428)

2026-07-15 23:48:28 -04:00

seedvr_vae_forward_test.py

Add SeedVR2 support (CORE-6) (#14424)

2026-07-10 15:07:42 +08:00

test_seedvr2_dtype.py

Try to fix some issues with the seedvr VAE. (#14877)

2026-07-10 19:54:28 -04:00

test_seedvr2_internals.py

Add SeedVR2 support (CORE-6) (#14424)

2026-07-10 15:07:42 +08:00

test_seedvr2_model.py

Add SeedVR2 support (CORE-6) (#14424)

2026-07-10 15:07:42 +08:00

test_seedvr2_vae_decode.py

Add SeedVR2 support (CORE-6) (#14424)

2026-07-10 15:07:42 +08:00

test_seedvr2_vae_tiled.py

Try to fix some issues with the seedvr VAE. (#14877)

2026-07-10 19:54:28 -04:00

test_vae_decode_tiled_nested.py

Fix VAEDecodeTiled crash on NestedTensor latents (MiniMax H3) (#15477)

2026-08-10 23:00:23 -04:00