From 539884b9ba126b69685a23402d9d9fedb99d266d Mon Sep 17 00:00:00 2001 From: Will Lin Date: Fri, 3 Jul 2026 23:49:44 +0000 Subject: [PATCH] [bugfix]: keep kandinsky5 example text encoders on GPU text_encoder_cpu_offload=True makes the Qwen encoder compute on CPU, where flash-attn has no kernels; with flash_attn installed both examples crash in the first text-encoding forward. The offloaded path only worked in envs without flash_attn via the silent SDPA fallback. Co-Authored-By: Claude Fable 5 --- examples/inference/basic/basic_kandinsky5_i2v.py | 2 +- examples/inference/basic/basic_kandinsky5_t2v.py | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/examples/inference/basic/basic_kandinsky5_i2v.py b/examples/inference/basic/basic_kandinsky5_i2v.py index c30ca2f48..861ffedac 100644 --- a/examples/inference/basic/basic_kandinsky5_i2v.py +++ b/examples/inference/basic/basic_kandinsky5_i2v.py @@ -14,7 +14,7 @@ def main(): use_fsdp_inference=False, dit_cpu_offload=False, vae_cpu_offload=False, - text_encoder_cpu_offload=True, + text_encoder_cpu_offload=False, pin_cpu_memory=True, # image_encoder_cpu_offload=False, ) diff --git a/examples/inference/basic/basic_kandinsky5_t2v.py b/examples/inference/basic/basic_kandinsky5_t2v.py index a5da476d8..bfb52c044 100644 --- a/examples/inference/basic/basic_kandinsky5_t2v.py +++ b/examples/inference/basic/basic_kandinsky5_t2v.py @@ -12,7 +12,7 @@ def main(): use_fsdp_inference=False, dit_cpu_offload=False, vae_cpu_offload=False, - text_encoder_cpu_offload=True, + text_encoder_cpu_offload=False, pin_cpu_memory=True, # image_encoder_cpu_offload=False, )