[bugfix]: swap cpu-offloaded text encoders to GPU for the forward

text_encoder_cpu_offload means params rest on CPU between calls but compute
on GPU: FSDP-wrapped encoders (T5/llama/clip) already get this via
CPUOffloadPolicy streaming, but encoders without _fsdp_shard_conditions
(reason1/Qwen2.5-VL) were parked on CPU and, after the encoder-device input
routing added for kandinsky5, silently ran their 7B forward on CPU (crashing
outright with flash-attn installed). Move plain CPU-parked encoders to the
target device for the forward and back afterwards, mirroring the
image-encoder/VAE offload pattern; FSDP (DTensor) encoders keep the existing
input routing. Revert the examples to text_encoder_cpu_offload=True now that
the default works. Verified e2e: T2V/I2V outputs are metric-identical to
GPU-resident runs, with ~12GB lower VRAM during denoising.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

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=False,
+        text_encoder_cpu_offload=True,
         pin_cpu_memory=True,
         # image_encoder_cpu_offload=False,
     )

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=False,
+        text_encoder_cpu_offload=True,
         pin_cpu_memory=True,
         # image_encoder_cpu_offload=False,
     )