fix: run text encoders on MPS GPU instead of CPU for Apple Silicon (Comfy-Org#12809)

On Apple Silicon, `vram_state` is set to `VRAMState.SHARED` because
CPU and GPU share unified memory. However, `text_encoder_device()`
only checked for `HIGH_VRAM` and `NORMAL_VRAM`, causing all text
encoders to fall back to CPU on MPS devices.

Adding `VRAMState.SHARED` to the condition allows non-quantized text
encoders (e.g. bf16 Gemma 3 12B) to run on the MPS GPU, providing
significant speedup for text encoding and prompt generation.

Note: quantized models (fp4/fp8) that use float8_e4m3fn internally
will still fall back to CPU via the `supports_cast()` check in
`CLIP.__init__()`, since MPS does not support fp8 dtypes.

Author: k06a

comfy/model_management.py

@@ -1003,7 +1003,7 @@ def text_encoder_offload_device():
 def text_encoder_device():
     if args.gpu_only:
         return get_torch_device()
-    elif vram_state in (VRAMState.HIGH_VRAM, VRAMState.NORMAL_VRAM) or comfy.memory_management.aimdo_enabled:
+    elif vram_state in (VRAMState.HIGH_VRAM, VRAMState.NORMAL_VRAM, VRAMState.SHARED) or comfy.memory_management.aimdo_enabled:
         if should_use_fp16(prioritize_performance=False):
             return get_torch_device()
         else: