The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
Use fp16 as the default vae dtype for the audio VAE.
comfyanonymous6425252c4f2f
diff --git a/comfy/model_management.py b/comfy/model_management.pyindex 8b8d3ff06..047193290 100644--- a/comfy/model_management.py+++ b/comfy/model_management.py@@ -167,7 +167,7 @@ if args.use_pytorch_cross_attention:ENABLE_PYTORCH_ATTENTION = TrueXFORMERS_IS_AVAILABLE = False-VAE_DTYPE = torch.float32+VAE_DTYPES = [torch.float32]try:if is_nvidia():@@ -176,7 +176,7 @@ try:if ENABLE_PYTORCH_ATTENTION == False and args.use_split_cross_attention == False and args.use_quad_cross_attention == False:ENABLE_PYTORCH_ATTENTION = Trueif torch.cuda.is_bf16_supported() and torch.cuda.get_device_properties(torch.cuda.current_device()).major >= 8:- VAE_DTYPE = torch.bfloat16+ VAE_DTYPES = [torch.bfloat16] + VAE_DTYPESif is_intel_xpu():if args.use_split_cross_attention == False and args.use_quad_cross_attention == False:ENABLE_PYTORCH_ATTENTION = True@@ -184,17 +184,10 @@ except:passif is_intel_xpu():- VAE_DTYPE = torch.bfloat16+ VAE_DTYPES = [torch.bfloat16] + VAE_DTYPESif args.cpu_vae:- VAE_DTYPE = torch.float32--if args.fp16_vae:- VAE_DTYPE = torch.float16-elif args.bf16_vae:- VAE_DTYPE = torch.bfloat16-elif args.fp32_vae:- VAE_DTYPE = torch.float32+ VAE_DTYPES = [torch.float32]if ENABLE_PYTORCH_ATTENTION:@@ -258,7 +251,6 @@ try:except:logging.warning("Could not pick default device.")-logging.info("VAE dtype: {}".format(VAE_DTYPE))current_loaded_models = []@@ -619,9 +611,22 @@ def vae_offload_device():else:return torch.device("cpu")-def vae_dtype():- global VAE_DTYPE- return VAE_DTYPE+def vae_dtype(device=None, allowed_dtypes=[]):+ global VAE_DTYPES+ if args.fp16_vae:+ return torch.float16+ elif args.bf16_vae:+ return torch.bfloat16+ elif args.fp32_vae:+ return torch.float32++ for d in allowed_dtypes:+ if d == torch.float16 and should_use_fp16(device, prioritize_performance=False):+ return d+ if d in VAE_DTYPES:+ return d++ return VAE_DTYPES[0]def get_autocast_device(dev):if hasattr(dev, 'type'):diff --git a/comfy/sd.py b/comfy/sd.pyindex e16cd8e53..58a858aa0 100644--- a/comfy/sd.py+++ b/comfy/sd.py@@ -178,6 +178,7 @@ class VAE:self.output_channels = 3self.process_input = lambda image: image * 2.0 - 1.0self.process_output = lambda image: torch.clamp((image + 1.0) / 2.0, min=0.0, max=1.0)+ self.working_dtypes = [torch.bfloat16, torch.float32]if config is None:if "decoder.mid.block_1.mix_factor" in sd:@@ -245,6 +246,7 @@ class VAE:self.downscale_ratio = 2048self.process_output = lambda audio: audioself.process_input = lambda audio: audio+ self.working_dtypes = [torch.float16, torch.bfloat16, torch.float32]else:logging.warning("WARNING: No VAE weights detected, VAE not initalized.")self.first_stage_model = None@@ -265,12 +267,13 @@ class VAE:self.device = deviceoffload_device = model_management.vae_offload_device()if dtype is None:- dtype = model_management.vae_dtype()+ dtype = model_management.vae_dtype(self.device, self.working_dtypes)self.vae_dtype = dtypeself.first_stage_model.to(self.vae_dtype)self.output_device = model_management.intermediate_device()self.patcher = comfy.model_patcher.ModelPatcher(self.first_stage_model, load_device=self.device, offload_device=offload_device)+ logging.debug("VAE load device: {}, offload device: {}, dtype: {}".format(self.device, offload_device, self.vae_dtype))def vae_encode_crop_pixels(self, pixels):dims = pixels.shape[1:-1]