The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
Fix MageFlow on cards that don't support bf16 (#15081)
Jukka Seppänen806e092ed427
diff --git a/comfy/ldm/mage_flow/model.py b/comfy/ldm/mage_flow/model.pyindex 92a5faa52..ac29bb610 100644--- a/comfy/ldm/mage_flow/model.py+++ b/comfy/ldm/mage_flow/model.py@@ -22,7 +22,6 @@ class MageTimestepProjEmbeddings(nn.Module):)def forward(self, timestep, hidden_states):- timestep = timestep.to(hidden_states.dtype)half_dim = 128exponent = -math.log(10000) * torch.arange(half_dim, dtype=torch.float32, device=timestep.device) / half_dimemb = torch.exp(exponent).to(timestep.dtype)diff --git a/comfy/model_base.py b/comfy/model_base.pyindex 50c73a431..ee6dc57a2 100644--- a/comfy/model_base.py+++ b/comfy/model_base.py@@ -2279,6 +2279,10 @@ class MageFlow(QwenImage):def __init__(self, model_config, model_type=ModelType.FLOW, device=None):super().__init__(model_config, model_type, device=device, unet_model=comfy.ldm.mage_flow.model.MageFlowTransformer2DModel)+ def process_timestep(self, timestep, **kwargs):+ # Mage runs in bf16 and rounds its timestep frequency table to the timestep dtype, keep that on fp32 devices.+ return timestep.to(torch.bfloat16)+def extra_conds_shapes(self, **kwargs):out = {}ref_latents = kwargs.get("reference_latents", None)