Skip to content

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.

Add Flux fp16 support hack.

comfyanonymous8115d8cce97a
diff --git a/comfy/ldm/flux/layers.py b/comfy/ldm/flux/layers.py
index 99f498106..4a0bd40c6 100644
--- a/comfy/ldm/flux/layers.py
+++ b/comfy/ldm/flux/layers.py
@@ -188,6 +188,10 @@ class DoubleStreamBlock(nn.Module):
# calculate the txt bloks
txt = txt + txt_mod1.gate * self.txt_attn.proj(txt_attn)
txt = txt + txt_mod2.gate * self.txt_mlp((1 + txt_mod2.scale) * self.txt_norm2(txt) + txt_mod2.shift)
+
+ if txt.dtype == torch.float16:
+ txt = txt.clip(-65504, 65504)
+
return img, txt
@@ -239,7 +243,10 @@ class SingleStreamBlock(nn.Module):
attn = attention(q, k, v, pe=pe)
# compute activation in mlp stream, cat again and run second linear layer
output = self.linear2(torch.cat((attn, self.mlp_act(mlp)), 2))
- return x + mod.gate * output
+ x = x + mod.gate * output
+ if x.dtype == torch.float16:
+ x = x.clip(-65504, 65504)
+ return x
class LastLayer(nn.Module):
diff --git a/comfy/supported_models.py b/comfy/supported_models.py
index 6cecb9a02..d07a7106c 100644
--- a/comfy/supported_models.py
+++ b/comfy/supported_models.py
@@ -642,7 +642,7 @@ class Flux(supported_models_base.BASE):
memory_usage_factor = 2.8
- supported_inference_dtypes = [torch.bfloat16, torch.float32]
+ supported_inference_dtypes = [torch.bfloat16, torch.float16, torch.float32]
vae_key_prefix = ["vae."]
text_encoder_key_prefix = ["text_encoders."]