nodes/cubiq-ComfyUI_IPAdapter_plus
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

hopefully found the right combination for dtype

Browse Source
This commit is contained in:
matt3o
2023-12-17 20:02:43 +01:00
parent cb1c5e8a1a
commit a4e71ccbbf
1 changed files with 5 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
IPAdapterPlus.py
+5 -6
View File
@@ -411,8 +411,7 @@ class IPAdapterApply:
CATEGORY = "ipadapter"
def apply_ipadapter(self, ipadapter, model, weight, clip_vision=None, image=None, weight_type="original", noise=None, embeds=None, attn_mask=None, start_at=0.0, end_at=1.0, unfold_batch=False):
self.dtype = model.model.diffusion_model.dtype
self.embeds_dtype = comfy.model_management.text_encoder_dtype(comfy.model_management.text_encoder_device()) # text and image embeds should have the same dtype
self.dtype = torch.float16 if comfy.model_management.should_use_fp16() else torch.float32
self.device = comfy.model_management.get_torch_device()
self.weight = weight
self.is_full = "proj.0.weight" in ipadapter["image_proj"]
@@ -460,11 +459,11 @@ class IPAdapterApply:
is_full=self.is_full,
)
self.ipadapter.to(self.device, dtype=self.embeds_dtype)
self.ipadapter.to(self.device, dtype=self.dtype)
image_prompt_embeds, uncond_image_prompt_embeds = self.ipadapter.get_image_embeds(clip_embed.to(self.device, dtype=self.embeds_dtype), clip_embed_zeroed.to(self.device, dtype=self.embeds_dtype))
image_prompt_embeds = image_prompt_embeds.to(self.device, dtype=self.embeds_dtype)
uncond_image_prompt_embeds = uncond_image_prompt_embeds.to(self.device, dtype=self.embeds_dtype)
image_prompt_embeds, uncond_image_prompt_embeds = self.ipadapter.get_image_embeds(clip_embed.to(self.device, dtype=self.dtype), clip_embed_zeroed.to(self.device, dtype=self.dtype))
image_prompt_embeds = image_prompt_embeds.to(self.device, dtype=self.dtype)
uncond_image_prompt_embeds = uncond_image_prompt_embeds.to(self.device, dtype=self.dtype)
work_model = model.clone()