wangbo/ComfyUI
1
0
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2026-01-21 20:00:17 +08:00
Code Issues Actions 9 Packages Projects Releases Wiki Activity

Merge branch 'comfyanonymous:master' into master

Browse Source
This commit is contained in:
patientx 2025-12-04 23:49:41 +03:00 committed by GitHub
commit f270afdff3
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 5 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
comfy/sd.py
View File

@ -471,7 +471,7 @@ class VAE:
decoder_config={'target': "comfy.ldm.hunyuan_video.vae_refiner.Decoder", 'params': ddconfig}) decoder_config={'target': "comfy.ldm.hunyuan_video.vae_refiner.Decoder", 'params': ddconfig})
self.memory_used_encode = lambda shape, dtype: (1400 * 9 * shape[-2] * shape[-1]) * model_management.dtype_size(dtype) self.memory_used_encode = lambda shape, dtype: (1400 * 9 * shape[-2] * shape[-1]) * model_management.dtype_size(dtype)
self.memory_used_decode = lambda shape, dtype: (2800 * 4 * shape[-2] * shape[-1] * 16 * 16) * model_management.dtype_size(dtype) self.memory_used_decode = lambda shape, dtype: (3600 * 4 * shape[-2] * shape[-1] * 16 * 16) * model_management.dtype_size(dtype)
elif "decoder.conv_in.conv.weight" in sd: elif "decoder.conv_in.conv.weight" in sd:
ddconfig = {'double_z': True, 'z_channels': 4, 'resolution': 256, 'in_channels': 3, 'out_ch': 3, 'ch': 128, 'ch_mult': [1, 2, 4, 4], 'num_res_blocks': 2, 'attn_resolutions': [], 'dropout': 0.0} ddconfig = {'double_z': True, 'z_channels': 4, 'resolution': 256, 'in_channels': 3, 'out_ch': 3, 'ch': 128, 'ch_mult': [1, 2, 4, 4], 'num_res_blocks': 2, 'attn_resolutions': [], 'dropout': 0.0}
ddconfig["conv3d"] = True ddconfig["conv3d"] = True
@ -483,8 +483,10 @@ class VAE:
self.latent_dim = 3 self.latent_dim = 3
self.latent_channels = ddconfig['z_channels'] = sd["decoder.conv_in.conv.weight"].shape[1] self.latent_channels = ddconfig['z_channels'] = sd["decoder.conv_in.conv.weight"].shape[1]
self.first_stage_model = AutoencoderKL(ddconfig=ddconfig, embed_dim=sd['post_quant_conv.weight'].shape[1]) self.first_stage_model = AutoencoderKL(ddconfig=ddconfig, embed_dim=sd['post_quant_conv.weight'].shape[1])
self.memory_used_decode = lambda shape, dtype: (1500 * shape[2] * shape[3] * shape[4] * (4 * 8 * 8)) * model_management.dtype_size(dtype) #This is likely to significantly over-estimate with single image or low frame counts as the
self.memory_used_encode = lambda shape, dtype: (900 * max(shape[2], 2) * shape[3] * shape[4]) * model_management.dtype_size(dtype) #implementation is able to completely skip caching. Rework if used as an image only VAE
self.memory_used_decode = lambda shape, dtype: (2800 * min(8, ((shape[2] - 1) * 4) + 1) * shape[3] * shape[4] * (8 * 8)) * model_management.dtype_size(dtype)
self.memory_used_encode = lambda shape, dtype: (1400 * min(9, shape[2]) * shape[3] * shape[4]) * model_management.dtype_size(dtype)
self.working_dtypes = [torch.bfloat16, torch.float16, torch.float32] self.working_dtypes = [torch.bfloat16, torch.float16, torch.float32]
elif "decoder.unpatcher3d.wavelets" in sd: elif "decoder.unpatcher3d.wavelets" in sd:
self.upscale_ratio = (lambda a: max(0, a * 8 - 7), 8, 8) self.upscale_ratio = (lambda a: max(0, a * 8 - 7), 8, 8)