netadmin/test
0
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

sync with repo 28.08

Browse Source
This commit is contained in:
netadmin
2024-08-28 19:33:34 +03:00
parent 727693318c
commit ad1e3ecbcb
134 changed files with 112534 additions and 12635 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
comfy/model_detection.py
View File

@@ -131,14 +131,14 @@ def detect_unet_config(state_dict, key_prefix):
if '{}double_blocks.0.img_attn.norm.key_norm.scale'.format(key_prefix) in state_dict_keys: #Flux
dit_config = {}
dit_config["image_model"] = "flux"
dit_config["in_channels"] = 64
dit_config["in_channels"] = 16
dit_config["vec_in_dim"] = 768
dit_config["context_in_dim"] = 4096
dit_config["hidden_size"] = 3072
dit_config["mlp_ratio"] = 4.0
dit_config["num_heads"] = 24
dit_config["depth"] = 19
dit_config["depth_single_blocks"] = 38
dit_config["depth"] = count_blocks(state_dict_keys, '{}double_blocks.'.format(key_prefix) + '{}.')
dit_config["depth_single_blocks"] = count_blocks(state_dict_keys, '{}single_blocks.'.format(key_prefix) + '{}.')
dit_config["axes_dim"] = [16, 56, 56]
dit_config["theta"] = 10000
dit_config["qkv_bias"] = True
@@ -472,9 +472,15 @@ def unet_config_from_diffusers_unet(state_dict, dtype=None):
'transformer_depth': [0, 1, 1], 'channel_mult': [1, 2, 4], 'transformer_depth_middle': -2, 'use_linear_in_transformer': False,
'context_dim': 768, 'num_head_channels': 64, 'transformer_depth_output': [0, 0, 1, 1, 1, 1],
'use_temporal_attention': False, 'use_temporal_resblock': False}
SD15_diffusers_inpaint = {'use_checkpoint': False, 'image_size': 32, 'out_channels': 4, 'use_spatial_transformer': True, 'legacy': False, 'adm_in_channels': None,
'dtype': dtype, 'in_channels': 9, 'model_channels': 320, 'num_res_blocks': [2, 2, 2, 2], 'transformer_depth': [1, 1, 1, 1, 1, 1, 0, 0],
'channel_mult': [1, 2, 4, 4], 'transformer_depth_middle': 1, 'use_linear_in_transformer': False, 'context_dim': 768, 'num_heads': 8,
'transformer_depth_output': [1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0],
'use_temporal_attention': False, 'use_temporal_resblock': False}
supported_models = [SDXL, SDXL_refiner, SD21, SD15, SD21_uncliph, SD21_unclipl, SDXL_mid_cnet, SDXL_small_cnet, SDXL_diffusers_inpaint, SSD_1B, Segmind_Vega, KOALA_700M, KOALA_1B, SD09_XS, SD_XS, SDXL_diffusers_ip2p]
supported_models = [SDXL, SDXL_refiner, SD21, SD15, SD21_uncliph, SD21_unclipl, SDXL_mid_cnet, SDXL_small_cnet, SDXL_diffusers_inpaint, SSD_1B, Segmind_Vega, KOALA_700M, KOALA_1B, SD09_XS, SD_XS, SDXL_diffusers_ip2p, SD15_diffusers_inpaint]
for unet_config in supported_models:
matches = True
@@ -495,7 +501,12 @@ def model_config_from_diffusers_unet(state_dict):
def convert_diffusers_mmdit(state_dict, output_prefix=""):
out_sd = {}
if 'transformer_blocks.0.attn.add_q_proj.weight' in state_dict: #SD3
if 'transformer_blocks.0.attn.norm_added_k.weight' in state_dict: #Flux
depth = count_blocks(state_dict, 'transformer_blocks.{}.')
depth_single_blocks = count_blocks(state_dict, 'single_transformer_blocks.{}.')
hidden_size = state_dict["x_embedder.bias"].shape[0]
sd_map = comfy.utils.flux_to_diffusers({"depth": depth, "depth_single_blocks": depth_single_blocks, "hidden_size": hidden_size}, output_prefix=output_prefix)
elif 'transformer_blocks.0.attn.add_q_proj.weight' in state_dict: #SD3
num_blocks = count_blocks(state_dict, 'transformer_blocks.{}.')
depth = state_dict["pos_embed.proj.weight"].shape[0] // 64
sd_map = comfy.utils.mmdit_to_diffusers({"depth": depth, "num_blocks": num_blocks}, output_prefix=output_prefix)
@@ -521,7 +532,12 @@ def convert_diffusers_mmdit(state_dict, output_prefix=""):
old_weight = out_sd.get(t[0], None)
if old_weight is None:
old_weight = torch.empty_like(weight)
old_weight = old_weight.repeat([3] + [1] * (len(old_weight.shape) - 1))
if old_weight.shape[offset[0]] < offset[1] + offset[2]:
exp = list(weight.shape)
exp[offset[0]] = offset[1] + offset[2]
new = torch.empty(exp, device=weight.device, dtype=weight.dtype)
new[:old_weight.shape[0]] = old_weight
old_weight = new
w = old_weight.narrow(offset[0], offset[1], offset[2])
else: