sdxl_permutation_spec

Créé Le diff n'expire jamais
16 suppressions
250 lignes
34 ajouts
269 lignes
from modules.merging.merge_rebasin import PermutationSpec, permutation_spec_from_axes_to_perm
# Ref: https://github.com/vladmandic/automatic/blob/dev/modules/merging/merge_PermSpec_SDXL.py
# Tldr: Awful. So many layers are incorrect and missing!

from PermSpec_Base import PermutationSpec, permutation_spec_from_axes_to_perm


def sdxl_permutation_spec() -> PermutationSpec:
def sdxl_permutation_spec() -> PermutationSpec:
conv = lambda name, p_in, p_out: { # pylint: disable=unnecessary-lambda-assignment
conv = lambda name, p_in, p_out: { # pylint: disable=unnecessary-lambda-assignment
f"{name}.weight": (
f"{name}.weight": (
p_out,
p_out,
p_in,
p_in,
),
),
f"{name}.bias": (p_out,),
f"{name}.bias": (p_out,),
}
}
norm = lambda name, p: {f"{name}.weight": (p,), f"{name}.bias": (p,)} # pylint: disable=unnecessary-lambda-assignment
norm = lambda name, p: {f"{name}.weight": (p,), f"{name}.bias": (p,)} # pylint: disable=unnecessary-lambda-assignment
dense = (
dense = (
lambda name, p_in, p_out, bias=True: { # pylint: disable=unnecessary-lambda-assignment
lambda name, p_in, p_out, bias=True: { # pylint: disable=unnecessary-lambda-assignment
f"{name}.weight": (p_out, p_in),
f"{name}.weight": (p_out, p_in),
f"{name}.bias": (p_out,),
f"{name}.bias": (p_out,),
}
}
if bias
if bias
else {f"{name}.weight": (p_out, p_in)}
else {f"{name}.weight": (p_out, p_in)}
)
)
# Layer name is somehow changed!
dense2 = (
lambda name, p_in, p_out, bias=True: { # pylint: disable=unnecessary-lambda-assignment
f"{name}_weight": (p_out, p_in),
f"{name}_bias": (p_out,),
}
if bias
else {f"{name}_weight": (p_out, p_in)}
)
skip = lambda name, p_in, p_out: { # pylint: disable=unnecessary-lambda-assignment
skip = lambda name, p_in, p_out: { # pylint: disable=unnecessary-lambda-assignment
f"{name}": (
f"{name}": (
p_out,
p_out,
p_in,
p_in,
None,
None,
None,
None,
)
)
}
}


# Unet Res blocks
# Unet Res blocks
easyblock = lambda name, p_in, p_out: { # pylint: disable=unnecessary-lambda-assignment
easyblock = lambda name, p_in, p_out: { # pylint: disable=unnecessary-lambda-assignment
**norm(f"{name}.in_layers.0", p_in),
**norm(f"{name}.in_layers.0", p_in),
**conv(f"{name}.in_layers.2", p_in, f"P_{name}_inner"),
**conv(f"{name}.in_layers.2", p_in, f"P_{name}_inner"),
**dense(
**dense(
f"{name}.emb_layers.1", f"P_{name}_inner2", f"P_{name}_inner3", bias=True
f"{name}.emb_layers.1", f"P_{name}_inner2", f"P_{name}_inner3", bias=True
),
),
**norm(f"{name}.out_layers.0", f"P_{name}_inner4"),
**norm(f"{name}.out_layers.0", f"P_{name}_inner4"),
**conv(f"{name}.out_layers.3", f"P_{name}_inner4", p_out),
**conv(f"{name}.out_layers.3", f"P_{name}_inner4", p_out),
}
}


# Text Encoder blocks
easyblock2 = lambda name, p: {
**norm(f"{name}.norm1", p),
**conv(f"{name}.conv1", p, f"P_{name}_inner"),
**norm(f"{name}.norm2", f"P_{name}_inner"),
**conv(f"{name}.conv2", f"P_{name}_inner", p),
}

# This is for blocks that use a residual connection, but change the number of channels via a Conv.
shortcutblock = lambda name, p_in, p_out: {
**norm(f"{name}.norm1", p_in),
**conv(f"{name}.conv1", p_in, f"P_{name}_inner"),
**norm(f"{name}.norm2", f"P_{name}_inner"),
**conv(f"{name}.conv2", f"P_{name}_inner", p_out),
**conv(f"{name}.nin_shortcut", p_in, p_out),
**norm(f"{name}.nin_shortcut", p_out),
}

return permutation_spec_from_axes_to_perm(
return permutation_spec_from_axes_to_perm(
{
{
# Skipped Layers
# Skipped Layers
**skip("betas", None, None),
**skip("betas", None, None),
**skip("alphas_cumprod", None, None),
**skip("alphas_cumprod", None, None),
**skip("alphas_cumprod_prev", None, None),
**skip("alphas_cumprod_prev", None, None),
**skip("sqrt_alphas_cumprod", None, None),
**skip("sqrt_alphas_cumprod", None, None),
**skip("sqrt_one_minus_alphas_cumprod", None, None),
**skip("sqrt_one_minus_alphas_cumprod", None, None),
**skip("log_one_minus_alphas_cumprods", None, None),
**skip("log_one_minus_alphas_cumprods", None, None),
**skip("sqrt_recip_alphas_cumprod", None, None),
**skip("sqrt_recip_alphas_cumprod", None, None),
**skip("sqrt_recipm1_alphas_cumprod", None, None),
**skip("sqrt_recipm1_alphas_cumprod", None, None),
**skip("posterior_variance", None, None),
**skip("posterior_variance", None, None),
**skip("posterior_log_variance_clipped", None, None),
**skip("posterior_log_variance_clipped", None, None),
**skip("posterior_mean_coef1", None, None),
**skip("posterior_mean_coef1", None, None),
**skip("posterior_mean_coef2", None, None),
**skip("posterior_mean_coef2", None, None),
**skip("log_one_minus_alphas_cumprod", None, None),
**skip("log_one_minus_alphas_cumprod", None, None),
**skip("model_ema.decay", None, None),
**skip("model_ema.decay", None, None),
**skip("model_ema.num_updates", None, None),
**skip("model_ema.num_updates", None, None),
**skip("conditioner.embedders.0.transformer.text_model.embeddings.position_ids", None, None),
**skip("conditioner.embedders.1.model.logit_scale", None, None),
**skip("conditioner.embedders.1.model.positional_embedding", None, None),
**skip("conditioner.embedders.1.model.text_projection", None, None),
**conv("model.diffusion_model.input_blocks.0.0", "P_bg0", "P_bg1"),
**conv("model.diffusion_model.input_blocks.0.0", "P_bg0", "P_bg1"),
**easyblock("model.diffusion_model.input_blocks.1.0", "P_bg2", "P_bg3"),
**easyblock("model.diffusion_model.input_blocks.1.0", "P_bg2", "P_bg3"),
**easyblock("model.diffusion_model.input_blocks.2.0", "P_bg4", "P_bg5"),
**easyblock("model.diffusion_model.input_blocks.2.0", "P_bg4", "P_bg5"),
**conv("model.diffusion_model.input_blocks.3.0.op", "P_bg6", "P_bg7"),
**conv("model.diffusion_model.input_blocks.3.0.op", "P_bg6", "P_bg7"),
**easyblock("model.diffusion_model.input_blocks.4.0", "P_bg8", "P_bg9"),
**easyblock("model.diffusion_model.input_blocks.4.0", "P_bg8", "P_bg9"),
**conv("model.diffusion_model.input_blocks.4.0.skip_connection", "P_bg10", "P_bg11"),
**conv("model.diffusion_model.input_blocks.4.0.skip_connection", "P_bg10", "P_bg11"),
**norm("model.diffusion_model.input_blocks.4.1.norm", "P_bg12"),
**norm("model.diffusion_model.input_blocks.4.1.norm", "P_bg12"),
**conv("model.diffusion_model.input_blocks.4.1.proj_in", "P_bg12", "P_bg13"),
**conv("model.diffusion_model.input_blocks.4.1.proj_in", "P_bg12", "P_bg13"),
**conv("model.diffusion_model.input_blocks.4.1.proj_out", "P_bg14", "P_bg15"),
**conv("model.diffusion_model.input_blocks.4.1.proj_out", "P_bg14", "P_bg15"),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_k", "P_bg16", "P_bg17", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_k", "P_bg16", "P_bg17", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_out.0", "P_bg16", "P_bg17", bias=True),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_out.0", "P_bg16", "P_bg17", bias=True),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_q", "P_bg16", "P_bg17", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_q", "P_bg16", "P_bg17", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_v", "P_bg16", "P_bg17", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_v", "P_bg16", "P_bg17", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_k", "P_bg18", "P_bg19", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_k", "P_bg18", "P_bg19", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_out.0", "P_bg20", "P_bg21", bias=True),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_out.0", "P_bg20", "P_bg21", bias=True),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_q", "P_bg20", "P_bg21", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_q", "P_bg20", "P_bg21", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_v", "P_bg18", "P_bg19", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_v", "P_bg18", "P_bg19", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.ff.net.0.proj", "P_bg22", "P_bg23", bias=True),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.ff.net.0.proj", "P_bg22", "P_bg23", bias=True),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.ff.net.2", "P_bg24", "P_bg25", bias=True),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.ff.net.2", "P_bg24", "P_bg25", bias=True),
**norm("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm1", "P_bg26"),
**norm("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm1", "P_bg26"),
**norm("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm2", "P_bg26"),
**norm("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm2", "P_bg26"),
**norm("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm3", "P_bg26"),
**norm("model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm3", "P_bg26"),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn1.to_k", "P_bg27", "P_bg28", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn1.to_k", "P_bg27", "P_bg28", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn1.to_out.0", "P_bg27", "P_bg28", bias=True),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn1.to_out.0", "P_bg27", "P_bg28", bias=True),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn1.to_q", "P_bg27", "P_bg28", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn1.to_q", "P_bg27", "P_bg28", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn1.to_v", "P_bg27", "P_bg28", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn1.to_v", "P_bg27", "P_bg28", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn2.to_k", "P_bg29", "P_bg30", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn2.to_k", "P_bg29", "P_bg30", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn2.to_out.0", "P_bg31", "P_bg32", bias=True),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn2.to_out.0", "P_bg31", "P_bg32", bias=True),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn2.to_q", "P_bg31", "P_bg32", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn2.to_q", "P_bg31", "P_bg32", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn2.to_v", "P_bg33", "P_bg34", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn2.to_v", "P_bg33", "P_bg34", bias=False),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.ff.net.0.proj", "P_bg35", "P_bg36", bias=True),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.ff.net.0.proj", "P_bg35", "P_bg36", bias=True),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.ff.net.2", "P_bg37", "P_bg38", bias=True),
**dense("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.ff.net.2", "P_bg37", "P_bg38", bias=True),
**norm("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.norm1", "P_bg39"),
**norm("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.norm1", "P_bg39"),
**norm("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.norm2", "P_bg39"),
**norm("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.norm2", "P_bg39"),
**norm("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.norm3", "P_bg39"),
**norm("model.diffusion_model.input_blocks.4.1.transformer_blocks.1.norm3", "P_bg39"),
**easyblock("model.diffusion_model.input_blocks.5.0", "P_bg40", "P_bg41"),
**easyblock("model.diffusion_model.input_blocks.5.0", "P_bg40", "P_bg41"),
**norm("model.diffusion_model.input_blocks.5.1.norm", "P_bg42"),
**norm("model.diffusion_model.input_blocks.5.1.norm", "P_bg42"),
**conv("model.diffusion_model.input_blocks.5.1.proj_in", "P_bg43", "P_bg44"),
**conv("model.diffusion_model.input_blocks.5.1.proj_in", "P_bg43", "P_bg44"),
**conv("model.diffusion_model.input_blocks.5.1.proj_out", "P_bg45", "P_bg46"),
**conv("model.diffusion_model.input_blocks.5.1.proj_out", "P_bg45", "P_bg46"),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_k", "P_bg47", "P_bg48", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_k", "P_bg47", "P_bg48", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_out.0", "P_bg47", "P_bg48", bias=True),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_out.0", "P_bg47", "P_bg48", bias=True),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_q", "P_bg47", "P_bg48", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_q", "P_bg47", "P_bg48", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_v", "P_bg47", "P_bg48", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_v", "P_bg47", "P_bg48", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn2.to_k", "P_bg49", "P_bg50", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn2.to_k", "P_bg49", "P_bg50", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn2.to_out.0", "P_bg51", "P_bg52", bias=True),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn2.to_out.0", "P_bg51", "P_bg52", bias=True),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn2.to_q", "P_bg53", "P_bg54", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn2.to_q", "P_bg53", "P_bg54", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn2.to_v", "P_bg55", "P_bg56", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn2.to_v", "P_bg55", "P_bg56", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.ff.net.0.proj", "P_bg57", "P_bg58", bias=True),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.ff.net.0.proj", "P_bg57", "P_bg58", bias=True),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.ff.net.2", "P_bg59", "P_bg60", bias=True),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.ff.net.2", "P_bg59", "P_bg60", bias=True),
**norm("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm1", "P_bg61"),
**norm("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm1", "P_bg61"),
**norm("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm2", "P_bg61"),
**norm("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm2", "P_bg61"),
**norm("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm3", "P_bg61"),
**norm("model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm3", "P_bg61"),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn1.to_k", "P_bg62", "P_bg63", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn1.to_k", "P_bg62", "P_bg63", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn1.to_out.0", "P_bg62", "P_bg63", bias=True),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn1.to_out.0", "P_bg62", "P_bg63", bias=True),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn1.to_q", "P_bg62", "P_bg63", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn1.to_q", "P_bg62", "P_bg63", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn1.to_v", "P_bg62", "P_bg63", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn1.to_v", "P_bg62", "P_bg63", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn2.to_k", "P_bg64", "P_bg65", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn2.to_k", "P_bg64", "P_bg65", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn2.to_out.0", "P_bg66", "P_bg67", bias=True),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn2.to_out.0", "P_bg66", "P_bg67", bias=True),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn2.to_q", "P_bg68", "P_bg69", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn2.to_q", "P_bg68", "P_bg69", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn2.to_v", "P_bg70", "P_bg71", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn2.to_v", "P_bg70", "P_bg71", bias=False),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.ff.net.0.proj", "P_bg72", "P_bg73", bias=True),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.ff.net.0.proj", "P_bg72", "P_bg73", bias=True),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.ff.net.2", "P_bg74", "P_bg75", bias=True),
**dense("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.ff.net.2", "P_bg74", "P_bg75", bias=True),
**norm("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.norm1", "P_bg76"),
**norm("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.norm1", "P_bg76"),
**norm("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.norm2", "P_bg76"),
**norm("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.norm2", "P_bg76"),
**norm("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.norm3", "P_bg76"),
**norm("model.diffusion_model.input_blocks.5.1.transformer_blocks.1.norm3", "P_bg76"),
**conv("model.diffusion_model.input_blocks.6.0.op", "P_bg77", "P_bg78"),
**conv("model.diffusion_model.input_blocks.6.0.op", "P_bg77", "P_bg78"),
**easyblock("model.diffusion_model.input_blocks.7.0", "P_bg78", "P_bg79"),
**easyblock("model.diffusion_model.input_blocks.7.0", "P_bg78", "P_bg79"),
**conv("model.diffusion_model.input_blocks.7.0.skip_connection", "P_bg80", "P_bg81"),
**conv("model.diffusion_model.input_blocks.7.0.skip_connection", "P_bg80", "P_bg81"),
**norm("model.diffusion_model.input_blocks.7.1.norm", "P_bg82"),
**norm("model.diffusion_model.input_blocks.7.1.norm", "P_bg82"),
**conv("model.diffusion_model.input_blocks.7.1.proj_in", "P_bg82", "P_bg83"),
**conv("model.diffusion_model.input_blocks.7.1.proj_in", "P_bg82", "P_bg83"),
**conv("model.diffusion_model.input_blocks.7.1.proj_out", "P_bg84", "P_bg85"),
**conv("model.diffusion_model.input_blocks.7.1.proj_out", "P_bg84", "P_bg85"),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_k", "P_bg86", "P_bg87", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_k", "P_bg86", "P_bg87", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_out.0", "P_bg86", "P_bg87", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_out.0", "P_bg86", "P_bg87", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_q", "P_bg86", "P_bg87", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_q", "P_bg86", "P_bg87", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_v", "P_bg86", "P_bg87", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_v", "P_bg86", "P_bg87", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn2.to_k", "P_bg88", "P_bg89", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn2.to_k", "P_bg88", "P_bg89", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn2.to_out.0", "P_bg90", "P_bg91", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn2.to_out.0", "P_bg90", "P_bg91", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn2.to_q", "P_bg92", "P_bg93", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn2.to_q", "P_bg92", "P_bg93", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn2.to_v", "P_bg94", "P_bg95", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn2.to_v", "P_bg94", "P_bg95", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.ff.net.0.proj", "P_bg96", "P_bg97", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.ff.net.0.proj", "P_bg96", "P_bg97", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.ff.net.2", "P_bg98", "P_bg99", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.ff.net.2", "P_bg98", "P_bg99", bias=True),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm1", "P_bg100"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm1", "P_bg100"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm2", "P_bg100"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm2", "P_bg100"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm3", "P_bg100"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm3", "P_bg100"),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.attn1.to_k", "P_bg101", "P_bg102", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.attn1.to_k", "P_bg101", "P_bg102", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.attn1.to_out.0", "P_bg101", "P_bg102", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.attn1.to_out.0", "P_bg101", "P_bg102", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.attn1.to_q", "P_bg101", "P_bg102", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.attn1.to_q", "P_bg101", "P_bg102", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.attn1.to_v", "P_bg101", "P_bg102", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.attn1.to_v", "P_bg101", "P_bg102", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.attn2.to_k", "P_bg103", "P_bg104", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.attn2.to_k", "P_bg103", "P_bg104", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.attn2.to_out.0", "P_bg105", "P_bg106", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.attn2.to_out.0", "P_bg105", "P_bg106", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.attn2.to_q", "P_bg107", "P_bg108", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.attn2.to_q", "P_bg107", "P_bg108", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.attn2.to_v", "P_bg109", "P_bg110", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.attn2.to_v", "P_bg109", "P_bg110", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.ff.net.0.proj", "P_bg111", "P_bg112", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.ff.net.0.proj", "P_bg111", "P_bg112", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.ff.net.2", "P_bg113", "P_bg114", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.ff.net.2", "P_bg113", "P_bg114", bias=True),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.norm1", "P_bg115"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.norm1", "P_bg115"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.norm2", "P_bg115"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.norm2", "P_bg115"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.norm3", "P_bg115"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.1.norm3", "P_bg115"),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.attn1.to_k", "P_bg116", "P_bg117", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.attn1.to_k", "P_bg116", "P_bg117", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.attn1.to_out.0", "P_bg116", "P_bg117", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.attn1.to_out.0", "P_bg116", "P_bg117", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.attn1.to_q", "P_bg116", "P_bg117", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.attn1.to_q", "P_bg116", "P_bg117", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.attn1.to_v", "P_bg116", "P_bg117", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.attn1.to_v", "P_bg116", "P_bg117", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.attn2.to_k", "P_bg118", "P_bg119", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.attn2.to_k", "P_bg118", "P_bg119", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.attn2.to_out.0", "P_bg120", "P_bg121", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.attn2.to_out.0", "P_bg120", "P_bg121", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.attn2.to_q", "P_bg122", "P_bg123", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.attn2.to_q", "P_bg122", "P_bg123", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.attn2.to_v", "P_bg124", "P_bg125", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.attn2.to_v", "P_bg124", "P_bg125", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.ff.net.0.proj", "P_bg126", "P_bg127", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.ff.net.0.proj", "P_bg126", "P_bg127", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.ff.net.2", "P_bg128", "P_bg129", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.ff.net.2", "P_bg128", "P_bg129", bias=True),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.norm1", "P_bg130"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.norm1", "P_bg130"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.norm2", "P_bg130"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.norm2", "P_bg130"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.norm3", "P_bg130"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.2.norm3", "P_bg130"),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn1.to_k", "P_bg110", "P_bg111", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn1.to_k", "P_bg110", "P_bg111", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn1.to_out.0", "P_bg131", "P_bg132", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn1.to_out.0", "P_bg131", "P_bg132", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn1.to_q", "P_bg131", "P_bg132", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn1.to_q", "P_bg131", "P_bg132", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn1.to_v", "P_bg131", "P_bg132", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn1.to_v", "P_bg131", "P_bg132", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn2.to_k", "P_bg133", "P_bg134", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn2.to_k", "P_bg133", "P_bg134", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn2.to_out.0", "P_bg135", "P_bg136", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn2.to_out.0", "P_bg135", "P_bg136", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn2.to_q", "P_bg137", "P_bg138", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn2.to_q", "P_bg137", "P_bg138", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn2.to_v", "P_bg139", "P_bg140", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn2.to_v", "P_bg139", "P_bg140", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.ff.net.0.proj", "P_bg141", "P_bg142", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.ff.net.0.proj", "P_bg141", "P_bg142", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.ff.net.2", "P_bg143", "P_bg144", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.ff.net.2", "P_bg143", "P_bg144", bias=True),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.norm1", "P_bg145"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.norm1", "P_bg145"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.norm2", "P_bg145"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.norm2", "P_bg145"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.norm3", "P_bg145"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.3.norm3", "P_bg145"),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn1.to_k", "P_bg146", "P_bg147", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn1.to_k", "P_bg146", "P_bg147", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn1.to_out.0", "P_bg146", "P_bg147", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn1.to_out.0", "P_bg146", "P_bg147", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn1.to_q", "P_bg146", "P_bg147", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn1.to_q", "P_bg146", "P_bg147", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn1.to_v", "P_bg146", "P_bg147", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn1.to_v", "P_bg146", "P_bg147", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn2.to_k", "P_bg148", "P_bg149", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn2.to_k", "P_bg148", "P_bg149", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn2.to_out.0", "P_bg150", "P_bg151", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn2.to_out.0", "P_bg150", "P_bg151", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn2.to_q", "P_bg152", "P_bg153", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn2.to_q", "P_bg152", "P_bg153", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn2.to_v", "P_bg154", "P_bg155", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn2.to_v", "P_bg154", "P_bg155", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.ff.net.0.proj", "P_bg156", "P_bg157", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.ff.net.0.proj", "P_bg156", "P_bg157", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.ff.net.2", "P_bg158", "P_bg159", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.ff.net.2", "P_bg158", "P_bg159", bias=True),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.norm1", "P_bg160"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.norm1", "P_bg160"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.norm2", "P_bg160"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.norm2", "P_bg160"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.norm3", "P_bg160"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.4.norm3", "P_bg160"),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn1.to_k", "P_bg161", "P_bg162", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn1.to_k", "P_bg161", "P_bg162", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn1.to_out.0", "P_bg161", "P_bg162", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn1.to_out.0", "P_bg161", "P_bg162", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn1.to_q", "P_bg161", "P_bg162", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn1.to_q", "P_bg161", "P_bg162", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn1.to_v", "P_bg161", "P_bg162", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn1.to_v", "P_bg161", "P_bg162", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn2.to_k", "P_bg163", "P_bg164", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn2.to_k", "P_bg163", "P_bg164", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn2.to_out.0", "P_bg165", "P_bg166", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn2.to_out.0", "P_bg165", "P_bg166", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn2.to_q", "P_bg167", "P_bg168", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn2.to_q", "P_bg167", "P_bg168", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn2.to_v", "P_bg169", "P_bg170", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn2.to_v", "P_bg169", "P_bg170", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.ff.net.0.proj", "P_bg171", "P_bg172", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.ff.net.0.proj", "P_bg171", "P_bg172", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.ff.net.2", "P_bg173", "P_bg174", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.ff.net.2", "P_bg173", "P_bg174", bias=True),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.norm1", "P_bg175"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.norm1", "P_bg175"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.norm2", "P_bg175"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.norm2", "P_bg175"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.norm3", "P_bg175"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.5.norm3", "P_bg175"),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn1.to_k", "P_bg176", "P_bg177", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn1.to_k", "P_bg176", "P_bg177", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn1.to_out.0", "P_bg176", "P_bg177", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn1.to_out.0", "P_bg176", "P_bg177", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn1.to_q", "P_bg176", "P_bg177", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn1.to_q", "P_bg176", "P_bg177", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn1.to_v", "P_bg176", "P_bg177", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn1.to_v", "P_bg176", "P_bg177", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn2.to_k", "P_bg178", "P_bg179", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn2.to_k", "P_bg178", "P_bg179", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn2.to_out.0", "P_bg180", "P_bg181", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn2.to_out.0", "P_bg180", "P_bg181", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn2.to_q", "P_bg182", "P_bg183", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn2.to_q", "P_bg182", "P_bg183", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn2.to_v", "P_bg184", "P_bg185", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn2.to_v", "P_bg184", "P_bg185", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.ff.net.0.proj", "P_bg186", "P_bg187", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.ff.net.0.proj", "P_bg186", "P_bg187", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.ff.net.2", "P_bg188", "P_bg189", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.ff.net.2", "P_bg188", "P_bg189", bias=True),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.norm1", "P_bg190"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.norm1", "P_bg190"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.norm2", "P_bg190"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.norm2", "P_bg190"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.norm3", "P_bg190"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.6.norm3", "P_bg190"),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.attn1.to_k", "P_bg191", "P_bg192", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.attn1.to_k", "P_bg191", "P_bg192", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.attn1.to_out.0", "P_bg193", "P_bg194", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.attn1.to_out.0", "P_bg193", "P_bg194", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.attn1.to_q", "P_bg195", "P_bg196", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.attn1.to_q", "P_bg195", "P_bg196", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.attn1.to_v", "P_bg197", "P_bg198", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.attn1.to_v", "P_bg197", "P_bg198", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.attn2.to_k", "P_bg199", "P_bg200", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.attn2.to_k", "P_bg199", "P_bg200", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.attn2.to_out.0", "P_bg201", "P_bg202", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.attn2.to_out.0", "P_bg201", "P_bg202", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.attn2.to_q", "P_bg203", "P_bg204", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.attn2.to_q", "P_bg203", "P_bg204", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.attn2.to_v", "P_bg205", "P_bg206", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.attn2.to_v", "P_bg205", "P_bg206", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.ff.net.0.proj", "P_bg207", "P_bg208", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.ff.net.0.proj", "P_bg207", "P_bg208", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.ff.net.2", "P_bg209", "P_bg210", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.ff.net.2", "P_bg209", "P_bg210", bias=True),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.norm1", "P_bg211"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.norm1", "P_bg211"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.norm2", "P_bg211"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.norm2", "P_bg211"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.norm3", "P_bg211"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.7.norm3", "P_bg211"),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.attn1.to_k", "P_bg212", "P_bg213", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.attn1.to_k", "P_bg212", "P_bg213", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.attn1.to_out.0", "P_bg212", "P_bg213", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.attn1.to_out.0", "P_bg212", "P_bg213", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.attn1.to_q", "P_bg212", "P_bg213", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.attn1.to_q", "P_bg212", "P_bg213", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.attn1.to_v", "P_bg212", "P_bg213", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.attn1.to_v", "P_bg212", "P_bg213", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.attn2.to_k", "P_bg214", "P_bg215", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.attn2.to_k", "P_bg214", "P_bg215", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.attn2.to_out.0", "P_bg216", "P_bg217", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.attn2.to_out.0", "P_bg216", "P_bg217", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.attn2.to_q", "P_bg218", "P_bg219", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.attn2.to_v", "P_bg220", "P_bg221", bias=False),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.ff.net.0.proj", "P_bg222", "P_bg223", bias=True),
**dense("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.ff.net.2", "P_bg224", "P_bg225", bias=True),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.norm1", "P_bg226"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.norm2", "P_bg226"),
**norm("model.diffusion_model.input_blocks.7.1.transformer_blocks.8.norm3", "P_bg226"),
**dense("m