hcpdiff 0.9.1__py3-none-any.whl → 2.1__py3-none-any.whl

hcpdiff-0.9.1.data/data/hcpdiff/cfgs/infer/img2img_controlnet.yaml

@@ -1,19 +0,0 @@
-_base_: [cfgs/infer/text2img.yaml]
-
-ex_input:
-  cond:
-    _target_: hcpdiff.data.data_processor.ControlNetProcessor
-    image: 'cond_img.png'
-
-merge:
-  plugin_cfg: cfgs/plugins/plugin_controlnet.yaml
-
-  group1:
-    type: 'unet'
-    base_model_alpha: 1.0 # base model weight to merge with lora or part
-    lora: null
-    part: null
-    plugin:
-      controlnet1:
-        path: 'ckpts/controlnet.ckpt'
-        layers: 'all'

hcpdiff-0.9.1.data/data/hcpdiff/cfgs/infer/inpaint.yaml

@@ -1,11 +0,0 @@
-_base_: [cfgs/infer/text2img.yaml]
-
-condition:
-  type: inpaint
-  image: 'cond_img.png'
-  mask: 'mask.png'
-
-infer_args:
-  guidance_scale: 7.5
-  num_inference_steps: 50
-  strength: 0.75

hcpdiff-0.9.1.data/data/hcpdiff/cfgs/infer/load_lora.yaml

@@ -1,26 +0,0 @@
-_base_: [cfgs/infer/text2img.yaml]
-
-merge:
-  group1:
-    type: 'unet'
-    base_model_alpha: 1.0 # base model weight to merge with lora or part
-    lora:
-      - path: 'lora1-unet.safetensors'
-        alpha: 0.8
-        layers: 'all'
-      - path: 'lora2-unet.safetensors'
-        alpha: 0.65
-        layers: 'all'
-    part: null
-
-  group2:
-    type: 'TE'
-    base_model_alpha: 1.0 # base model weight to infer with lora or part
-    lora:
-      - path: 'lora1-te.safetensors'
-        alpha: 0.8
-        layers: 'all'
-      - path: 'lora2-te.safetensors'
-        alpha: 0.65
-        layers: 'all'
-    part: null

hcpdiff-0.9.1.data/data/hcpdiff/cfgs/infer/load_unet_part.yaml

@@ -1,18 +0,0 @@
-_base_: [cfgs/infer/text2img.yaml]
-
-merge:
-  group1:
-    type: 'unet'
-    base_model_alpha: 0.0 # 基础模型权重0.0，part部分新模型权重1.0，表示用part模型覆盖原有模型
-    part:
-      - path: 'unet-100.safetensors的路径'
-        alpha: 1.0
-        layers: 'all'
-
-  group2: # 如果没有练text_encoder，这部分可以不加
-    type: 'TE'
-    base_model_alpha: 0.0
-    lora:
-      - path: 'text_encoder-100.safetensors的路径'
-        alpha: 1.0
-        layers: 'all'

hcpdiff-0.9.1.data/data/hcpdiff/cfgs/infer/offload_2GB.yaml

@@ -1,6 +0,0 @@
-_base_: [cfgs/infer/text2img.yaml]
-
-offload:
-  max_VRAM: 2GiB
-  max_RAM: 30GiB
-  vae_cpu: False

hcpdiff-0.9.1.data/data/hcpdiff/cfgs/infer/save_model.yaml

@@ -1,44 +0,0 @@
-_base_: [cfgs/infer/text2img.yaml]
-
-save_model:
-  path: ckpts/model
-  to_safetensors: true
-
-merge:
-
-  group_lora_unet:
-    type: 'unet'
-    base_model_alpha: 1.0 # base model weight to merge with lora or part
-    lora:
-      - path: 'exps/lora1/ckpts/unet-600.safetensors'
-        alpha: 0.8
-        layers: 'all'
-      - path: 'exps/lora2/ckpts/unet-800.safetensors'
-        alpha: 0.7
-        layers:
-          - 're:.*\.to_k$'
-          - 're:.*\.to_v$'
-
-  group_lora_TE:
-    type: 'TE'
-    base_model_alpha: 1.0 # base model weight to infer with lora or part
-    lora:
-      - path: 'exps/lora1/ckpts/text_encoder-600.safetensors'
-        alpha: 0.8
-        layers: 'all'
-
-  group_part_unet:
-    type: 'unet'
-    base_model_alpha: 0.4 # base model weight to merge with lora or part
-    part:
-      - path: 'exps/part1/ckpts/unet-500.safetensors'
-        alpha: 0.6
-        layers: 'all'
-
-  group_part_TE:
-    type: 'TE'
-    base_model_alpha: 0.4 # base model weight to merge with lora or part
-    part:
-      - path: 'exps/part1/ckpts/text_encoder-500.safetensors'
-        alpha: 0.6
-        layers: 'all'

hcpdiff-0.9.1.data/data/hcpdiff/cfgs/infer/text2img.yaml

@@ -1,53 +0,0 @@
-# base_state*base_model_alpha + (lora_state[i]*lora_scale[i]*lora_alpha[i]) + (part_state[k]*part_alpha[k])
-
-pretrained_model: ''
-prompt: ''
-neg_prompt: 'lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry'
-emb_dir: 'embs/'
-N_repeats: 1
-clip_skip: 0
-clip_final_norm: True
-encoder_attention_mask: True
-bs: 4
-num: 1
-seed: null
-dtype: 'fp16'
-amp: True
-
-condition: null
-
-ex_input: {}
-
-# Syntactic sugar for interface
-save:
-  out_dir: 'output/'
-  save_cfg: True
-  image_type: png
-  quality: 95
-#  image_type: webp
-#  quality: 75
-
-offload: null
-
-#vae_optimize: null
-vae_optimize:
-  tiling: False
-  slicing: False
-
-interface:
-  - _target_: hcpdiff.vis.DiskInterface
-    show_steps: 0
-    save_root: ${save.out_dir}
-    save_cfg: ${save.save_cfg}
-    image_type: ${save.image_type}
-    quality: ${save.quality}
-
-infer_args:
-  width: 512
-  height: 512
-  guidance_scale: 7.5
-  num_inference_steps: 50
-
-new_components: {}
-
-merge: null

hcpdiff-0.9.1.data/data/hcpdiff/cfgs/infer/text2img_DA++.yaml

@@ -1,34 +0,0 @@
-_base_: [cfgs/infer/text2img.yaml]
-
-merge:
-  exp_dir: '2023-04-03-10-10-36'
-  alpha: 0.8
-  alpha_neg: 0.65
-
-  group1:
-    type: 'unet'
-    base_model_alpha: 1.0 # base model weight to merge with lora or part
-    lora:
-      - path: 'exps/${merge.exp_dir}/ckpts/unet-600.safetensors'
-        alpha: ${merge.alpha}
-        layers: 'all'
-        mask: [ 0.5, 1 ] #
-      - path: 'exps/${merge.exp_dir}/ckpts/unet-neg-600.safetensors'
-        alpha: ${merge.alpha_neg}
-        layers: 'all'
-        mask: [ 0, 0.5 ]
-    part: null
-
-  group2:
-    type: 'TE'
-    base_model_alpha: 1.0 # base model weight to infer with lora or part
-    lora:
-      - path: 'exps/${merge.exp_dir}/ckpts/text_encoder-600.safetensors'
-        alpha: ${merge.alpha}
-        layers: 'all'
-        mask: [ 0.5, 1 ]
-      - path: 'exps/${merge.exp_dir}/ckpts/text_encoder-neg-600.safetensors'
-        alpha: ${merge.alpha_neg}
-        layers: 'all'
-        mask: [ 0, 0.5 ]
-    part: null

hcpdiff-0.9.1.data/data/hcpdiff/cfgs/infer/text2img_sdxl.yaml

@@ -1,9 +0,0 @@
-_base_: [cfgs/infer/text2img.yaml]
-
-emb_dir: embs_sdxl
-
-prompt: ''
-neg_prompt: ''
-
-clip_skip: 1
-clip_final_norm: False

hcpdiff-0.9.1.data/data/hcpdiff/cfgs/plugins/plugin_controlnet.yaml

@@ -1,17 +0,0 @@
-_base_: [cfgs/train/tuning_base.yaml]
-
-plugin_unet:
-  controlnet1:
-    _target_: hcpdiff.models.controlnet.ControlNetPlugin
-    _partial_: True
-    lr: 1e-4
-    from_layers:
-      - 'pre_hook:'
-      - 'pre_hook:conv_in' # to make forward inside autocast
-    to_layers:
-      - 'down_blocks.0'
-      - 'down_blocks.1'
-      - 'down_blocks.2'
-      - 'down_blocks.3'
-      - 'mid_block'
-      - 'pre_hook:up_blocks.3.resnets.2'

hcpdiff-0.9.1.data/data/hcpdiff/cfgs/te_struct.txt

@@ -1,193 +0,0 @@
-CLIPTextModel(
-  (text_model): CLIPTextTransformer(
-    (embeddings): CLIPTextEmbeddings(
-      (token_embedding): Embedding(49408, 768)
-      (position_embedding): Embedding(77, 768)
-    )
-    (encoder): CLIPEncoder(
-      (layers): ModuleList(
-        (0): CLIPEncoderLayer(
-          (self_attn): CLIPAttention(
-            (k_proj): Linear(in_features=768, out_features=768, bias=True)
-            (v_proj): Linear(in_features=768, out_features=768, bias=True)
-            (q_proj): Linear(in_features=768, out_features=768, bias=True)
-            (out_proj): Linear(in_features=768, out_features=768, bias=True)
-          )
-          (layer_norm1): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-          (mlp): CLIPMLP(
-            (activation_fn): QuickGELUActivation()
-            (fc1): Linear(in_features=768, out_features=3072, bias=True)
-            (fc2): Linear(in_features=3072, out_features=768, bias=True)
-          )
-          (layer_norm2): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-        )
-        (1): CLIPEncoderLayer(
-          (self_attn): CLIPAttention(
-            (k_proj): Linear(in_features=768, out_features=768, bias=True)
-            (v_proj): Linear(in_features=768, out_features=768, bias=True)
-            (q_proj): Linear(in_features=768, out_features=768, bias=True)
-            (out_proj): Linear(in_features=768, out_features=768, bias=True)
-          )
-          (layer_norm1): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-          (mlp): CLIPMLP(
-            (activation_fn): QuickGELUActivation()
-            (fc1): Linear(in_features=768, out_features=3072, bias=True)
-            (fc2): Linear(in_features=3072, out_features=768, bias=True)
-          )
-          (layer_norm2): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-        )
-        (2): CLIPEncoderLayer(
-          (self_attn): CLIPAttention(
-            (k_proj): Linear(in_features=768, out_features=768, bias=True)
-            (v_proj): Linear(in_features=768, out_features=768, bias=True)
-            (q_proj): Linear(in_features=768, out_features=768, bias=True)
-            (out_proj): Linear(in_features=768, out_features=768, bias=True)
-          )
-          (layer_norm1): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-          (mlp): CLIPMLP(
-            (activation_fn): QuickGELUActivation()
-            (fc1): Linear(in_features=768, out_features=3072, bias=True)
-            (fc2): Linear(in_features=3072, out_features=768, bias=True)
-          )
-          (layer_norm2): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-        )
-        (3): CLIPEncoderLayer(
-          (self_attn): CLIPAttention(
-            (k_proj): Linear(in_features=768, out_features=768, bias=True)
-            (v_proj): Linear(in_features=768, out_features=768, bias=True)
-            (q_proj): Linear(in_features=768, out_features=768, bias=True)
-            (out_proj): Linear(in_features=768, out_features=768, bias=True)
-          )
-          (layer_norm1): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-          (mlp): CLIPMLP(
-            (activation_fn): QuickGELUActivation()
-            (fc1): Linear(in_features=768, out_features=3072, bias=True)
-            (fc2): Linear(in_features=3072, out_features=768, bias=True)
-          )
-          (layer_norm2): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-        )
-        (4): CLIPEncoderLayer(
-          (self_attn): CLIPAttention(
-            (k_proj): Linear(in_features=768, out_features=768, bias=True)
-            (v_proj): Linear(in_features=768, out_features=768, bias=True)
-            (q_proj): Linear(in_features=768, out_features=768, bias=True)
-            (out_proj): Linear(in_features=768, out_features=768, bias=True)
-          )
-          (layer_norm1): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-          (mlp): CLIPMLP(
-            (activation_fn): QuickGELUActivation()
-            (fc1): Linear(in_features=768, out_features=3072, bias=True)
-            (fc2): Linear(in_features=3072, out_features=768, bias=True)
-          )
-          (layer_norm2): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-        )
-        (5): CLIPEncoderLayer(
-          (self_attn): CLIPAttention(
-            (k_proj): Linear(in_features=768, out_features=768, bias=True)
-            (v_proj): Linear(in_features=768, out_features=768, bias=True)
-            (q_proj): Linear(in_features=768, out_features=768, bias=True)
-            (out_proj): Linear(in_features=768, out_features=768, bias=True)
-          )
-          (layer_norm1): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-          (mlp): CLIPMLP(
-            (activation_fn): QuickGELUActivation()
-            (fc1): Linear(in_features=768, out_features=3072, bias=True)
-            (fc2): Linear(in_features=3072, out_features=768, bias=True)
-          )
-          (layer_norm2): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-        )
-        (6): CLIPEncoderLayer(
-          (self_attn): CLIPAttention(
-            (k_proj): Linear(in_features=768, out_features=768, bias=True)
-            (v_proj): Linear(in_features=768, out_features=768, bias=True)
-            (q_proj): Linear(in_features=768, out_features=768, bias=True)
-            (out_proj): Linear(in_features=768, out_features=768, bias=True)
-          )
-          (layer_norm1): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-          (mlp): CLIPMLP(
-            (activation_fn): QuickGELUActivation()
-            (fc1): Linear(in_features=768, out_features=3072, bias=True)
-            (fc2): Linear(in_features=3072, out_features=768, bias=True)
-          )
-          (layer_norm2): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-        )
-        (7): CLIPEncoderLayer(
-          (self_attn): CLIPAttention(
-            (k_proj): Linear(in_features=768, out_features=768, bias=True)
-            (v_proj): Linear(in_features=768, out_features=768, bias=True)
-            (q_proj): Linear(in_features=768, out_features=768, bias=True)
-            (out_proj): Linear(in_features=768, out_features=768, bias=True)
-          )
-          (layer_norm1): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-          (mlp): CLIPMLP(
-            (activation_fn): QuickGELUActivation()
-            (fc1): Linear(in_features=768, out_features=3072, bias=True)
-            (fc2): Linear(in_features=3072, out_features=768, bias=True)
-          )
-          (layer_norm2): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-        )
-        (8): CLIPEncoderLayer(
-          (self_attn): CLIPAttention(
-            (k_proj): Linear(in_features=768, out_features=768, bias=True)
-            (v_proj): Linear(in_features=768, out_features=768, bias=True)
-            (q_proj): Linear(in_features=768, out_features=768, bias=True)
-            (out_proj): Linear(in_features=768, out_features=768, bias=True)
-          )
-          (layer_norm1): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-          (mlp): CLIPMLP(
-            (activation_fn): QuickGELUActivation()
-            (fc1): Linear(in_features=768, out_features=3072, bias=True)
-            (fc2): Linear(in_features=3072, out_features=768, bias=True)
-          )
-          (layer_norm2): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-        )
-        (9): CLIPEncoderLayer(
-          (self_attn): CLIPAttention(
-            (k_proj): Linear(in_features=768, out_features=768, bias=True)
-            (v_proj): Linear(in_features=768, out_features=768, bias=True)
-            (q_proj): Linear(in_features=768, out_features=768, bias=True)
-            (out_proj): Linear(in_features=768, out_features=768, bias=True)
-          )
-          (layer_norm1): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-          (mlp): CLIPMLP(
-            (activation_fn): QuickGELUActivation()
-            (fc1): Linear(in_features=768, out_features=3072, bias=True)
-            (fc2): Linear(in_features=3072, out_features=768, bias=True)
-          )
-          (layer_norm2): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-        )
-        (10): CLIPEncoderLayer(
-          (self_attn): CLIPAttention(
-            (k_proj): Linear(in_features=768, out_features=768, bias=True)
-            (v_proj): Linear(in_features=768, out_features=768, bias=True)
-            (q_proj): Linear(in_features=768, out_features=768, bias=True)
-            (out_proj): Linear(in_features=768, out_features=768, bias=True)
-          )
-          (layer_norm1): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-          (mlp): CLIPMLP(
-            (activation_fn): QuickGELUActivation()
-            (fc1): Linear(in_features=768, out_features=3072, bias=True)
-            (fc2): Linear(in_features=3072, out_features=768, bias=True)
-          )
-          (layer_norm2): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-        )
-        (11): CLIPEncoderLayer(
-          (self_attn): CLIPAttention(
-            (k_proj): Linear(in_features=768, out_features=768, bias=True)
-            (v_proj): Linear(in_features=768, out_features=768, bias=True)
-            (q_proj): Linear(in_features=768, out_features=768, bias=True)
-            (out_proj): Linear(in_features=768, out_features=768, bias=True)
-          )
-          (layer_norm1): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-          (mlp): CLIPMLP(
-            (activation_fn): QuickGELUActivation()
-            (fc1): Linear(in_features=768, out_features=3072, bias=True)
-            (fc2): Linear(in_features=3072, out_features=768, bias=True)
-          )
-          (layer_norm2): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-        )
-      )
-    )
-    (final_layer_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
-  )
-)

hcpdiff-0.9.1.data/data/hcpdiff/cfgs/train/dataset/base_dataset.yaml

@@ -1,29 +0,0 @@
-data:
-  dataset1:
-    _target_: hcpdiff.data.TextImagePairDataset
-    _partial_: True # Not directly instantiate the object here. There are other parameters to be added in the runtime.
-    batch_size: 4
-    cache_latents: True
-    att_mask_encode: False
-    loss_weight: 1.0
-
-    source:
-      data_source1:
-        _target_: hcpdiff.data.source.Text2ImageAttMapSource
-        img_root: 'imgs/'
-        prompt_template: 'prompt_tuning_template/object.txt'
-        caption_file: null # path to image captions (file_words)
-        att_mask: null
-        bg_color: [ 255, 255, 255 ] # RGB; for ARGB -> RGB
-
-        word_names: {}
-
-        text_transforms:
-          _target_: torchvision.transforms.Compose
-          transforms:
-            - _target_: hcpdiff.utils.caption_tools.TemplateFill
-              word_names: ${....word_names}
-    bucket:
-      _target_: hcpdiff.data.bucket.RatioBucket.from_files # aspect ratio bucket
-      target_area: ${hcp.eval:"512*512"}
-      num_bucket: 5

hcpdiff-0.9.1.data/data/hcpdiff/cfgs/train/dataset/regularization_dataset.yaml

@@ -1,31 +0,0 @@
-_base_: [cfgs/train/dataset/base_dataset.yaml]
-
-data:
-  dataset_class:
-    _target_: hcpdiff.data.TextImagePairDataset
-    _partial_: True
-    batch_size: 1
-    cache_latents: True
-    att_mask_encode: False
-    loss_weight: 1.0
-
-    source:
-      data_source1:
-        _target_: hcpdiff.data.source.Text2ImageAttMapSource
-        img_root: 'imgs/db_class'
-        prompt_template: 'prompt_tuning_template/object.txt'
-        caption_file: null
-        att_mask: null
-        bg_color: [ 255, 255, 255 ] # RGB; for ARGB -> RGB
-
-        word_names:
-          pt1: ''
-
-        text_transforms:
-          _target_: torchvision.transforms.Compose
-          transforms:
-            - _target_: hcpdiff.utils.caption_tools.TemplateFill
-              word_names: ${....word_names}
-    bucket:
-      _target_: hcpdiff.data.bucket.FixedBucket
-      target_size: [ 512, 512 ]

hcpdiff-0.9.1.data/data/hcpdiff/cfgs/train/examples/CustomDiffusion.yaml

@@ -1,74 +0,0 @@
-_base_:
-  - cfgs/train/dataset/regularization_dataset.yaml
-  - cfgs/train/train_base.yaml
-  - cfgs/train/tuning_base.yaml
-
-unet:
-  -
-    lr: 1e-6
-    layers: # k,v of cross attention
-      - 're:.*attn2\.to_k$'
-      - 're:.*attn2\.to_v$'
-
-## lora version of CustomDiffusion
-#lora_unet:
-#  -
-#    lr: 1e-4
-#    layers:
-#      - 're:.*attn2\.to_k$'
-#      - 're:.*attn2\.to_v$'
-
-tokenizer_pt:
-  train: # prompt tuning embeddings, needs to be created in advance
-    - { name: 'pt-dog1', lr: 0.003 }
-
-train:
-  gradient_accumulation_steps: 1
-  save_step: 100
-
-  scheduler:
-    name: 'constant_with_warmup'
-    num_warmup_steps: 50
-    num_training_steps: 600
-
-model:
-  pretrained_model_name_or_path: 'runwayml/stable-diffusion-v1-5'
-  tokenizer_repeats: 1
-  ema_unet: 0
-  ema_text_encoder: 0
-
-data:
-  dataset1:
-    batch_size: 4
-    cache_latents: True
-
-    source:
-      data_source1:
-        img_root: 'imgs/'
-        prompt_template: 'prompt_tuning_template/object.txt'
-        caption_file: null # path to image captions (file_words)
-
-        word_names:
-          pt1: sks
-          class: dog
-    bucket:
-      _target_: hcpdiff.data.bucket.RatioBucket.from_files # aspect ratio bucket
-      target_area: ${hcp.eval:"512*512"}
-      num_bucket: 1
-
-  dataset_class:
-    batch_size: 1
-    cache_latents: True
-    loss_weight: 1.0
-
-    source:
-      data_source1:
-        img_root: 'imgs/db_class'
-        prompt_template: 'prompt_tuning_template/object.txt'
-        caption_file: null
-
-        word_names:
-          class: dog
-    bucket:
-      _target_: hcpdiff.data.bucket.FixedBucket
-      target_size: [512, 512]