| from transformers.configuration_utils import PretrainedConfig |
| from transformers.modeling_rope_utils import rope_config_validation |
| from transformers.utils import logging |
|
|
| logger = logging.get_logger(__name__) |
|
|
| class VAEEncoderConfig(PretrainedConfig): |
| model_type = "umm" |
| base_config_key = "vae_encoder_config" |
|
|
| def __init__( |
| self, |
| resolution=256, |
| in_channels=3, |
| ch=128, |
| ch_mult=[1,2,4,4], |
| num_res_blocks=2, |
| z_channels=32, |
| **kwargs |
| ): |
| self.resolution = resolution |
| self.in_channels = in_channels |
| self.ch = ch |
| self.ch_mult = ch_mult |
| self.num_res_blocks = num_res_blocks |
| self.z_channels = z_channels |
| super().__init__(**kwargs) |
|
|
| class VAEDecoderConfig(PretrainedConfig): |
| model_type = "umm" |
| base_config_key = "vae_decoder_config" |
|
|
| def __init__( |
| self, |
| ch=128, |
| out_ch=3, |
| ch_mult=[1,2,4,4], |
| num_res_blocks=2, |
| in_channels=3, |
| resolution=256, |
| z_channels=32, |
| **kwargs |
| ): |
| self.resolution = resolution |
| self.in_channels = in_channels |
| self.ch = ch |
| self.out_ch = out_ch |
| self.ch_mult = ch_mult |
| self.num_res_blocks = num_res_blocks |
| self.z_channels = z_channels |
| super().__init__(**kwargs) |
|
|
| class Siglip2VisionConfig(PretrainedConfig): |
| model_type = "umm" |
| base_config_key = "vision_representation_config" |
|
|
| def __init__( |
| self, |
| hidden_size=768, |
| intermediate_size=3072, |
| num_hidden_layers=12, |
| num_attention_heads=12, |
| num_channels=3, |
| image_size=256, |
| patch_size=16, |
| hidden_act="gelu_pytorch_tanh", |
| layer_norm_eps=1e-6, |
| attention_dropout=0.0, |
| **kwargs, |
| ): |
| super().__init__(**kwargs) |
|
|
| self.hidden_size = hidden_size |
| self.intermediate_size = intermediate_size |
| self.num_hidden_layers = num_hidden_layers |
| self.num_attention_heads = num_attention_heads |
| self.num_channels = num_channels |
| self.patch_size = patch_size |
| self.attention_dropout = attention_dropout |
| self.layer_norm_eps = layer_norm_eps |
| self.hidden_act = hidden_act |
| self.image_size = image_size |
|
|
| class Qwen2Config(PretrainedConfig): |
| model_type = "umm" |
| base_config_key = "text_config" |
|
|
| def __init__( |
| self, |
| vocab_size=152064, |
| hidden_size=3584, |
| intermediate_size=18944, |
| num_hidden_layers=28, |
| num_attention_heads=28, |
| num_key_value_heads=4, |
| hidden_act="silu", |
| max_position_embeddings=131072, |
| initializer_range=0.02, |
| rms_norm_eps=1e-6, |
| use_cache=True, |
| tie_word_embeddings=False, |
| rope_theta=1000000.0, |
| rope_scaling=None, |
| use_sliding_window=False, |
| sliding_window=131072, |
| max_window_layers=28, |
| layer_types=None, |
| attention_dropout=0.0, |
| **kwargs, |
| ): |
| self.vocab_size = vocab_size |
| self.max_position_embeddings = max_position_embeddings |
| self.hidden_size = hidden_size |
| self.intermediate_size = intermediate_size |
| self.num_hidden_layers = num_hidden_layers |
| self.num_attention_heads = num_attention_heads |
| self.use_sliding_window = use_sliding_window |
| self.sliding_window = sliding_window if self.use_sliding_window else None |
| self.max_window_layers = max_window_layers |
| |
| if num_key_value_heads is None: |
| num_key_value_heads = num_attention_heads |
|
|
| self.num_key_value_heads = num_key_value_heads |
| self.hidden_act = hidden_act |
| self.initializer_range = initializer_range |
| self.rms_norm_eps = rms_norm_eps |
| self.use_cache = use_cache |
| self.rope_theta = rope_theta |
| self.rope_scaling = rope_scaling |
| self.attention_dropout = attention_dropout |
| |
| |
| if self.rope_scaling is not None and "type" in self.rope_scaling: |
| self.rope_scaling["rope_type"] = self.rope_scaling["type"] |
| rope_config_validation(self) |
|
|
| self.layer_types = layer_types |
| if self.layer_types is None: |
| self.layer_types = [ |
| "sliding_attention" |
| if self.sliding_window is not None and i >= self.max_window_layers |
| else "full_attention" |
| for i in range(self.num_hidden_layers) |
| ] |
|
|
| super().__init__( |
| tie_word_embeddings=tie_word_embeddings, |
| **kwargs, |
| ) |
|
|
| class UMMConfig(PretrainedConfig): |
| model_type = "umm" |
| sub_configs = { |
| "vision_representation_config": Siglip2VisionConfig, |
| "vae_encoder_config": VAEEncoderConfig, |
| "vae_decoder_config": VAEDecoderConfig, |
| "text_config": Qwen2Config, |
| } |
| keys_to_ignore_at_inference = ["past_key_values"] |
| def __init__( |
| self, |
| vision_representation_config=None, |
| vae_encoder_config=None, |
| vae_decoder_config=None, |
| text_config=None, |
| **kwargs, |
| ): |
| if isinstance(vision_representation_config, dict): |
| self.vision_representation_config = self.sub_configs["vision_representation_config"](**vision_representation_config) |
| elif vision_representation_config is None: |
| self.vision_representation_config = self.sub_configs["vision_representation_config"]() |
|
|
| if isinstance(vae_encoder_config, dict): |
| self.vae_encoder_config = self.sub_configs["vae_encoder_config"](**vae_encoder_config) |
| elif vae_encoder_config is None: |
| self.vae_encoder_config = self.sub_configs["vae_encoder_config"]() |
| |
| if isinstance(vae_decoder_config, dict): |
| self.vae_decoder_config = self.sub_configs["vae_decoder_config"](**vae_decoder_config) |
| elif vae_decoder_config is None: |
| self.vae_decoder_config = self.sub_configs["vae_decoder_config"]() |
| |
| if isinstance(text_config, dict): |
| self.text_config = self.sub_configs["text_config"](**text_config) |
| elif text_config is None: |
| self.text_config = self.sub_configs["text_config"]() |
|
|
| super().__init__(**kwargs) |
|
|
| __all__ = ["UMMConfig"] |
