ai-edge-torch-nightly 0.2.0.dev20240609__py3-none-any.whl → 0.2.0.dev20240610__py3-none-any.whl

ai_edge_torch/generative/examples/gemma/gemma.py

@@ -116,7 +116,7 @@ def get_model_config_2b(kv_cache_max_len: int = 1024) -> cfg.ModelConfig:
   )
   ff_config = cfg.FeedForwardConfig(
       type=cfg.FeedForwardType.GATED,
-      activation=cfg.ActivationType.GELU_TANH,
+      activation=cfg.ActivationConfig(cfg.ActivationType.GELU_TANH),
       intermediate_size=16384,
   )
   norm_config = cfg.NormalizationConfig(

ai_edge_torch/generative/examples/phi2/phi2.py

@@ -112,7 +112,7 @@ def get_model_config(kv_cache_max_len: int = 1024) -> cfg.ModelConfig:
   )
   ff_config = cfg.FeedForwardConfig(
       type=cfg.FeedForwardType.SEQUENTIAL,
-      activation=cfg.ActivationType.GELU_TANH,
+      activation=cfg.ActivationConfig(cfg.ActivationType.GELU_TANH),
       intermediate_size=10240,
       use_bias=True,
   )

ai_edge_torch/generative/examples/stable_diffusion/clip.py

@@ -90,7 +90,7 @@ def get_model_config() -> cfg.ModelConfig:
 
   ff_config = cfg.FeedForwardConfig(
       type=cfg.FeedForwardType.SEQUENTIAL,
-      activation=cfg.ActivationType.GELU_QUICK,
+      activation=cfg.ActivationConfig(cfg.ActivationType.GELU_QUICK),
       intermediate_size=embedding_dim * 4,
       use_bias=True,
   )

ai_edge_torch/generative/examples/stable_diffusion/decoder.py

@@ -221,7 +221,7 @@ class Decoder(nn.Module):
                   in_channels=prev_output_channel,
                   out_channels=block_out_channels,
                   normalization_config=config.normalization_config,
-                  activation_type=config.activation_type,
+                  activation_config=config.activation_config,
                   num_layers=config.layers_per_block,
                   add_upsample=not_final_block,
                   upsample_conv=True,
@@ -235,7 +235,7 @@ class Decoder(nn.Module):
     self.final_norm = layers_builder.build_norm(
         block_out_channels, config.normalization_config
     )
-    self.act_fn = layers_builder.get_activation(config.activation_type)
+    self.act_fn = layers_builder.get_activation(config.activation_config)
     self.conv_out = nn.Conv2d(
         block_out_channels,
         config.out_channels,
@@ -287,7 +287,7 @@ def get_model_config() -> unet_cfg.AutoEncoderConfig:
   mid_block_config = unet_cfg.MidBlock2DConfig(
       in_channels=block_out_channels[-1],
       normalization_config=norm_config,
-      activation_type=layers_cfg.ActivationType.SILU,
+      activation_config=layers_cfg.ActivationConfig(layers_cfg.ActivationType.SILU),
       num_layers=1,
       attention_block_config=att_config,
   )
@@ -296,7 +296,7 @@ def get_model_config() -> unet_cfg.AutoEncoderConfig:
       in_channels=in_channels,
       latent_channels=latent_channels,
       out_channels=out_channels,
-      activation_type=layers_cfg.ActivationType.SILU,
+      activation_config=layers_cfg.ActivationConfig(layers_cfg.ActivationType.SILU),
       block_out_channels=block_out_channels,
       scaling_factor=scaling_factor,
       layers_per_block=layers_per_block,

ai_edge_torch/generative/examples/stable_diffusion/diffusion.py

@@ -130,7 +130,7 @@ class Upsample(nn.Module):
     self.conv = nn.Conv2d(channels, channels, kernel_size=3, padding=1)
 
   def forward(self, x):
-    x = F.interpolate(x, scale_factor=2, mode='nearest')
+    x = F.interpolate(x, scale_factor=2, mode="nearest")
     return self.conv(x)
 
 
@@ -237,3 +237,8 @@ class Diffusion(nn.Module):
     output = self.unet(latent, context, time)
     output = self.final(output)
     return output
+
+
+if __name__ == "__main__":
+  diffusion = Diffusion()
+  print(diffusion.state_dict().keys())

ai_edge_torch/generative/examples/t5/t5.py

@@ -349,7 +349,7 @@ def get_model_config_t5() -> cfg.ModelConfig:
   )
   ff_config = cfg.FeedForwardConfig(
       type=cfg.FeedForwardType.SEQUENTIAL,
-      activation=cfg.ActivationType.RELU,
+      activation=cfg.ActivationConfig(cfg.ActivationType.RELU),
       intermediate_size=3072,
   )
   # T5 Confirmed as RMS Norm and eps = 1e-6 TJA.

ai_edge_torch/generative/examples/test_models/toy_model.py

@@ -76,7 +76,7 @@ def define_and_run() -> None:
   )
   ff_config = cfg.FeedForwardConfig(
       type=cfg.FeedForwardType.GATED,
-      activation=cfg.ActivationType.SILU,
+      activation=cfg.ActivationConfig(cfg.ActivationType.SILU),
       intermediate_size=256,
   )
   norm_config = cfg.NormalizationConfig(type=cfg.NormalizationType.RMS_NORM)

ai_edge_torch/generative/examples/test_models/toy_model_with_external_kv_cache.py

@@ -95,7 +95,7 @@ def get_model_config() -> cfg.ModelConfig:
   )
   ff_config = cfg.FeedForwardConfig(
       type=cfg.FeedForwardType.GATED,
-      activation=cfg.ActivationType.SILU,
+      activation=cfg.ActivationConfig(cfg.ActivationType.SILU),
       intermediate_size=256,
   )
   norm_config = cfg.NormalizationConfig(type=cfg.NormalizationType.RMS_NORM)

ai_edge_torch/generative/examples/test_models/toy_model_with_kv_cache.py

@@ -83,7 +83,7 @@ def get_model_config() -> cfg.ModelConfig:
   )
   ff_config = cfg.FeedForwardConfig(
       type=cfg.FeedForwardType.GATED,
-      activation=cfg.ActivationType.SILU,
+      activation=cfg.ActivationConfig(cfg.ActivationType.SILU),
       intermediate_size=256,
   )
   norm_config = cfg.NormalizationConfig(type=cfg.NormalizationType.RMS_NORM)

ai_edge_torch/generative/examples/tiny_llama/tiny_llama.py

@@ -112,7 +112,7 @@ def get_model_config(kv_cache_max_len: int = 1024) -> cfg.ModelConfig:
   )
   ff_config = cfg.FeedForwardConfig(
       type=cfg.FeedForwardType.GATED,
-      activation=cfg.ActivationType.SILU,
+      activation=cfg.ActivationConfig(cfg.ActivationType.SILU),
       intermediate_size=5632,
   )
   norm_config = cfg.NormalizationConfig(type=cfg.NormalizationType.RMS_NORM)

ai_edge_torch/generative/layers/builder.py

@@ -13,6 +13,7 @@
 # limitations under the License.
 # ==============================================================================
 # Builder class for individual components.
+import torch
 from torch import nn
 import torch.nn.functional as F
 
@@ -21,6 +22,23 @@ import ai_edge_torch.generative.layers.model_config as cfg
 import ai_edge_torch.generative.layers.normalization as normalization
 
 
+class GeGLU(nn.Module):
+  """GeGLU is an activation function which is a variant of GELU.
+
+  GeGLU(x) = (xW+b) * GELU(xV+c)
+  See: https://arxiv.org/abs/2002.05202v1
+
+  """
+
+  def __init__(self, d_in: int, d_out: int):
+    super().__init__()
+    self.proj = nn.Linear(d_in, d_out * 2)
+
+  def forward(self, x: torch.Tensor):
+    x, gate = self.proj(x).chunk(2, dim=-1)
+    return x * F.gelu(gate)
+
+
 def build_norm(dim: int, config: cfg.NormalizationConfig):
   """Builder function for normalizers.
 
@@ -81,29 +99,33 @@ def build_ff(dim: int, config: cfg.FeedForwardConfig):
   )
 
 
-def get_activation(type_: cfg.ActivationType):
-  """Get pytorch callable activation from the name.
+def get_activation(config: cfg.ActivationConfig):
+  """Get pytorch callable activation from the activation config.
 
   Args:
-    name (string): activation's name.
+    config (cfg.ActivationConfig): activation config.
 
   Returns:
     Activation function.
 
   Raises:
-    ValueError: If activation name is not supported.
+    ValueError: If activation config is not supported.
   """
-  if type_ == cfg.ActivationType.SILU:
+  if config.type == cfg.ActivationType.LINEAR:
+    return lambda x: x
+  elif config.type == cfg.ActivationType.SILU:
     return F.silu
-  elif type_ == cfg.ActivationType.GELU:
+  elif config.type == cfg.ActivationType.GELU:
     return F.gelu
-  elif type_ == cfg.ActivationType.GELU_TANH:
+  elif config.type == cfg.ActivationType.GELU_TANH:
     return lambda x: F.gelu(x, approximate="tanh")
-  elif type_ == cfg.ActivationType.GELU_QUICK:
+  elif config.type == cfg.ActivationType.GELU_QUICK:
     # GELU approximation that is fast but somewhat inaccurate.
     # See: https://github.com/hendrycks/GELUs
     return lambda x: x * F.sigmoid(1.702 * x)
-  elif type_ == cfg.ActivationType.RELU:
+  elif config.type == cfg.ActivationType.GE_GLU:
+    return GeGLU(config.dim_in, config.dim_out)
+  elif config.type == cfg.ActivationType.RELU:
     return F.relu
   else:
     raise ValueError("Unsupported activation type.")

ai_edge_torch/generative/layers/model_config.py

@@ -27,7 +27,7 @@ class ActivationType(enum.Enum):
   SILU = enum.auto()
   GELU = enum.auto()
   GELU_TANH = enum.auto()
-  GELU_QUICK = enum.auto()
+  GE_GLU = enum.auto()
   RELU = enum.auto()
 
 
@@ -74,12 +74,20 @@ class AttentionConfig:
   relative_attention_max_distance: int = 0
 
 
+@dataclass
+class ActivationConfig:
+  type: ActivationType = ActivationType.LINEAR
+  # Dimension of input and output, used in GeGLU.
+  dim_in: Optional[int] = None
+  dim_out: Optional[int] = None
+
+
 @dataclass
 class FeedForwardConfig:
   """FeedForward module's parameters."""
 
   type: FeedForwardType
-  activation: ActivationType
+  activation: ActivationConfig
   intermediate_size: int
   use_bias: bool = False
 

ai_edge_torch/generative/layers/unet/blocks_2d.py

@@ -53,7 +53,7 @@ class ResidualBlock2D(nn.Module):
     self.conv_2 = nn.Conv2d(
         config.out_channels, config.out_channels, kernel_size=3, stride=1, padding=1
     )
-    self.act_fn = layers_builder.get_activation(config.activation_type)
+    self.act_fn = layers_builder.get_activation(config.activation_config)
     if config.in_channels == config.out_channels:
       self.residual_layer = nn.Identity()
     else:
@@ -167,7 +167,7 @@ class UpDecoderBlock2D(nn.Module):
                   out_channels=config.out_channels,
                   time_embedding_channels=config.time_embedding_channels,
                   normalization_config=config.normalization_config,
-                  activation_type=config.activation_type,
+                  activation_config=config.activation_config,
               )
           )
       )
@@ -244,7 +244,7 @@ class MidBlock2D(nn.Module):
                 out_channels=config.in_channels,
                 time_embedding_channels=config.time_embedding_channels,
                 normalization_config=config.normalization_config,
-                activation_type=config.activation_type,
+                activation_config=config.activation_config,
             )
         )
     ]
@@ -259,7 +259,7 @@ class MidBlock2D(nn.Module):
                   out_channels=config.in_channels,
                   time_embedding_channels=config.time_embedding_channels,
                   normalization_config=config.normalization_config,
-                  activation_type=config.activation_type,
+                  activation_config=config.activation_config,
               )
           )
       )

ai_edge_torch/generative/layers/unet/model_config.py

@@ -39,7 +39,7 @@ class ResidualBlock2DConfig:
   in_channels: int
   out_channels: int
   normalization_config: layers_cfg.NormalizationConfig
-  activation_type: layers_cfg.ActivationType
+  activation_config: layers_cfg.ActivationConfig
   # Optional time embedding channels if the residual block takes a time embedding context as input
   time_embedding_channels: Optional[int] = None
 
@@ -56,7 +56,7 @@ class UpDecoderBlock2DConfig:
   in_channels: int
   out_channels: int
   normalization_config: layers_cfg.NormalizationConfig
-  activation_type: layers_cfg.ActivationType
+  activation_config: layers_cfg.ActivationConfig
   num_layers: int
   # Optional time embedding channels if the residual blocks take a time embedding context as input
   time_embedding_channels: Optional[int] = None
@@ -72,7 +72,7 @@ class UpDecoderBlock2DConfig:
 class MidBlock2DConfig:
   in_channels: int
   normalization_config: layers_cfg.NormalizationConfig
-  activation_type: layers_cfg.ActivationType
+  activation_config: layers_cfg.ActivationConfig
   num_layers: int
   # Optional time embedding channels if the residual blocks take a time embedding context as input
   time_embedding_channels: Optional[int] = None
@@ -85,7 +85,7 @@ class AutoEncoderConfig:
   """Configurations of encoder/decoder in the autoencoder model."""
 
   # The activation type of encoder/decoder blocks.
-  activation_type: layers_cfg.ActivationType
+  activation_config: layers_cfg.ActivationConfig
 
   # The output channels of each block.
   block_out_channels: List[int]

{ai_edge_torch_nightly-0.2.0.dev20240609.dist-info → ai_edge_torch_nightly-0.2.0.dev20240610.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: ai-edge-torch-nightly
-Version: 0.2.0.dev20240609
+Version: 0.2.0.dev20240610
 Summary: Supporting PyTorch models with the Google AI Edge TFLite runtime.
 Home-page: https://github.com/google-ai-edge/ai-edge-torch
 Keywords: On-Device ML,AI,Google,TFLite,PyTorch,LLMs,GenAI

{ai_edge_torch_nightly-0.2.0.dev20240609.dist-info → ai_edge_torch_nightly-0.2.0.dev20240610.dist-info}/RECORD

@@ -34,16 +34,16 @@ ai_edge_torch/generative/__init__.py,sha256=hHLluseD2R0Hh4W6XZRIXY_dRQeYudjsrKGf
 ai_edge_torch/generative/examples/__init__.py,sha256=hHLluseD2R0Hh4W6XZRIXY_dRQeYudjsrKGf6LZz65g,671
 ai_edge_torch/generative/examples/gemma/__init__.py,sha256=hHLluseD2R0Hh4W6XZRIXY_dRQeYudjsrKGf6LZz65g,671
 ai_edge_torch/generative/examples/gemma/convert_to_tflite.py,sha256=dZv3r24uHsTMokEdnl3nf7LpmV0q7FLnVtCuHn5AuUs,2538
-ai_edge_torch/generative/examples/gemma/gemma.py,sha256=YF4Ua-1lnL3qhQnh1sY5-HlYw2Dq6ZRm227XyDe7WAw,5913
+ai_edge_torch/generative/examples/gemma/gemma.py,sha256=1lZfXGHmbII4rFu0U2B9NzlJCRhphxtmQtkCHQ39_uw,5935
 ai_edge_torch/generative/examples/phi2/__init__.py,sha256=hHLluseD2R0Hh4W6XZRIXY_dRQeYudjsrKGf6LZz65g,671
 ai_edge_torch/generative/examples/phi2/convert_to_tflite.py,sha256=6nOuwx9q3AUlYcjXRRXSr_3M2JKqdJ-vUf-uE3VFYHE,2512
-ai_edge_torch/generative/examples/phi2/phi2.py,sha256=VvigzPQ_LJHeADTsMliwFwPe2BcnOhFgKDqr_WZ2JQ8,5540
+ai_edge_torch/generative/examples/phi2/phi2.py,sha256=PMhKC6JCAMYSj2F3UmWHWK4rTcXD-B6PuehaoDccRqk,5562
 ai_edge_torch/generative/examples/stable_diffusion/__init__.py,sha256=hHLluseD2R0Hh4W6XZRIXY_dRQeYudjsrKGf6LZz65g,671
 ai_edge_torch/generative/examples/stable_diffusion/attention.py,sha256=Lo4Dq7a3Kg-lyH56iqGtqCo5UaClQHRCTDdNagXGTo8,3535
-ai_edge_torch/generative/examples/stable_diffusion/clip.py,sha256=yUCJemEh4n8ez-yLgVU0HZAki-PZ9nY04DFjgpx9PUc,3698
+ai_edge_torch/generative/examples/stable_diffusion/clip.py,sha256=qU1wVEcn_biwCuDguZljhlLGzpLIqgqC31Dh_lXquQc,3720
 ai_edge_torch/generative/examples/stable_diffusion/convert_to_tflite.py,sha256=NmgDo5uAefrhMUbYku0TKHlqzO0NVWI_M1ue8tddQR4,4024
-ai_edge_torch/generative/examples/stable_diffusion/decoder.py,sha256=Z1bnZvYtPdwNy706kixVDfL32X-R87B_WF3CcHwiz0o,11038
-ai_edge_torch/generative/examples/stable_diffusion/diffusion.py,sha256=TfbfsmuKoGsBENF9fYIAN_SMEQNhj-kjNdqQXFJGxpg,7784
+ai_edge_torch/generative/examples/stable_diffusion/decoder.py,sha256=meW8t-3BDdjFs5vCAf76cn6lGx49a_GcEvnVa9R5if4,11106
+ai_edge_torch/generative/examples/stable_diffusion/diffusion.py,sha256=_gEeUxa9Xyd3iLb_fyeUefHKuELVDorDlQs8e7wdXKg,7878
 ai_edge_torch/generative/examples/stable_diffusion/encoder.py,sha256=mgbxkeFDMkNIGmnbcFTIFPu8EWKokghiviYIOB2lE3Q,3437
 ai_edge_torch/generative/examples/stable_diffusion/pipeline.py,sha256=FCbnwlkpYYb-tF7KscbSYjNEdg7XnuLju1cDuIRoQv8,8277
 ai_edge_torch/generative/examples/stable_diffusion/tokenizer.py,sha256=r9RqbyNvuvXOGu3ojtl7ZmbC7o4Pt8aUKAhN1yCdtEc,3397
@@ -55,29 +55,29 @@ ai_edge_torch/generative/examples/stable_diffusion/samplers/k_lms.py,sha256=iPYX
 ai_edge_torch/generative/examples/stable_diffusion/samplers/sampler.py,sha256=5iRfU5MO6GR6K3WrdddIU_9U7ZZGEEb7zGKVY1WFl-8,1340
 ai_edge_torch/generative/examples/t5/__init__.py,sha256=hHLluseD2R0Hh4W6XZRIXY_dRQeYudjsrKGf6LZz65g,671
 ai_edge_torch/generative/examples/t5/convert_to_tflite.py,sha256=bWtwtUacvJOEDUpuYvLTgkP7oTkXKJA-Tf4FPxlD1Cw,4536
-ai_edge_torch/generative/examples/t5/t5.py,sha256=q2gG5RRo7RgNzvHXYC0Juh6Tgt5d_RTMSWFaYvOKiZU,21065
+ai_edge_torch/generative/examples/t5/t5.py,sha256=L6YrVzUEzP-Imb8W28LdukFGrx1aWSzz1kyYK_9RFZM,21087
 ai_edge_torch/generative/examples/t5/t5_attention.py,sha256=rRgwCEdVtzcJEaGbbBjw8HxCxrCX3pXA5nelawdYiME,9036
 ai_edge_torch/generative/examples/test_models/__init__.py,sha256=hHLluseD2R0Hh4W6XZRIXY_dRQeYudjsrKGf6LZz65g,671
-ai_edge_torch/generative/examples/test_models/toy_model.py,sha256=EV07_MEG3fv9g0ZGu9gbBd5BjjrGkxCT1pv7dvhz4TI,3791
-ai_edge_torch/generative/examples/test_models/toy_model_with_external_kv_cache.py,sha256=rzL5h7Z5DIEgfpc1pWgYHdKt2aR8ha_CUqTKQBSPBaU,5521
-ai_edge_torch/generative/examples/test_models/toy_model_with_kv_cache.py,sha256=MUr6fSj2hBuYSlNbZtrBBpzqB_0WY-l_xYcd_TFFUjY,4831
+ai_edge_torch/generative/examples/test_models/toy_model.py,sha256=CUXsQ_IU96NaCg9jyfeKI0Zz2iWDkJUsPJyPR1Pgz7I,3813
+ai_edge_torch/generative/examples/test_models/toy_model_with_external_kv_cache.py,sha256=zwCmCnhr-vhBwHqv9i7xMasdBGVNqAGxZvWsncsJn58,5543
+ai_edge_torch/generative/examples/test_models/toy_model_with_kv_cache.py,sha256=lfYUiem_Pbn3vGgPx84BeI8n7rN3-1fImwCLm8Eo2U8,4853
 ai_edge_torch/generative/examples/tiny_llama/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 ai_edge_torch/generative/examples/tiny_llama/convert_to_tflite.py,sha256=E4I5OlC4zyl5cxiiu7uTED-zcwYRu210lP1zuT3xLBE,2566
-ai_edge_torch/generative/examples/tiny_llama/tiny_llama.py,sha256=hVGpuI8gpj4Rn9k4otsRE22MSLFHBDlUOgioY6Ru6VI,5629
+ai_edge_torch/generative/examples/tiny_llama/tiny_llama.py,sha256=IFRLPG9wz_aLl_zV_6CETCjSM03ukA6bZqqyDLVACuw,5651
 ai_edge_torch/generative/layers/__init__.py,sha256=hHLluseD2R0Hh4W6XZRIXY_dRQeYudjsrKGf6LZz65g,671
 ai_edge_torch/generative/layers/attention.py,sha256=Z8gXHYs6h8gaRiYAdvYUbHzg_2EmqfxiChsf_SYraAc,7902
 ai_edge_torch/generative/layers/attention_utils.py,sha256=hXhuyKblPPxKIRzlAf1YNlwHgpbj-6nReRLhRHELx5k,6350
-ai_edge_torch/generative/layers/builder.py,sha256=8cPL1NAutjT6Dwtyy2X7NSaTl9WCUJM5SIrBIDcEvVY,3520
+ai_edge_torch/generative/layers/builder.py,sha256=jAyrR5hsSI0aimKZumyvxdJ1GovERIfsK0g-dezX2gs,4163
 ai_edge_torch/generative/layers/feed_forward.py,sha256=4j2QaSCw59Jkk_ixKDpKEj7FLRauzuExTiSNRzAjAhE,2820
 ai_edge_torch/generative/layers/kv_cache.py,sha256=4uiZLO3om5G3--kT04Jt0esEYznbkJ7QLzSHfb8mjc4,3090
-ai_edge_torch/generative/layers/model_config.py,sha256=72DXOsFC0buvzZp6YyVjuTVrpphAubBJ5NJWfs3kEwk,4362
+ai_edge_torch/generative/layers/model_config.py,sha256=g_XJXcQOCkE-mt58fSH4-T4GY_uLeMilg6mxwDMCfz4,4557
 ai_edge_torch/generative/layers/normalization.py,sha256=M27eW3TcNK20oaXClXtfnu0lLWrAGrSKSsbegRWnj3c,1867
 ai_edge_torch/generative/layers/rotary_position_embedding.py,sha256=12SsCuoRuLNCwnFGe_pHDOZEBwBcqXs87Aj0PaWWw4E,1383
 ai_edge_torch/generative/layers/scaled_dot_product_attention.py,sha256=dYafGC205QE5CLIbBTCI-7eVvEGZEHzs1toPEhemeDs,3391
 ai_edge_torch/generative/layers/unet/__init__.py,sha256=hHLluseD2R0Hh4W6XZRIXY_dRQeYudjsrKGf6LZz65g,671
-ai_edge_torch/generative/layers/unet/blocks_2d.py,sha256=KuZd2oZhkCQSknSgXMBla-sfYBPUv5bZNf9RYKXHfGg,10052
+ai_edge_torch/generative/layers/unet/blocks_2d.py,sha256=7mHyJYq9lq5zVYp4mEz-R8Az3FFngi711YC20KP6ED8,10066
 ai_edge_torch/generative/layers/unet/builder.py,sha256=iH0_nuY9TF2ap5h1JbGNCOonPTfrXQHcF8U0slrIREM,1210
-ai_edge_torch/generative/layers/unet/model_config.py,sha256=LeRGB34fQ73UlknlFpjM9U-SZIRcQDnSmDltJivX-UA,4044
+ai_edge_torch/generative/layers/unet/model_config.py,sha256=sbtbDEHmMV9GLKngwjsNvqm8wovLxnlidkQbXdXkXKs,4060
 ai_edge_torch/generative/quantize/__init__.py,sha256=hHLluseD2R0Hh4W6XZRIXY_dRQeYudjsrKGf6LZz65g,671
 ai_edge_torch/generative/quantize/example.py,sha256=t-YwyKSPAG-OZC1DfH-0vfie2RHHpTSQjxUY-tmhu5g,1543
 ai_edge_torch/generative/quantize/quant_attrs.py,sha256=ffBALrrbrfiG_mrOr-f3B1Gc6PlAma9gtvVnfP7SDzI,1862
@@ -107,8 +107,8 @@ ai_edge_torch/quantize/quant_config.py,sha256=ExThdTXqnWmGC3-F6sdXbXr8nYzkEe_qCz
 ai_edge_torch/testing/__init__.py,sha256=hHLluseD2R0Hh4W6XZRIXY_dRQeYudjsrKGf6LZz65g,671
 ai_edge_torch/testing/model_coverage/__init__.py,sha256=5P8J6Zk5YYtDvTBucFvB9NGSRI7Gw_24WnrbhXgycEE,765
 ai_edge_torch/testing/model_coverage/model_coverage.py,sha256=EIyKz-HY70DguWuSrJal8LpYXQ5ZSEUf3ZrVl7jikFM,4286
-ai_edge_torch_nightly-0.2.0.dev20240609.dist-info/LICENSE,sha256=z8d0m5b2O9McPEK1xHG_dWgUBT6EfBDz6wA0F7xSPTA,11358
-ai_edge_torch_nightly-0.2.0.dev20240609.dist-info/METADATA,sha256=THUdn03pVtqLTQHdigvUA8B32EHwfV5UTvBRQCpR8v0,1748
-ai_edge_torch_nightly-0.2.0.dev20240609.dist-info/WHEEL,sha256=GJ7t_kWBFywbagK5eo9IoUwLW6oyOeTKmQ-9iHFVNxQ,92
-ai_edge_torch_nightly-0.2.0.dev20240609.dist-info/top_level.txt,sha256=5KXRaF2hwkApYxf7Y8y_tVb9aulGTlbOoNdbx1aKRkE,14
-ai_edge_torch_nightly-0.2.0.dev20240609.dist-info/RECORD,,
+ai_edge_torch_nightly-0.2.0.dev20240610.dist-info/LICENSE,sha256=z8d0m5b2O9McPEK1xHG_dWgUBT6EfBDz6wA0F7xSPTA,11358
+ai_edge_torch_nightly-0.2.0.dev20240610.dist-info/METADATA,sha256=6hL5PV3S56VU2l6xqS-YrmzMZeajtXsikIdR7kDYcWE,1748
+ai_edge_torch_nightly-0.2.0.dev20240610.dist-info/WHEEL,sha256=GJ7t_kWBFywbagK5eo9IoUwLW6oyOeTKmQ-9iHFVNxQ,92
+ai_edge_torch_nightly-0.2.0.dev20240610.dist-info/top_level.txt,sha256=5KXRaF2hwkApYxf7Y8y_tVb9aulGTlbOoNdbx1aKRkE,14
+ai_edge_torch_nightly-0.2.0.dev20240610.dist-info/RECORD,,