keras-hub-nightly 0.22.0.dev202508110431__tar.gz → 0.22.0.dev202508140419__tar.gz

{keras_hub_nightly-0.22.0.dev202508110431 → keras_hub_nightly-0.22.0.dev202508140419}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: keras-hub-nightly
-Version: 0.22.0.dev202508110431
+Version: 0.22.0.dev202508140419
 Summary: Pretrained models for Keras.
 Author-email: Keras team <keras-users@googlegroups.com>
 License-Expression: Apache-2.0
@@ -18,9 +18,9 @@ Classifier: Operating System :: MacOS
 Classifier: Intended Audience :: Science/Research
 Classifier: Topic :: Scientific/Engineering
 Classifier: Topic :: Software Development
-Requires-Python: >=3.9
+Requires-Python: >=3.10
 Description-Content-Type: text/markdown
-Requires-Dist: keras>=3.5
+Requires-Dist: keras>=3.8
 Requires-Dist: absl-py
 Requires-Dist: numpy
 Requires-Dist: packaging
@@ -31,7 +31,7 @@ Requires-Dist: tensorflow-text; platform_system != "Windows"
 
 # KerasHub: Multi-framework Pretrained Models
 [![](https://github.com/keras-team/keras-hub/workflows/Tests/badge.svg?branch=master)](https://github.com/keras-team/keras-hub/actions?query=workflow%3ATests+branch%3Amaster)
-![Python](https://img.shields.io/badge/python-v3.9.0+-success.svg)
+![Python](https://img.shields.io/badge/python-v3.10.0+-success.svg)
 [![contributions welcome](https://img.shields.io/badge/contributions-welcome-brightgreen.svg?style=flat)](https://github.com/keras-team/keras-hub/issues)
 
 > [!IMPORTANT]

{keras_hub_nightly-0.22.0.dev202508110431 → keras_hub_nightly-0.22.0.dev202508140419}/README.md

@@ -1,6 +1,6 @@
 # KerasHub: Multi-framework Pretrained Models
 [![](https://github.com/keras-team/keras-hub/workflows/Tests/badge.svg?branch=master)](https://github.com/keras-team/keras-hub/actions?query=workflow%3ATests+branch%3Amaster)
-![Python](https://img.shields.io/badge/python-v3.9.0+-success.svg)
+![Python](https://img.shields.io/badge/python-v3.10.0+-success.svg)
 [![contributions welcome](https://img.shields.io/badge/contributions-welcome-brightgreen.svg?style=flat)](https://github.com/keras-team/keras-hub/issues)
 
 > [!IMPORTANT]

keras_hub_nightly-0.22.0.dev202508140419/keras_hub/src/models/deit/__init__.py

@@ -0,0 +1,5 @@
+from keras_hub.src.models.deit.deit_backbone import DeiTBackbone
+from keras_hub.src.models.deit.deit_presets import backbone_presets
+from keras_hub.src.utils.preset_utils import register_presets
+
+register_presets(backbone_presets, DeiTBackbone)

{keras_hub_nightly-0.22.0.dev202508110431 → keras_hub_nightly-0.22.0.dev202508140419}/keras_hub/src/models/deit/deit_presets.py

@@ -2,46 +2,46 @@
 
 # Metadata for loading pretrained model weights.
 backbone_presets = {
-    "deit-base-distilled-patch16-384_imagenet": {
+    "deit_base_distilled_patch16_384_imagenet": {
         "metadata": {
             "description": (
                 "DeiT-B16 model pre-trained on the ImageNet 1k dataset with "
                 "image resolution of 384x384 "
             ),
-            "params": 86092032,
+            "params": 86_092_032,
             "path": "deit",
         },
         "kaggle_handle": "kaggle://keras/deit/keras/deit_base_distilled_patch16_384_imagenet/1",
     },
-    "deit-base-distilled-patch16-224_imagenet": {
+    "deit_base_distilled_patch16_224_imagenet": {
         "metadata": {
             "description": (
                 "DeiT-B16 model pre-trained on the ImageNet 1k dataset with "
                 "image resolution of 224x224 "
             ),
-            "params": 85800192,
+            "params": 85_800_192,
             "path": "deit",
         },
         "kaggle_handle": "kaggle://keras/deit/keras/deit_base_distilled_patch16_224_imagenet/1",
     },
-    "deit-tiny-distilled-patch16-224_imagenet": {
+    "deit_tiny_distilled_patch16_224_imagenet": {
         "metadata": {
             "description": (
                 "DeiT-T16 model pre-trained on the ImageNet 1k dataset with "
                 "image resolution of 224x224 "
             ),
-            "params": 5524800,
+            "params": 5_524_800,
             "path": "deit",
         },
         "kaggle_handle": "kaggle://keras/deit/keras/deit_tiny_distilled_patch16_224_imagenet/1",
     },
-    "deit-small-distilled-patch16-224_imagenet": {
+    "deit_small_distilled_patch16_224_imagenet": {
         "metadata": {
             "description": (
                 "DeiT-S16 model pre-trained on the ImageNet 1k dataset with "
                 "image resolution of 224x224 "
             ),
-            "params": 21666432,
+            "params": 21_666_432,
             "path": "deit",
         },
         "kaggle_handle": "kaggle://keras/deit/keras/deit_small_distilled_patch16_224_imagenet/1",

keras_hub_nightly-0.22.0.dev202508140419/keras_hub/src/models/esm/__init__.py

@@ -0,0 +1,5 @@
+from keras_hub.src.models.esm.esm_backbone import ESMBackbone
+from keras_hub.src.models.esm.esm_presets import backbone_presets
+from keras_hub.src.utils.preset_utils import register_presets
+
+register_presets(backbone_presets, ESMBackbone)

keras_hub_nightly-0.22.0.dev202508140419/keras_hub/src/models/esm/esm_presets.py

@@ -0,0 +1,53 @@
+"""ESM model preset configurations."""
+
+# Metadata for loading pretrained model weights.
+backbone_presets = {
+    "esm2_t6_8M": {
+        "metadata": {
+            "description": (
+                "6 transformer layers version of the ESM-2 protein language "
+                "model, trained on the UniRef50 clustered protein sequence "
+                "dataset."
+            ),
+            "params": 7_408_960,
+            "path": "esm",
+        },
+        "kaggle_handle": "kaggle://keras/esm-2/keras/esm2_t6_8M/1",
+    },
+    "esm2_t12_35M": {
+        "metadata": {
+            "description": (
+                "12 transformer layers version of the ESM-2 protein language "
+                "model, trained on the UniRef50 clustered protein sequence "
+                "dataset."
+            ),
+            "params": 33_269_280,
+            "path": "esm",
+        },
+        "kaggle_handle": "kaggle://keras/esm-2/keras/esm2_t12_35M/1",
+    },
+    "esm2_t30_150M": {
+        "metadata": {
+            "description": (
+                "30 transformer layers version of the ESM-2 protein language "
+                "model, trained on the UniRef50 clustered protein sequence "
+                "dataset."
+            ),
+            "params": 147_728_000,
+            "path": "esm",
+        },
+        "kaggle_handle": "kaggle://keras/esm-2/keras/esm2_t30_150M/1",
+    },
+    "esm2_t33_650M": {
+        "metadata": {
+            "description": (
+                "33 transformer layers version of the ESM-2 protein language "
+                "model, trained on the UniRef50 clustered protein sequence "
+                "dataset."
+            ),
+            "params": 649_400_320,
+            "path": "esm",
+        },
+        "kaggle_handle": "kaggle://keras/esm-2/keras/esm2_t33_650M/1",
+    },
+}

{keras_hub_nightly-0.22.0.dev202508110431 → keras_hub_nightly-0.22.0.dev202508140419}/keras_hub/src/models/mobilenet/mobilenet_presets.py

@@ -8,7 +8,7 @@ backbone_presets = {
                 "dataset at a 224x224 resolution. Has half channel multiplier."
             ),
             "params": 278784,
-            "path": "mobilenetv3",
+            "path": "mobilenet",
         },
         "kaggle_handle": "kaggle://keras/mobilenetv3/keras/mobilenet_v3_small_050_imagenet/1",
     },
@@ -20,7 +20,7 @@ backbone_presets = {
                 "multiplier."
             ),
             "params": 939120,
-            "path": "mobilenetv3",
+            "path": "mobilenet",
         },
         "kaggle_handle": "kaggle://keras/mobilenetv3/keras/mobilenet_v3_small_100_imagenet/1",
     },
@@ -32,7 +32,7 @@ backbone_presets = {
                 "multiplier."
             ),
             "params": 2996352,
-            "path": "mobilenetv3",
+            "path": "mobilenet",
         },
         "kaggle_handle": "kaggle://keras/mobilenetv3/keras/mobilenet_v3_large_100_imagenet/1",
     },
@@ -44,7 +44,7 @@ backbone_presets = {
                 "multiplier."
             ),
             "params": 2996352,
-            "path": "mobilenetv3",
+            "path": "mobilenet",
         },
         "kaggle_handle": "kaggle://keras/mobilenetv3/keras/mobilenet_v3_large_100_imagenet_21k/1",
     },

{keras_hub_nightly-0.22.0.dev202508110431 → keras_hub_nightly-0.22.0.dev202508140419}/keras_hub/src/models/pali_gemma/pali_gemma_presets.py

@@ -63,7 +63,7 @@ backbone_presets = {
             ),
             "params": 3032979696,
             "official_name": "PaliGemma2",
-            "path": "pali_gemma2",
+            "path": "pali_gemma",
             "model_card": "https://www.kaggle.com/models/google/paligemma-2",
         },
         "kaggle_handle": "kaggle://keras/paligemma2/keras/pali_gemma_2_ft_docci_3b_448/2",
@@ -78,7 +78,7 @@ backbone_presets = {
             ),
             "params": 9663294192,
             "official_name": "PaliGemma2",
-            "path": "pali_gemma2",
+            "path": "pali_gemma",
             "model_card": "https://www.kaggle.com/models/google/paligemma-2",
         },
         "kaggle_handle": "kaggle://keras/paligemma2/keras/pali_gemma2_ft_docci_10b_448/3",
@@ -93,7 +93,7 @@ backbone_presets = {
             ),
             "params": 3032094960,
             "official_name": "PaliGemma2",
-            "path": "pali_gemma2",
+            "path": "pali_gemma",
             "model_card": "https://www.kaggle.com/models/google/paligemma-2",
         },
         "kaggle_handle": "kaggle://keras/paligemma2/keras/pali_gemma2_mix_3b_224/2",
@@ -108,7 +108,7 @@ backbone_presets = {
             ),
             "params": 3032979696,
             "official_name": "PaliGemma2",
-            "path": "pali_gemma2",
+            "path": "pali_gemma",
             "model_card": "https://www.kaggle.com/models/google/paligemma-2",
         },
         "kaggle_handle": "kaggle://keras/paligemma2/keras/pali_gemma2_mix_3b_448/2",
@@ -123,7 +123,7 @@ backbone_presets = {
             ),
             "params": 9662409456,
             "official_name": "PaliGemma2",
-            "path": "pali_gemma2",
+            "path": "pali_gemma",
             "model_card": "https://www.kaggle.com/models/google/paligemma-2",
         },
         "kaggle_handle": "kaggle://keras/paligemma2/keras/pali_gemma2_mix_10b_224/3",
@@ -138,7 +138,7 @@ backbone_presets = {
             ),
             "params": 9663294192,
             "official_name": "PaliGemma2",
-            "path": "pali_gemma2",
+            "path": "pali_gemma",
             "model_card": "https://www.kaggle.com/models/google/paligemma-2",
         },
         "kaggle_handle": "kaggle://keras/paligemma2/keras/pali_gemma2_mix_10b_448/3",
@@ -153,7 +153,7 @@ backbone_presets = {
             ),
             "params": 27650192112,
             "official_name": "PaliGemma2",
-            "path": "pali_gemma2",
+            "path": "pali_gemma",
             "model_card": "https://www.kaggle.com/models/google/paligemma-2",
         },
         "kaggle_handle": "kaggle://keras/paligemma2/keras/pali_gemma2_28b_mix_224/3",
@@ -168,7 +168,7 @@ backbone_presets = {
             ),
             "params": 27650192112,
             "official_name": "PaliGemma2",
-            "path": "pali_gemma2",
+            "path": "pali_gemma",
             "model_card": "https://www.kaggle.com/models/google/paligemma-2",
         },
         "kaggle_handle": "kaggle://keras/paligemma2/keras/pali_gemma2_28b_mix_448/3",
@@ -183,7 +183,7 @@ backbone_presets = {
             ),
             "params": 3032094960,
             "official_name": "PaliGemma2",
-            "path": "pali_gemma2",
+            "path": "pali_gemma",
             "model_card": "https://www.kaggle.com/models/google/paligemma-2",
         },
         "kaggle_handle": "kaggle://keras/paligemma2/keras/pali_gemma2_pt_3b_224/2",
@@ -198,7 +198,7 @@ backbone_presets = {
             ),
             "params": 3032979696,
             "official_name": "PaliGemma2",
-            "path": "pali_gemma2",
+            "path": "pali_gemma",
             "model_card": "https://www.kaggle.com/models/google/paligemma-2",
         },
         "kaggle_handle": "kaggle://keras/paligemma2/keras/pali_gemma2_pt_3b_448/2",
@@ -213,7 +213,7 @@ backbone_presets = {
             ),
             "params": 3036518640,
             "official_name": "PaliGemma2",
-            "path": "pali_gemma2",
+            "path": "pali_gemma",
             "model_card": "https://www.kaggle.com/models/google/paligemma-2",
         },
         "kaggle_handle": "kaggle://keras/paligemma2/keras/pali_gemma2_pt_3b_896/2",
@@ -228,7 +228,7 @@ backbone_presets = {
             ),
             "params": 9662409456,
             "official_name": "PaliGemma2",
-            "path": "pali_gemma2",
+            "path": "pali_gemma",
             "model_card": "https://www.kaggle.com/models/google/paligemma-2",
         },
         "kaggle_handle": "kaggle://keras/paligemma2/keras/pali_gemma2_pt_10b_224/3",
@@ -243,7 +243,7 @@ backbone_presets = {
             ),
             "params": 9663294192,
             "official_name": "PaliGemma2",
-            "path": "pali_gemma2",
+            "path": "pali_gemma",
             "model_card": "https://www.kaggle.com/models/google/paligemma-2",
         },
         "kaggle_handle": "kaggle://keras/paligemma2/keras/pali_gemma2_pt_10b_448/3",
@@ -258,7 +258,7 @@ backbone_presets = {
             ),
             "params": 9666833136,
             "official_name": "PaliGemma2",
-            "path": "pali_gemma2",
+            "path": "pali_gemma",
             "model_card": "https://www.kaggle.com/models/google/paligemma-2",
         },
         "kaggle_handle": "kaggle://keras/paligemma2/keras/pali_gemma2_pt_10b_896/3",
@@ -273,7 +273,7 @@ backbone_presets = {
             ),
             "params": 27650192112,
             "official_name": "PaliGemma2",
-            "path": "pali_gemma2",
+            "path": "pali_gemma",
             "model_card": "https://www.kaggle.com/models/google/paligemma-2",
         },
         "kaggle_handle": "kaggle://keras/paligemma2/keras/pali_gemma2_pt_28b_224/4",
@@ -288,7 +288,7 @@ backbone_presets = {
             ),
             "params": 27650192112,
             "official_name": "PaliGemma2",
-            "path": "pali_gemma2",
+            "path": "pali_gemma",
             "model_card": "https://www.kaggle.com/models/google/paligemma-2",
         },
         "kaggle_handle": "kaggle://keras/paligemma2/keras/pali_gemma2_pt_28b_448/3",
@@ -303,7 +303,7 @@ backbone_presets = {
             ),
             "params": 27650192112,
             "official_name": "PaliGemma2",
-            "path": "pali_gemma2",
+            "path": "pali_gemma",
             "model_card": "https://www.kaggle.com/models/google/paligemma-2",
         },
         "kaggle_handle": "kaggle://keras/paligemma2/keras/pali_gemma2_pt_28b_896/3",

{keras_hub_nightly-0.22.0.dev202508110431 → keras_hub_nightly-0.22.0.dev202508140419}/keras_hub/src/models/qwen3/qwen3_presets.py

@@ -8,7 +8,7 @@ backbone_presets = {
                 "efficiency and fast inference on resource-constrained devices."
             ),
             "params": 596049920,
-            "path": "qwen-3",
+            "path": "qwen3",
         },
         "kaggle_handle": "kaggle://keras/qwen-3/keras/qwen3_0.6b_en/1",
     },
@@ -19,7 +19,7 @@ backbone_presets = {
                 "a good balance between performance and resource usage."
             ),
             "params": 1720574976,
-            "path": "qwen-3",
+            "path": "qwen3",
         },
         "kaggle_handle": "kaggle://keras/qwen-3/keras/qwen3_1.7b_en/1",
     },
@@ -31,7 +31,7 @@ backbone_presets = {
                 "variants."
             ),
             "params": 4022468096,
-            "path": "qwen-3",
+            "path": "qwen3",
         },
         "kaggle_handle": "kaggle://keras/qwen-3/keras/qwen3_4b_en/1",
     },
@@ -43,7 +43,7 @@ backbone_presets = {
                 "capabilities."
             ),
             "params": 8190735360,
-            "path": "qwen-3",
+            "path": "qwen3",
         },
         "kaggle_handle": "kaggle://keras/qwen-3/keras/qwen3_8b_en/1",
     },
@@ -54,7 +54,7 @@ backbone_presets = {
                 "advanced reasoning, coding, and multilingual capabilities."
             ),
             "params": 14768307200,
-            "path": "qwen-3",
+            "path": "qwen3",
         },
         "kaggle_handle": "kaggle://keras/qwen-3/keras/qwen3_14b_en/1",
     },
@@ -66,7 +66,7 @@ backbone_presets = {
                 "general language tasks."
             ),
             "params": 32762123264,
-            "path": "qwen-3",
+            "path": "qwen3",
         },
         "kaggle_handle": "kaggle://keras/qwen-3/keras/qwen3_32b_en/1",
     },

{keras_hub_nightly-0.22.0.dev202508110431 → keras_hub_nightly-0.22.0.dev202508140419}/keras_hub/src/models/qwen_moe/qwen_moe_presets.py

@@ -8,7 +8,7 @@ backbone_presets = {
                 "and 8 experts per MoE layer."
             ),
             "params": 14315784192,
-            "path": "qwen-1.5-moe",
+            "path": "qwen_moe",
         },
         "kaggle_handle": "kaggle://keras/qwen-1.5-moe/Keras/qwen1.5_moe_2.7b_en/4",
     },

{keras_hub_nightly-0.22.0.dev202508110431 → keras_hub_nightly-0.22.0.dev202508140419}/keras_hub/src/models/roformer_v2/roformer_v2_attention.py

@@ -1,7 +1,6 @@
 import keras
 from keras import initializers
 from keras import ops
-from packaging import version
 
 
 class RoformerNorm(keras.layers.Layer):
@@ -180,8 +179,6 @@ class RoformerAttention(keras.layers.Layer):
         vw = ops.reshape(vw, (b, s, self.heads, self.head_size))
 
         qw, kw = self.rotary_embedding_layer([qw, kw])
-        if version.parse(keras.__version__) < version.parse("3.6"):
-            raise ("Please make sure your Keras version is >=3.6.")
         flash_attention = keras.config.is_flash_attention_enabled()
         attention_mask = ops.reshape(attention_mask, [b, 1, s, 1])
         if keras.config.backend() == "torch":

{keras_hub_nightly-0.22.0.dev202508110431 → keras_hub_nightly-0.22.0.dev202508140419}/keras_hub/src/models/stable_diffusion_3/stable_diffusion_3_backbone.py

@@ -1,4 +1,6 @@
 import keras
+from keras import backend
+from keras import distribution
 from keras import layers
 from keras import ops
 
@@ -621,11 +623,17 @@ class StableDiffusion3Backbone(Backbone):
                 config["vae"]["config"]["dtype"] = dtype_config
 
         # Text encoders default to float16 dtype if not specified.
+        # TODO: JAX CPU doesn't support float16 in `nn.dot_product_attention`.
+        is_jax_cpu = (
+            backend.backend() == "jax"
+            and "cpu" in distribution.list_devices()[0].lower()
+        )
         for text_encoder in ("clip_l", "clip_g", "t5"):
             if (
                 text_encoder in config
                 and config[text_encoder] is not None
                 and "dtype" not in config[text_encoder]["config"]
+                and not is_jax_cpu
             ):
                 config[text_encoder]["config"]["dtype"] = "float16"
 

{keras_hub_nightly-0.22.0.dev202508110431 → keras_hub_nightly-0.22.0.dev202508140419}/keras_hub/src/utils/transformers/convert_llama3.py

@@ -127,6 +127,12 @@ def convert_tokenizer(cls, preset, **kwargs):
     vocab = tokenizer_config["model"]["vocab"]
     merges = tokenizer_config["model"]["merges"]
 
+    # Handle different merge formats
+    if merges and isinstance(merges[0], list) and len(merges[0]) == 2:
+        # Convert list of lists format [["Ġ", "a"], ["Ġ", "b"]]
+        # to space-separated strings
+        merges = [" ".join(merge) for merge in merges]
+
     # Load all special tokens with the exception of "reserved" ones.
     special_tokens = set()
     for token in tokenizer_config["added_tokens"]:

{keras_hub_nightly-0.22.0.dev202508110431 → keras_hub_nightly-0.22.0.dev202508140419}/keras_hub/src/version.py

@@ -1,7 +1,7 @@
 from keras_hub.src.api_export import keras_hub_export
 
 # Unique source of truth for the version number.
-__version__ = "0.22.0.dev202508110431"
+__version__ = "0.22.0.dev202508140419"
 
 
 @keras_hub_export("keras_hub.version")

{keras_hub_nightly-0.22.0.dev202508110431 → keras_hub_nightly-0.22.0.dev202508140419}/keras_hub_nightly.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: keras-hub-nightly
-Version: 0.22.0.dev202508110431
+Version: 0.22.0.dev202508140419
 Summary: Pretrained models for Keras.
 Author-email: Keras team <keras-users@googlegroups.com>
 License-Expression: Apache-2.0
@@ -18,9 +18,9 @@ Classifier: Operating System :: MacOS
 Classifier: Intended Audience :: Science/Research
 Classifier: Topic :: Scientific/Engineering
 Classifier: Topic :: Software Development
-Requires-Python: >=3.9
+Requires-Python: >=3.10
 Description-Content-Type: text/markdown
-Requires-Dist: keras>=3.5
+Requires-Dist: keras>=3.8
 Requires-Dist: absl-py
 Requires-Dist: numpy
 Requires-Dist: packaging
@@ -31,7 +31,7 @@ Requires-Dist: tensorflow-text; platform_system != "Windows"
 
 # KerasHub: Multi-framework Pretrained Models
 [![](https://github.com/keras-team/keras-hub/workflows/Tests/badge.svg?branch=master)](https://github.com/keras-team/keras-hub/actions?query=workflow%3ATests+branch%3Amaster)
-![Python](https://img.shields.io/badge/python-v3.9.0+-success.svg)
+![Python](https://img.shields.io/badge/python-v3.10.0+-success.svg)
 [![contributions welcome](https://img.shields.io/badge/contributions-welcome-brightgreen.svg?style=flat)](https://github.com/keras-team/keras-hub/issues)
 
 > [!IMPORTANT]

{keras_hub_nightly-0.22.0.dev202508110431 → keras_hub_nightly-0.22.0.dev202508140419}/keras_hub_nightly.egg-info/requires.txt

@@ -1,4 +1,4 @@
-keras>=3.5
+keras>=3.8
 absl-py
 numpy
 packaging

{keras_hub_nightly-0.22.0.dev202508110431 → keras_hub_nightly-0.22.0.dev202508140419}/pyproject.toml

@@ -9,7 +9,7 @@ authors = [
 ]
 description = "Pretrained models for Keras."
 readme = "README.md"
-requires-python = ">=3.9"
+requires-python = ">=3.10"
 license = "Apache-2.0"
 dynamic = ["version"]
 classifiers = [
@@ -27,7 +27,7 @@ classifiers = [
     "Topic :: Software Development",
 ]
 dependencies = [
-    "keras>=3.5",
+    "keras>=3.8",
     "absl-py",
     "numpy",
     "packaging",