Re-convert with vision tower (fix blank-image loading)

kf_config.json

@@ -4,7 +4,7 @@
   "model_module": "kerasformers.models.gemma3n",
   "model_class": "Gemma3nForConditionalGeneration",
   "variant": "gemma-3n-e2b",
-  "weights": "model.weights.json",
+  "weights": "model.weights.h5",
   "schema_version": 2,
   "weight_dtype": "bfloat16",
   "model_type": "gemma3n",
@@ -128,6 +128,14 @@
       0.0
     ]
   },
+  "vision_config": {
+    "architecture": "mobilenetv5_300m_enc",
+    "hidden_size": 2048,
+    "vocab_size": 128,
+    "vocab_offset": 262144,
+    "rms_norm_eps": 1e-06,
+    "do_pooling": false
+  },
   "audio_config": {
     "vocab_size": 128,
     "vocab_offset": 262272,

model.weights.json

@@ -1351,184 +1351,184 @@
         "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_2/norm/vars": "model_00001.weights.h5",
         "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_3/layer_scale/vars": "model_00001.weights.h5",
         "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_3/pw_exp/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_3/pw_exp/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_3/pw_proj/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_3/pw_proj/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_3/attn/key_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_3/attn/output_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_3/attn/query_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_3/attn/value_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_3/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_3/norm/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_4/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_4/pw_exp/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_4/pw_exp/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_4/pw_proj/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_4/pw_proj/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_4/attn/key_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_4/attn/output_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_4/attn/query_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_4/attn/value_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_4/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_4/norm/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_5/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_5/pw_exp/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_5/pw_exp/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_5/pw_proj/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_5/pw_proj/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_5/attn/key_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_5/attn/output_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_5/attn/query_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_5/attn/value_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_5/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_5/norm/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_6/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_6/pw_exp/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_6/pw_exp/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_6/pw_proj/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_6/pw_proj/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_6/attn/key_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_6/attn/output_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_6/attn/query_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_6/attn/value_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_6/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_6/norm/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_7/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_7/pw_exp/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_7/pw_exp/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_7/pw_proj/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_7/pw_proj/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_7/attn/key_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_7/attn/output_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_7/attn/query_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_7/attn/value_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_7/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_7/norm/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_8/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_8/pw_exp/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_8/pw_exp/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_8/pw_proj/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_8/pw_proj/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_8/attn/key_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_8/attn/output_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_8/attn/query_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_8/attn/value_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_8/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_8/norm/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_9/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_9/pw_exp/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_9/pw_exp/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_9/pw_proj/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_9/pw_proj/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_9/attn/key_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_9/attn/output_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_9/attn/query_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_9/attn/value_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_9/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_9/norm/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_10/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_10/pw_exp/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_10/pw_exp/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_10/pw_proj/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_10/pw_proj/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_10/attn/key_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_10/attn/output_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_10/attn/query_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_10/attn/value_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_10/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_10/norm/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_11/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_11/pw_exp/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_11/pw_exp/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_11/pw_proj/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_11/pw_proj/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_11/attn/key_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_11/attn/output_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_11/attn/query_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_11/attn/value_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_11/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_11/norm/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_12/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_12/pw_exp/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_12/pw_exp/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_12/pw_proj/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_12/pw_proj/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_12/attn/key_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_12/attn/output_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_12/attn/query_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_12/attn/value_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_12/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_12/norm/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_13/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_13/pw_exp/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_13/pw_exp/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_13/pw_proj/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_13/pw_proj/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_13/attn/key_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_13/attn/output_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_13/attn/query_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_13/attn/value_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_13/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_13/norm/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_14/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_14/pw_exp/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_14/pw_exp/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_14/pw_proj/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_14/pw_proj/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_14/attn/key_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_14/attn/output_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_14/attn/query_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_14/attn/value_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_14/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_14/norm/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_15/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_15/pw_exp/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_15/pw_exp/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_15/pw_proj/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_15/pw_proj/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_15/attn/key_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_15/attn/output_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_15/attn/query_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_15/attn/value_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_15/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_15/norm/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_16/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_16/pw_exp/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_16/pw_exp/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_16/pw_proj/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_16/pw_proj/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_16/attn/key_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_16/attn/output_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_16/attn/query_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_16/attn/value_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_16/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_16/norm/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_17/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_17/pw_exp/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_17/pw_exp/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_17/pw_proj/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_17/pw_proj/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_17/attn/key_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_17/attn/output_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_17/attn/query_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_17/attn/value_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_17/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_17/norm/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_18/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_18/pw_exp/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_18/pw_exp/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_18/pw_proj/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_18/pw_proj/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_18/attn/key_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_18/attn/output_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_18/attn/query_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_18/attn/value_proj/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_18/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_18/norm/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_19/layer_scale/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_19/pw_exp/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_19/pw_exp/conv/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_19/pw_proj/bn/vars": "model_00001.weights.h5",
-        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_19/pw_proj/conv/vars": "model_00001.weights.h5"
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_3/pw_exp/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_3/pw_proj/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_3/pw_proj/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_3/attn/key_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_3/attn/output_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_3/attn/query_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_3/attn/value_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_3/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_3/norm/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_4/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_4/pw_exp/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_4/pw_exp/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_4/pw_proj/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_4/pw_proj/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_4/attn/key_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_4/attn/output_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_4/attn/query_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_4/attn/value_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_4/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_4/norm/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_5/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_5/pw_exp/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_5/pw_exp/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_5/pw_proj/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_5/pw_proj/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_5/attn/key_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_5/attn/output_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_5/attn/query_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_5/attn/value_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_5/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_5/norm/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_6/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_6/pw_exp/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_6/pw_exp/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_6/pw_proj/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_6/pw_proj/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_6/attn/key_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_6/attn/output_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_6/attn/query_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_6/attn/value_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_6/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_6/norm/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_7/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_7/pw_exp/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_7/pw_exp/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_7/pw_proj/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_7/pw_proj/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_7/attn/key_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_7/attn/output_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_7/attn/query_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_7/attn/value_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_7/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_7/norm/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_8/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_8/pw_exp/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_8/pw_exp/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_8/pw_proj/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_8/pw_proj/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_8/attn/key_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_8/attn/output_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_8/attn/query_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_8/attn/value_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_8/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_8/norm/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_9/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_9/pw_exp/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_9/pw_exp/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_9/pw_proj/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_9/pw_proj/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_9/attn/key_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_9/attn/output_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_9/attn/query_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_9/attn/value_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_9/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_9/norm/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_10/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_10/pw_exp/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_10/pw_exp/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_10/pw_proj/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_10/pw_proj/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_10/attn/key_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_10/attn/output_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_10/attn/query_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_10/attn/value_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_10/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_10/norm/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_11/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_11/pw_exp/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_11/pw_exp/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_11/pw_proj/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_11/pw_proj/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_11/attn/key_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_11/attn/output_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_11/attn/query_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_11/attn/value_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_11/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_11/norm/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_12/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_12/pw_exp/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_12/pw_exp/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_12/pw_proj/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_12/pw_proj/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_12/attn/key_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_12/attn/output_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_12/attn/query_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_12/attn/value_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_12/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_12/norm/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_13/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_13/pw_exp/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_13/pw_exp/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_13/pw_proj/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_13/pw_proj/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_13/attn/key_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_13/attn/output_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_13/attn/query_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_13/attn/value_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_13/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_13/norm/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_14/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_14/pw_exp/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_14/pw_exp/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_14/pw_proj/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_14/pw_proj/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_14/attn/key_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_14/attn/output_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_14/attn/query_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_14/attn/value_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_14/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_14/norm/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_15/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_15/pw_exp/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_15/pw_exp/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_15/pw_proj/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_15/pw_proj/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_15/attn/key_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_15/attn/output_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_15/attn/query_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_15/attn/value_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_15/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_15/norm/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_16/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_16/pw_exp/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_16/pw_exp/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_16/pw_proj/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_16/pw_proj/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_16/attn/key_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_16/attn/output_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_16/attn/query_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_16/attn/value_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_16/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_16/norm/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_17/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_17/pw_exp/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_17/pw_exp/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_17/pw_proj/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_17/pw_proj/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_17/attn/key_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_17/attn/output_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_17/attn/query_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_17/attn/value_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_17/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_17/norm/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_18/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_18/pw_exp/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_18/pw_exp/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_18/pw_proj/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_18/pw_proj/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_18/attn/key_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_18/attn/output_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_18/attn/query_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_18/attn/value_proj/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_18/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/mobile_attention_18/norm/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_19/layer_scale/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_19/pw_exp/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_19/pw_exp/conv/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_19/pw_proj/bn/vars": "model_00002.weights.h5",
+        "/layers/mobile_net_v5_encoder/stages/container_3/universal_inverted_residual_19/pw_proj/conv/vars": "model_00002.weights.h5"
     }
 }

model_00001.weights.h5

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4a1b8aecd9d2d463a56d3a3303d9efed8c2dcd63b9c6072278616262dc941cdf
-size 5722960688
+oid sha256:83f1155d43c463b3371432a0df46211f72d3049bb7743442a0a0cf9f634c4537
+size 5365597336

model_00002.weights.h5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f8fee9647037a6a37db6239418cc4ccca1661d2bc834a56aef5d2c6f343ca262
+size 357371928