Update tiny models for Gemma3nModel

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

config.json

@@ -0,0 +1,160 @@
+{
+  "architectures": [
+    "Gemma3nModel"
+  ],
+  "audio_config": {
+    "conf_attention_chunk_size": 4,
+    "conf_attention_context_left": 5,
+    "conf_attention_context_right": 0,
+    "conf_attention_logit_cap": 50.0,
+    "conf_conv_kernel_size": 3,
+    "conf_num_attention_heads": 4,
+    "conf_num_hidden_layers": 2,
+    "conf_reduction_factor": 4,
+    "conf_residual_weight": 0.5,
+    "gradient_clipping": 10000000000.0,
+    "hidden_size": 32,
+    "input_feat_size": 32,
+    "model_type": "gemma3n_audio",
+    "rms_norm_eps": 1e-06,
+    "sscp_conv_channel_size": [
+      16,
+      8
+    ],
+    "sscp_conv_group_norm_eps": 0.001,
+    "sscp_conv_kernel_size": [
+      [
+        3,
+        3
+      ],
+      [
+        3,
+        3
+      ]
+    ],
+    "sscp_conv_stride_size": [
+      [
+        2,
+        2
+      ],
+      [
+        2,
+        2
+      ]
+    ],
+    "vocab_offset": 262272,
+    "vocab_size": 128
+  },
+  "audio_soft_tokens_per_image": 4,
+  "audio_token_id": 8,
+  "boa_token_id": 6,
+  "boi_token_id": 4,
+  "dtype": "float32",
+  "eoa_token_id": 7,
+  "eoi_token_id": 5,
+  "image_token_id": 3,
+  "initializer_range": 0.02,
+  "mm_tokens_per_image": 2,
+  "model_type": "gemma3n",
+  "text_config": {
+    "activation_sparsity_pattern": [
+      0.0,
+      0.0,
+      0.0,
+      0.0
+    ],
+    "altup_active_idx": 0,
+    "altup_coef_clip": 120.0,
+    "altup_correct_scale": true,
+    "altup_num_inputs": 2,
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "bos_token_id": 2,
+    "eos_token_id": 1,
+    "final_logit_softcapping": 30.0,
+    "head_dim": 8,
+    "hidden_activation": "gelu_pytorch_tanh",
+    "hidden_size": 16,
+    "hidden_size_per_layer_input": 16,
+    "initializer_range": 0.02,
+    "intermediate_size": [
+      22,
+      22,
+      22,
+      22
+    ],
+    "laurel_rank": 64,
+    "layer_types": [
+      "full_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention"
+    ],
+    "max_position_embeddings": 512,
+    "model_type": "gemma3n_text",
+    "num_attention_heads": 2,
+    "num_hidden_layers": 4,
+    "num_key_value_heads": 2,
+    "num_kv_shared_layers": 2,
+    "pad_token_id": 0,
+    "rms_norm_eps": 1e-06,
+    "rope_parameters": {
+      "full_attention": {
+        "rope_theta": 1000000.0,
+        "rope_type": "default"
+      },
+      "sliding_attention": {
+        "rope_theta": 10000.0,
+        "rope_type": "default"
+      }
+    },
+    "sliding_window": 512,
+    "tie_word_embeddings": true,
+    "use_cache": true,
+    "vocab_size": 262400,
+    "vocab_size_per_layer_input": 99
+  },
+  "tie_word_embeddings": true,
+  "transformers_version": "5.16.0.dev0",
+  "use_cache": true,
+  "vision_config": {
+    "architecture": "vit_pe_core_large_patch14_336",
+    "do_pooling": false,
+    "hidden_size": 32,
+    "image_size": 20,
+    "initializer_range": 0.02,
+    "intermediate_size": 37,
+    "is_training": true,
+    "label_names": [
+      "LABEL_0",
+      "LABEL_1"
+    ],
+    "model_args": {
+      "depth": 2,
+      "embed_dim": 64,
+      "global_pool": "",
+      "img_size": [
+        20,
+        20
+      ],
+      "init_values": 0.1,
+      "ref_feat_shape": [
+        1,
+        1
+      ],
+      "use_post_transformer_norm": false
+    },
+    "model_type": "gemma3n_vision",
+    "num_attention_heads": 4,
+    "num_channels": 3,
+    "num_classes": 2,
+    "num_hidden_layers": 2,
+    "num_key_value_heads": 1,
+    "patch_size": 5,
+    "rms_norm_eps": 1e-06,
+    "use_labels": true,
+    "vocab_offset": 262144,
+    "vocab_size": 128
+  },
+  "vision_soft_tokens_per_image": 4
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:40a8d0b42594eb912866ed4ba5dcff21fa61da3a98e3afaed8fbb2aed1b3dbb1
+size 17723392

preprocessor_config.json

@@ -0,0 +1,29 @@
+{
+  "crop_size": {
+    "height": 20,
+    "width": 20
+  },
+  "dither": 0.0,
+  "feature_extractor_type": "Gemma3nAudioFeatureExtractor",
+  "feature_size": 128,
+  "fft_length": 1024,
+  "fft_overdrive": true,
+  "frame_length": 512,
+  "hop_length": 160,
+  "input_scale_factor": 1.0,
+  "max_frequency": 7600.0,
+  "mel_floor": 1e-05,
+  "min_frequency": 125.0,
+  "padding_side": "right",
+  "padding_value": 0.0,
+  "per_bin_mean": null,
+  "per_bin_stddev": null,
+  "preemphasis": 0.97,
+  "preemphasis_htk_flavor": true,
+  "return_attention_mask": true,
+  "sampling_rate": 16000,
+  "size": {
+    "height": 20,
+    "width": 20
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6aa40b46d55d1fa5f1c0b218de3a695942c620858ba34aa94e2aa705efc6f071
+size 33442552

tokenizer_config.json

@@ -0,0 +1,31 @@
+{
+  "audio_token": "<audio_soft_token>",
+  "backend": "tokenizers",
+  "boa_token": "<start_of_audio>",
+  "boi_token": "<start_of_image>",
+  "bos_token": "<bos>",
+  "clean_up_tokenization_spaces": false,
+  "eoa_token": "<end_of_audio>",
+  "eoi_token": "<end_of_image>",
+  "eos_token": "<eos>",
+  "image_token": "<image_soft_token>",
+  "is_local": true,
+  "local_files_only": false,
+  "mask_token": "<mask>",
+  "model_max_length": 1000000000000000019884624838656,
+  "model_specific_special_tokens": {
+    "audio_token": "<audio_soft_token>",
+    "boa_token": "<start_of_audio>",
+    "boi_token": "<start_of_image>",
+    "eoa_token": "<end_of_audio>",
+    "eoi_token": "<end_of_image>",
+    "image_token": "<image_soft_token>"
+  },
+  "pad_token": "<pad>",
+  "processor_class": "Gemma3nProcessor",
+  "sp_model_kwargs": null,
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "GemmaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}