Upload Gemma3ForConditionalGeneration

config.json

@@ -6,7 +6,7 @@
   "eoi_token_index": 256000,
   "image_token_index": 262144,
   "initializer_range": 0.02,
-  "mm_tokens_per_image": 49,
+  "mm_tokens_per_image": 64,
   "model_type": "gemma3",
   "text_config": {
     "attention_bias": false,
@@ -16,13 +16,13 @@
     "final_logit_softcapping": null,
     "head_dim": 256,
     "hidden_activation": "gelu_pytorch_tanh",
-    "hidden_size": 256,
+    "hidden_size": 128,
     "initializer_range": 0.02,
     "intermediate_size": 512,
     "max_position_embeddings": 131072,
     "model_type": "gemma3_text",
     "num_attention_heads": 4,
-    "num_hidden_layers": 4,
+    "num_hidden_layers": 8,
     "num_key_value_heads": 2,
     "query_pre_attn_scalar": 256,
     "rms_norm_eps": 1e-06,
@@ -39,14 +39,14 @@
   "vision_config": {
     "attention_dropout": 0.1,
     "hidden_act": "gelu_pytorch_tanh",
-    "hidden_size": 256,
-    "image_size": 224,
+    "hidden_size": 128,
+    "image_size": 32,
     "intermediate_size": 512,
     "layer_norm_eps": 1e-06,
     "model_type": "siglip_vision_model",
     "num_attention_heads": 4,
     "num_channels": 3,
     "num_hidden_layers": 4,
-    "patch_size": 16
+    "patch_size": 2
   }
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:175717feeb80f912aade8b1b4d6dbb4537e9199c0e272fb7764b7e35e1169534
-size 299212928
+oid sha256:53c279eae6c9dab986b2b03645a0cc9d6ce0bd54af0552d913e805a2bfc12946
+size 157351984