Update config.json (#2)

- Update config.json (8dc2e95f36594c8e2fa45a6b0a54e7b019530d38)


Co-authored-by: Xuchen Pan <panxuchen@users.noreply.huggingface.co>

config.json

@@ -18,9 +18,9 @@
     "full_attention_interval": 4,
     "head_dim": 256,
     "hidden_act": "silu",
-    "hidden_size": 4096,
+    "hidden_size": 2048,
     "initializer_range": 0.02,
-    "intermediate_size": 12288,
+    "intermediate_size": 6144,
     "layer_types": [
       "linear_attention",
       "linear_attention",
@@ -45,20 +45,12 @@
       "linear_attention",
       "linear_attention",
       "linear_attention",
-      "full_attention",
-      "linear_attention",
-      "linear_attention",
-      "linear_attention",
-      "full_attention",
-      "linear_attention",
-      "linear_attention",
-      "linear_attention",
       "full_attention"
     ],
     "linear_conv_kernel_dim": 4,
     "linear_key_head_dim": 128,
     "linear_num_key_heads": 16,
-    "linear_num_value_heads": 32,
+    "linear_num_value_heads": 16,
     "linear_value_head_dim": 128,
     "mamba_ssm_dtype": "float32",
     "max_position_embeddings": 262144,
@@ -66,9 +58,9 @@
     "model_type": "qwen3_5_text",
     "mtp_num_hidden_layers": 1,
     "mtp_use_dedicated_embeddings": false,
-    "num_attention_heads": 16,
-    "num_hidden_layers": 32,
-    "num_key_value_heads": 4,
+    "num_attention_heads": 8,
+    "num_hidden_layers": 24,
+    "num_key_value_heads": 2,
     "pad_token_id": null,
     "partial_rotary_factor": 0.25,
     "rms_norm_eps": 1e-06,
@@ -83,26 +75,26 @@
       "rope_theta": 10000000,
       "rope_type": "default"
     },
-    "tie_word_embeddings": false,
+    "tie_word_embeddings": true,
     "use_cache": true,
     "vocab_size": 248320
   },
-  "tie_word_embeddings": false,
+  "tie_word_embeddings": true,
   "transformers_version": "5.3.0",
   "video_token_id": 248057,
   "vision_config": {
     "deepstack_visual_indexes": [],
-    "depth": 27,
+    "depth": 24,
     "dtype": "float32",
     "hidden_act": "gelu_pytorch_tanh",
-    "hidden_size": 1152,
+    "hidden_size": 1024,
     "in_channels": 3,
     "initializer_range": 0.02,
-    "intermediate_size": 4304,
+    "intermediate_size": 4096,
     "model_type": "qwen3_5",
     "num_heads": 16,
     "num_position_embeddings": 2304,
-    "out_hidden_size": 4096,
+    "out_hidden_size": 2048,
     "patch_size": 16,
     "spatial_merge_size": 2,
     "temporal_patch_size": 2