Upload Qwen2VLForConditionalGeneration

config.json

@@ -15,7 +15,7 @@
   "model_type": "qwen2_vl",
   "num_attention_heads": 1,
   "num_hidden_layers": 1,
-  "num_key_value_heads": 4,
+  "num_key_value_heads": 2,
   "rms_norm_eps": 1e-06,
   "rope_scaling": {
     "mrope_section": [
@@ -39,7 +39,7 @@
     "hidden_size": 64,
     "image_token_id": null,
     "initializer_range": 0.02,
-    "intermediate_size": 18944,
+    "intermediate_size": 8960,
     "layer_types": [
       "full_attention",
       "full_attention",
@@ -75,7 +75,7 @@
     "model_type": "qwen2_vl_text",
     "num_attention_heads": 1,
     "num_hidden_layers": 1,
-    "num_key_value_heads": 4,
+    "num_key_value_heads": 2,
     "rms_norm_eps": 1e-06,
     "rope_scaling": {
       "mrope_section": [
@@ -88,16 +88,16 @@
     },
     "rope_theta": 1000000.0,
     "sliding_window": null,
-    "torch_dtype": "bfloat16",
+    "tie_word_embeddings": true,
+    "torch_dtype": "float32",
     "use_cache": true,
     "use_sliding_window": false,
     "video_token_id": null,
     "vision_end_token_id": 151653,
     "vision_start_token_id": 151652,
     "vision_token_id": 151654,
-    "vocab_size": 152064
+    "vocab_size": 151936
   },
-  "tie_word_embeddings": false,
   "torch_dtype": "float32",
   "transformers_version": "4.53.0",
   "use_cache": true,
@@ -117,10 +117,11 @@
     "patch_size": 14,
     "spatial_merge_size": 2,
     "spatial_patch_size": 14,
-    "temporal_patch_size": 2
+    "temporal_patch_size": 2,
+    "torch_dtype": "float32"
   },
   "vision_end_token_id": 151653,
   "vision_start_token_id": 151652,
   "vision_token_id": 151654,
-  "vocab_size": 152064
+  "vocab_size": 151936
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b65dbbd6044b690ae2ad4a163f3706d4880c6096ffbdaebe9c59b5ee46947a1f
-size 66584432
+oid sha256:60669a8303b1d49faa61164d7f47c065ca1514d8f98680cd3d9cd4123e84dd1a
+size 46711048