Upload Pix2StructForConditionalGeneration

config.json

@@ -7,6 +7,7 @@
   "decoder_start_token_id": 0,
   "initializer_factor": 1.0,
   "initializer_range": 0.02,
+  "is_vqa": false,
   "model_type": "pix2struct",
   "pad_token_id": 0,
   "text_config": {
@@ -82,7 +83,7 @@
     "top_p": 1.0,
     "torch_dtype": null,
     "torchscript": false,
-    "transformers_version": "4.27.0.dev0",
+    "transformers_version": "4.28.0.dev0",
     "typical_p": 1.0,
     "use_bfloat16": false,
     "use_cache": false,
@@ -114,13 +115,11 @@
     "finetuning_task": null,
     "forced_bos_token_id": null,
     "forced_eos_token_id": null,
-    "hidden_dropout_prob": 0.0,
     "hidden_size": 1536,
     "id2label": {
       "0": "LABEL_0",
       "1": "LABEL_1"
     },
-    "image_size": 384,
     "initializer_factor": 1.0,
     "initializer_range": 0.02,
     "is_decoder": false,
@@ -134,7 +133,6 @@
     "length_penalty": 1.0,
     "max_length": 20,
     "min_length": 0,
-    "mlp_bias": false,
     "model_type": "pix2struct_vision_model",
     "no_repeat_ngram_size": 0,
     "num_attention_heads": 24,
@@ -153,7 +151,6 @@
     "problem_type": null,
     "projection_dim": 768,
     "pruned_heads": {},
-    "qkv_bias": false,
     "relative_attention_max_distance": 128,
     "relative_attention_num_buckets": 32,
     "remove_invalid_values": false,
@@ -173,7 +170,7 @@
     "top_p": 1.0,
     "torch_dtype": null,
     "torchscript": false,
-    "transformers_version": "4.27.0.dev0",
+    "transformers_version": "4.28.0.dev0",
     "typical_p": 1.0,
     "use_bfloat16": false
   }