Update model

config.json

@@ -9,7 +9,6 @@
   "d_latents": 1024,
   "d_model": 261,
   "hidden_act": "gelu",
-  "hidden_dropout_prob": 0.1,
   "id2label": {
     "0": "tench, Tinca tinca",
     "1": "goldfish, Carassius auratus",
@@ -2016,6 +2015,7 @@
     "zucchini, courgette": 939
   },
   "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 2048,
   "model_type": "perceiver",
   "num_blocks": 8,
   "num_cross_attention_heads": 1,
@@ -2023,17 +2023,21 @@
   "num_latents": 512,
   "num_self_attends_per_block": 6,
   "num_self_attention_heads": 8,
-  "position_embedding_init_scale": 0.02,
+  "output_shape": [
+    1,
+    16,
+    224,
+    224
+  ],
   "qk_channels": null,
   "samples_per_patch": 16,
   "self_attention_widening_factor": 1,
-  "seq_len": 2048,
   "torch_dtype": "float32",
   "train_size": [
     368,
     496
   ],
-  "transformers_version": "4.11.0.dev0",
+  "transformers_version": "4.12.0.dev0",
   "use_query_residual": true,
   "v_channels": null,
   "vocab_size": 262