Update config.json

config.json

@@ -67,7 +67,6 @@
     "top_p": 1.0,
     "torch_dtype": null,
     "torchscript": false,
-    "transformers_version": "4.16.0.dev0",
     "use_bfloat16": false,
     "vocab_size": 256,
     "use_attention_bias": false,
@@ -82,7 +81,7 @@
   },
   "text_config_dict": null,
   "transformers_version": null,
-  "vision_config": {
+  "speech_config": {
     "use_attention_bias": false,
     "_name_or_path": "",
     "add_cross_attention": false,
@@ -104,7 +103,7 @@
     "forced_eos_token_id": null,
     "hidden_act": "gelu",
     "hidden_size": 768,
-    "image_size": 224,
+    "vocab_size": 8192,
     "initializer_factor": 1.0,
     "initializer_range": 0.02,
     "intermediate_size": 1536,
@@ -114,7 +113,7 @@
     "length_penalty": 1.0,
     "max_length": 20,
     "min_length": 0,
-    "model_type": "clip_vision_model",
+    "model_type": "clip_speech_model",
     "no_repeat_ngram_size": 0,
     "num_attention_heads": 12,
     "num_beam_groups": 1,
@@ -154,5 +153,5 @@
     "use_rotary_embedding": true,
     "summary_type": "mean"
   },
-  "vision_config_dict": null
+  "speech_config_dict": null,
 }