feat: create config files

tools/train/config/medium/config.json

@@ -0,0 +1,33 @@
+{
+  "activation_dropout": 0.0,
+  "activation_function": "gelu",
+  "attention_dropout": 0.0,
+  "bos_token_id": 16385,
+  "classifier_dropout": 0.0,
+  "d_model": 1536,
+  "decoder_attention_heads": 16,
+  "decoder_ffn_dim": 4096,
+  "decoder_layerdrop": 0.0,
+  "decoder_layers": 18,
+  "decoder_start_token_id": 16384,
+  "dropout": 0.1,
+  "encoder_attention_heads": 16,
+  "encoder_ffn_dim": 4096,
+  "encoder_layerdrop": 0.0,
+  "encoder_layers": 18,
+  "encoder_vocab_size": 50264,
+  "eos_token_id": 16385,
+  "gradient_checkpointing": false,
+  "image_length": 256,
+  "image_vocab_size": 16384,
+  "init_std": 0.01,
+  "is_encoder_decoder": true,
+  "max_text_length": 64,
+  "model_type": "dallebart",
+  "normalize_text": true,
+  "pad_token_id": 16385,
+  "scale_embedding": false,
+  "tie_word_embeddings": false,
+  "transformers_version": "4.13.0.dev0",
+  "use_cache": true
+}

tools/train/config/mega/config.json

@@ -8,19 +8,19 @@
   "decoder_attention_heads": 16,
   "decoder_ffn_dim": 4096,
   "decoder_layerdrop": 0.0,
-  "decoder_layers": 20,
+  "decoder_layers": 31,
   "decoder_start_token_id": 16384,
   "dropout": 0.1,
   "encoder_attention_heads": 16,
   "encoder_ffn_dim": 4096,
   "encoder_layerdrop": 0.0,
-  "encoder_layers": 20,
+  "encoder_layers": 31,
   "encoder_vocab_size": 50264,
   "eos_token_id": 16385,
   "gradient_checkpointing": false,
   "image_length": 256,
   "image_vocab_size": 16384,
-  "init_std": 0.02,
+  "init_std": 0.01,
   "is_encoder_decoder": true,
   "max_text_length": 64,
   "model_type": "dallebart",

tools/train/config/mini/config.json

@@ -0,0 +1,33 @@
+{
+  "activation_dropout": 0.0,
+  "activation_function": "gelu",
+  "attention_dropout": 0.0,
+  "bos_token_id": 16385,
+  "classifier_dropout": 0.0,
+  "d_model": 1024,
+  "decoder_attention_heads": 16,
+  "decoder_ffn_dim": 4096,
+  "decoder_layerdrop": 0.0,
+  "decoder_layers": 12,
+  "decoder_start_token_id": 16384,
+  "dropout": 0.1,
+  "encoder_attention_heads": 16,
+  "encoder_ffn_dim": 4096,
+  "encoder_layerdrop": 0.0,
+  "encoder_layers": 12,
+  "encoder_vocab_size": 50264,
+  "eos_token_id": 16385,
+  "gradient_checkpointing": false,
+  "image_length": 256,
+  "image_vocab_size": 16384,
+  "init_std": 0.02,
+  "is_encoder_decoder": true,
+  "max_text_length": 64,
+  "model_type": "dallebart",
+  "normalize_text": true,
+  "pad_token_id": 16385,
+  "scale_embedding": false,
+  "tie_word_embeddings": false,
+  "transformers_version": "4.13.0.dev0",
+  "use_cache": true
+}