Upload config.json with huggingface_hub

config.json

@@ -1,24 +1,30 @@
 {
+  "name_or_path": "VarunGumma/rotary-indictrans2-en-indic-1B",
   "activation_dropout": 0.0,
   "activation_function": "gelu",
+  "auto_map": {
+    "AutoConfig": "configuration_rotary_indictrans.RotaryIndicTransConfig",
+    "AutoModelForSeq2SeqLM": "modeling_rotary_indictrans.RotaryIndicTransForConditionalGeneration"
+  },
+  "tokenizer_class": "IndicTransTokenizer",
   "architectures": [
     "RotaryIndicTransForConditionalGeneration"
   ],
   "attention_dropout": 0.0,
   "attn_implementation": "eager",
   "bos_token_id": 0,
-  "decoder_attention_heads": 8,
-  "decoder_embed_dim": 512,
-  "decoder_ffn_dim": 2048,
+  "decoder_attention_heads": 16,
+  "decoder_embed_dim": 1024,
+  "decoder_ffn_dim": 8192,
   "decoder_layerdrop": 0,
   "decoder_layers": 18,
   "decoder_normalize_before": true,
   "decoder_start_token_id": 2,
   "decoder_vocab_size": 122672,
   "dropout": 0.2,
-  "encoder_attention_heads": 8,
-  "encoder_embed_dim": 512,
-  "encoder_ffn_dim": 2048,
+  "encoder_attention_heads": 16,
+  "encoder_embed_dim": 1024,
+  "encoder_ffn_dim": 8192,
   "encoder_layerdrop": 0,
   "encoder_layers": 18,
   "encoder_normalize_before": true,
@@ -26,7 +32,7 @@
   "eos_token_id": 2,
   "init_std": 0.02,
   "is_encoder_decoder": true,
-  "layernorm_embedding": true,
+  "layernorm_embedding": false,
   "model_type": "RotaryIndicTrans",
   "num_hidden_layers": 18,
   "pad_token_id": 1,
@@ -34,7 +40,7 @@
     "theta": 10000
   },
   "scale_embedding": true,
-  "share_decoder_input_output_embed": true,
+  "share_decoder_input_output_embed": false,
   "torch_dtype": "float32",
   "transformers_version": "4.44.0",
   "use_cache": true