Training in progress, step 80000

config.json

@@ -6,10 +6,10 @@
   "d_kv": 64,
   "d_model": 768,
   "decoder_input_size": 1024,
-  "decoder_start_token_id": 1025,
+  "decoder_start_token_id": 0,
   "dense_act_fn": "relu",
   "dropout_rate": 0.2,
-  "eos_token_id": 1026,
+  "eos_token_id": 1,
   "feed_forward_proj": "relu",
   "initializer_factor": 1.0,
   "is_encoder_decoder": true,
@@ -22,11 +22,11 @@
   "num_decoder_layers": 10,
   "num_heads": 12,
   "num_layers": 10,
-  "pad_token_id": 1025,
+  "pad_token_id": 0,
   "relative_attention_max_distance": 128,
   "relative_attention_num_buckets": 32,
   "torch_dtype": "float32",
   "transformers_version": "4.26.0",
   "use_cache": true,
-  "vocab_size": 1026
+  "vocab_size": 1024
 }

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:66ee84e08cab9222909d92733cea65cf45c066b5281a6e68b6374775d431f8df
-size 475248041
+oid sha256:c19fa466e48d2881d4327ea2a8f3eba3d33144ba815bf1d8a05a1ac431957289
+size 475241897