Training in progress, step 4000

config.json

@@ -1,12 +1,12 @@
 {
-  "_name_or_path": "google/mt5-base",
+  "_name_or_path": "google/mt5-small",
   "architectures": [
     "MT5ForConditionalGeneration"
   ],
   "classifier_dropout": 0.0,
-  "d_ff": 2048,
+  "d_ff": 1024,
   "d_kv": 64,
-  "d_model": 768,
+  "d_model": 512,
   "decoder_start_token_id": 0,
   "dense_act_fn": "gelu_new",
   "dropout_rate": 0.1,
@@ -17,10 +17,9 @@
   "is_gated_act": true,
   "layer_norm_epsilon": 1e-06,
   "model_type": "mt5",
-  "num_decoder_layers": 12,
-  "num_heads": 12,
-  "num_layers": 12,
-  "output_past": true,
+  "num_decoder_layers": 8,
+  "num_heads": 6,
+  "num_layers": 8,
   "pad_token_id": 0,
   "relative_attention_max_distance": 128,
   "relative_attention_num_buckets": 32,

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e91bc3a4236c284739eae42e5c3afb83e677585afbc9d9ab081526b1e2565343
-size 2329702581
+oid sha256:dc1fe41902491e6f01ef0c1675c57890cd827062490ffdd23d749606198fd358
+size 1200772613

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ceb8ce40620efa895cc2c171ae88ef5acbede90334a352cb62a1a54e0cafde3e
+oid sha256:24d418b407dc881fc61125bde60c14914d17a529403328852171f8af9d7e11ce
 size 4155