Training in progress, step 500

config.json

@@ -3,9 +3,9 @@
     "T5ForConditionalGeneration"
   ],
   "classifier_dropout": 0.0,
-  "d_ff": 2048,
+  "d_ff": 1024,
   "d_kv": 64,
-  "d_model": 512,
+  "d_model": 128,
   "decoder_start_token_id": 3,
   "dense_act_fn": "relu",
   "dropout_rate": 0.1,
@@ -16,9 +16,9 @@
   "is_gated_act": true,
   "layer_norm_epsilon": 1e-06,
   "model_type": "t5",
-  "num_decoder_layers": 8,
-  "num_heads": 8,
-  "num_layers": 8,
+  "num_decoder_layers": 2,
+  "num_heads": 4,
+  "num_layers": 2,
   "pad_token_id": 1,
   "relative_attention_max_distance": 128,
   "relative_attention_num_buckets": 32,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:579b19b53147459d758e682e0261400ea4113e5d23ac645619998a0e2991f537
-size 311922600
+oid sha256:2a280b717c88bfe503ee7276d9654f02bd3d2c59e2651f617ec03826df8a9278
+size 11905784

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fb1791b2c49e468bed5edaeccb8cbf955693cedfc76960b5a9b53ce9ad295e21
+oid sha256:27282a459303c45a7ea1398f33e94094a56e6c8251876f324aca9c4ad47de520
 size 5624