Auto-upgrade v4: composite 0.8757 | GLEU 0.7792 | BERTScore 0.9793 | 1-WER 0.8687 | base=coedit-large, r=32, 256-token ctx, coedit+GEC data, faithfulness gate [merged weights → merged/]

merged/config.json

@@ -3,9 +3,9 @@
     "T5ForConditionalGeneration"
   ],
   "classifier_dropout": 0.0,
-  "d_ff": 1024,
+  "d_ff": 2816,
   "d_kv": 64,
-  "d_model": 512,
+  "d_model": 1024,
   "decoder_start_token_id": 0,
   "dense_act_fn": "gelu_new",
   "dropout_rate": 0.1,
@@ -19,45 +19,16 @@
   "layer_norm_epsilon": 1e-06,
   "model_type": "t5",
   "n_positions": 512,
-  "num_decoder_layers": 8,
-  "num_heads": 6,
-  "num_layers": 8,
+  "num_decoder_layers": 24,
+  "num_heads": 16,
+  "num_layers": 24,
   "output_past": true,
   "pad_token_id": 0,
   "relative_attention_max_distance": 128,
   "relative_attention_num_buckets": 32,
   "scale_decoder_outputs": false,
-  "task_specific_params": {
-    "summarization": {
-      "early_stopping": true,
-      "length_penalty": 2.0,
-      "max_length": 200,
-      "min_length": 30,
-      "no_repeat_ngram_size": 3,
-      "num_beams": 4,
-      "prefix": "summarize: "
-    },
-    "translation_en_to_de": {
-      "early_stopping": true,
-      "max_length": 300,
-      "num_beams": 4,
-      "prefix": "translate English to German: "
-    },
-    "translation_en_to_fr": {
-      "early_stopping": true,
-      "max_length": 300,
-      "num_beams": 4,
-      "prefix": "translate English to French: "
-    },
-    "translation_en_to_ro": {
-      "early_stopping": true,
-      "max_length": 300,
-      "num_beams": 4,
-      "prefix": "translate English to Romanian: "
-    }
-  },
   "tie_word_embeddings": false,
   "transformers_version": "5.8.0",
   "use_cache": false,
-  "vocab_size": 32128
+  "vocab_size": 32100
 }

merged/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:59635aecd8a0054a8c272a9ab2c488c1e06c8a65dedb03a35aceddc438ce9940
-size 307867048
+oid sha256:3b2c1883006f7bdb5a27c4fe2ea6305830e9943e55b1280d69d5d145b60a1243
+size 3132439432

merged/tokenizer_config.json

@@ -106,9 +106,13 @@
   ],
   "is_local": false,
   "local_files_only": false,
+  "max_length": 256,
   "model_max_length": 512,
   "pad_token": "<pad>",
   "sp_model_kwargs": {},
+  "stride": 0,
   "tokenizer_class": "T5Tokenizer",
+  "truncation_side": "right",
+  "truncation_strategy": "longest_first",
   "unk_token": "<unk>"
 }