Training selesai: clean generation_config + decoder_start_token_id

config.json

@@ -1,27 +1,24 @@
 {
   "architectures": [
-    "LongT5ForConditionalGeneration"
+    "T5ForConditionalGeneration"
   ],
-  "bos_token_id": null,
-  "d_ff": 2048,
+  "classifier_dropout": 0.0,
+  "d_ff": 3072,
   "d_kv": 64,
   "d_model": 768,
   "decoder_start_token_id": 0,
-  "dense_act_fn": "gelu_new",
+  "dense_act_fn": "relu",
   "dropout_rate": 0.1,
   "dtype": "float32",
-  "encoder_attention_type": "transient-global",
   "eos_token_id": 1,
-  "feed_forward_proj": "gated-gelu",
-  "global_block_size": 16,
+  "feed_forward_proj": "relu",
   "initializer_factor": 1.0,
   "is_decoder": false,
   "is_encoder_decoder": true,
-  "is_gated_act": true,
+  "is_gated_act": false,
   "layer_norm_epsilon": 1e-06,
-  "local_radius": 127,
-  "model_type": "longt5",
-  "n_positions": 4096,
+  "model_type": "t5",
+  "n_positions": 512,
   "num_decoder_layers": 12,
   "num_heads": 12,
   "num_layers": 12,
@@ -29,7 +26,37 @@
   "pad_token_id": 0,
   "relative_attention_max_distance": 128,
   "relative_attention_num_buckets": 32,
-  "tie_word_embeddings": false,
+  "scale_decoder_outputs": true,
+  "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": true,
   "transformers_version": "5.0.0",
   "use_cache": false,
   "vocab_size": 32100

generation_config.json

@@ -1,36 +1,14 @@
 {
-  "_from_model_config": false,
-  "assistant_confidence_threshold": 0.4,
-  "assistant_lookbehind": 10,
   "decoder_start_token_id": 0,
-  "diversity_penalty": 0.0,
-  "do_sample": true,
-  "early_stopping": false,
-  "encoder_no_repeat_ngram_size": 0,
-  "encoder_repetition_penalty": 1.0,
+  "do_sample": false,
+  "early_stopping": true,
   "eos_token_id": 1,
-  "epsilon_cutoff": 0.0,
-  "eta_cutoff": 0.0,
   "length_penalty": 1.0,
-  "max_length": 20,
   "max_new_tokens": 64,
-  "min_length": 0,
   "no_repeat_ngram_size": 2,
-  "num_assistant_tokens": 20,
-  "num_assistant_tokens_schedule": "constant",
-  "num_beam_groups": 1,
-  "num_beams": 8,
-  "num_return_sequences": 5,
-  "output_scores": false,
+  "num_beams": 4,
+  "num_return_sequences": 1,
   "pad_token_id": 0,
-  "remove_invalid_values": false,
-  "repetition_penalty": 1.0,
-  "return_dict_in_generate": false,
-  "target_lookbehind": 10,
-  "temperature": 0.8,
-  "top_k": 50,
-  "top_p": 0.95,
-  "transformers_version": "5.0.0",
-  "typical_p": 1.0,
-  "use_cache": true
+  "repetition_penalty": 1.2,
+  "transformers_version": "5.0.0"
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:eab6d229f31e6720b5515aa66638b8f40510040967f41ae5d5fce9f20117dc1e
-size 990214168
+oid sha256:ccccc4e7c8be7a47d5aac1299debac88d1291a59a2770245268b1db2e238d9dc
+size 891558696