Upload T5ForConditionalGeneration

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "trained_models/t5_qasrl-baseline-10.05.22",
+  "_name_or_path": "/home/nlp/kleinay/tmp/t5-tst-summarization/qasrl/full-data-linearization/all_by_role_ordering",
   "append_verb_form": true,
   "architectures": [
     "T5ForConditionalGeneration"
@@ -9,8 +9,9 @@
   "d_model": 512,
   "debug_mode": false,
   "decoder_start_token_id": 0,
-  "description": "optimal qasrl baseline config based on finer sweep1",
-  "dir_switch": "qasrl/baseline",
+  "dense_act_fn": "relu",
+  "description": "qasrl baseline with output_linearization=all_by_role_ordering",
+  "dir_switch": "full-data-linearization/all_by_role_ordering",
   "do_eval_on": "validation",
   "dropout_rate": 0.1,
   "eos_token_id": 1,
@@ -21,6 +22,7 @@
   "gradient_accumulation_steps": 8,
   "initializer_factor": 1.0,
   "is_encoder_decoder": true,
+  "is_gated_act": false,
   "layer_norm_epsilon": 1e-06,
   "learning_rate": 0.005,
   "load_best_model_at_end": true,
@@ -41,11 +43,12 @@
   "predicate_marker_type": "generic",
   "predict_with_generate": true,
   "preprocess_input_func": "input_predicate_marker",
+  "preprocess_output_func": "all_by_role_ordering",
   "preprocessing_kwargs": {
     "append_verb_form": true,
     "debug_mode": false,
-    "description": "optimal qasrl baseline config based on finer sweep1",
-    "dir_switch": "qasrl/baseline",
+    "description": "qasrl baseline with output_linearization=all_by_role_ordering",
+    "dir_switch": "full-data-linearization/all_by_role_ordering",
     "do_eval_on": "validation",
     "dropout_rate": 0.1,
     "eval_steps": 500,
@@ -65,16 +68,18 @@
     "predicate_marker_type": "generic",
     "predict_with_generate": true,
     "preprocess_input_func": "input_predicate_marker",
+    "preprocess_output_func": "all_by_role_ordering",
     "qanom_joint_factor": 1,
     "save_steps": 500,
     "save_strategy": "steps",
     "seed": 44,
     "source_prefix": "parse: ",
     "train_dataset": "qasrl",
-    "train_epochs": 5,
+    "train_epochs": 20,
     "use_bilateral_predicate_marker": true
   },
   "qanom_joint_factor": 1,
+  "relative_attention_max_distance": 128,
   "relative_attention_num_buckets": 32,
   "save_steps": 500,
   "save_strategy": "steps",
@@ -111,8 +116,8 @@
   },
   "torch_dtype": "float32",
   "train_dataset": "qasrl",
-  "train_epochs": 5,
-  "transformers_version": "4.17.0",
+  "train_epochs": 20,
+  "transformers_version": "4.26.1",
   "use_bilateral_predicate_marker": true,
   "use_cache": true,
   "vocab_size": 32101

generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "decoder_start_token_id": 0,
+  "eos_token_id": 1,
+  "pad_token_id": 0,
+  "transformers_version": "4.26.1"
+}

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:83821efedd28325976e71d0e003bed5678ea9de6bef45f29f90004118311f402
-size 242028475
+oid sha256:99bf08ccbd1a2886850e03cbf398a6c0599815818acf35f4cbc7bffe5bf4481d
+size 242016842