Upload TFT5ForConditionalGeneration (#2)

- Upload TFT5ForConditionalGeneration (939a128f9ec63093c168f8e9486b1a06ac3997fa)

README.md

@@ -1,6 +1,6 @@
 ---
 license: apache-2.0
-base_model: ai-forever/FRED-T5-large
+base_model: ai-forever/FRED-T5-1.7B
 tags:
 - generated_from_keras_callback
 model-index:
@@ -13,9 +13,9 @@ probably proofread and complete it, then remove this comment. -->
 
 # t5-jumoreski
 
-This model is a fine-tuned version of [ai-forever/FRED-T5-large](https://huggingface.co/ai-forever/FRED-T5-large) on an unknown dataset.
+This model is a fine-tuned version of [ai-forever/FRED-T5-1.7B](https://huggingface.co/ai-forever/FRED-T5-1.7B) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Train Loss: 0.1765
+- Train Loss: 0.3638
 - Epoch: 4
 
 ## Model description
@@ -35,18 +35,18 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- optimizer: {'name': 'Adam', 'weight_decay': 0.01, 'clipnorm': 1, 'global_clipnorm': None, 'clipvalue': None, 'use_ema': False, 'ema_momentum': 0.99, 'ema_overwrite_frequency': None, 'jit_compile': False, 'is_legacy_optimizer': False, 'learning_rate': {'class_name': 'CosineDecay', 'config': {'initial_learning_rate': 2e-06, 'decay_steps': 3290, 'alpha': 5e-09, 'name': None}}, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-07, 'amsgrad': False}
+- optimizer: {'name': 'Adam', 'weight_decay': 0.01, 'clipnorm': 1, 'global_clipnorm': None, 'clipvalue': None, 'use_ema': False, 'ema_momentum': 0.99, 'ema_overwrite_frequency': None, 'jit_compile': False, 'is_legacy_optimizer': False, 'learning_rate': {'class_name': 'CosineDecay', 'config': {'initial_learning_rate': 2e-06, 'decay_steps': 5345, 'alpha': 5e-09, 'name': None}}, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-07, 'amsgrad': False}
 - training_precision: mixed_bfloat16
 
 ### Training results
 
 | Train Loss | Epoch |
 |:----------:|:-----:|
-| 0.2797     | 0     |
-| 0.2091     | 1     |
-| 0.1885     | 2     |
-| 0.1804     | 3     |
-| 0.1765     | 4     |
+| 0.3673     | 0     |
+| 0.3652     | 1     |
+| 0.3645     | 2     |
+| 0.3645     | 3     |
+| 0.3638     | 4     |
 
 
 ### Framework versions

config.json

@@ -1,12 +1,12 @@
 {
-  "_name_or_path": "ai-forever/FRED-T5-large",
+  "_name_or_path": "ai-forever/FRED-T5-1.7B",
   "architectures": [
     "T5ForConditionalGeneration"
   ],
   "classifier_dropout": 0.0,
-  "d_ff": 2816,
+  "d_ff": 4096,
   "d_kv": 64,
-  "d_model": 1024,
+  "d_model": 1536,
   "decoder_start_token_id": 0,
   "dense_act_fn": "gelu_new",
   "dropout_rate": 0.1,
@@ -19,7 +19,7 @@
   "layer_norm_epsilon": 1e-06,
   "model_type": "t5",
   "num_decoder_layers": 24,
-  "num_heads": 16,
+  "num_heads": 24,
   "num_layers": 24,
   "output_past": true,
   "pad_token_id": 0,

tf_model.h5

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3086edab331e9280020590c659ad5aebd23aaf6ca1ac878a40a61d6d8bbf94e1
-size 3695699576
+oid sha256:7572db74655669bc1c017b9174e3a9f0d574f1cb44250ff3ce7b34024807a028
+size 7581417048