Upload folder using huggingface_hub

README.md

@@ -16,12 +16,12 @@ Multi-task transformer with shared MoE layers for:
 ## Performance
 
 ### Classification
-- Validation Accuracy: 0.3580
+- Validation Accuracy: 0.3420
 
 ### Summarization
-- ROUGE-1: 0.2154
-- ROUGE-2: 0.0298
-- ROUGE-L: 0.1842
+- ROUGE-1: 0.2250
+- ROUGE-2: 0.0333
+- ROUGE-L: 0.2078
 
 ## Benefits
 

history.csv

@@ -1,4 +1,4 @@
 epoch,train_sum_loss,train_class_loss,train_class_acc,val_sum_loss,val_class_loss,val_class_acc,val_rouge1,val_rouge2,val_rougeL
-1,6.539164045969645,1.080719949245453,0.407,6.074222045898438,1.0701570644378662,0.344,0.1856974871518223,0.01907572984467157,0.16932123951907738
-2,5.473755016326904,1.0669607858657837,0.3973333333333333,5.697721607208252,0.9951665096282959,0.506,0.2045861270829145,0.024241315885172792,0.17770470477158784
-3,4.5570032602945965,1.0329711267153423,0.4613333333333333,5.637807807922363,1.338168493270874,0.358,0.21535871469554163,0.02978786749856711,0.1842483145510327
+1,6.272962152862549,1.0185239918708802,0.4368,5.8008860549926755,1.1023406219482421,0.304,0.17660690248954364,0.018371370658335634,0.15689190075772386
+2,5.219159950256348,1.099994719696045,0.3462,5.416218318939209,1.0967657680511476,0.386,0.21779871449217764,0.028536134623968267,0.19329056631523484
+3,4.404316672515869,1.0899620735168456,0.361,5.30168932723999,1.7727555074691772,0.342,0.22501668712961082,0.03326989668183761,0.20778061653444058

metrics.json

@@ -2,39 +2,39 @@
   "history": [
     {
       "epoch": 1,
-      "train_sum_loss": 6.539164045969645,
-      "train_class_loss": 1.080719949245453,
-      "train_class_acc": 0.407,
-      "val_sum_loss": 6.074222045898438,
-      "val_class_loss": 1.0701570644378662,
-      "val_class_acc": 0.344,
-      "val_rouge1": 0.1856974871518223,
-      "val_rouge2": 0.01907572984467157,
-      "val_rougeL": 0.16932123951907738
+      "train_sum_loss": 6.272962152862549,
+      "train_class_loss": 1.0185239918708802,
+      "train_class_acc": 0.4368,
+      "val_sum_loss": 5.8008860549926755,
+      "val_class_loss": 1.1023406219482421,
+      "val_class_acc": 0.304,
+      "val_rouge1": 0.17660690248954364,
+      "val_rouge2": 0.018371370658335634,
+      "val_rougeL": 0.15689190075772386
     },
     {
       "epoch": 2,
-      "train_sum_loss": 5.473755016326904,
-      "train_class_loss": 1.0669607858657837,
-      "train_class_acc": 0.3973333333333333,
-      "val_sum_loss": 5.697721607208252,
-      "val_class_loss": 0.9951665096282959,
-      "val_class_acc": 0.506,
-      "val_rouge1": 0.2045861270829145,
-      "val_rouge2": 0.024241315885172792,
-      "val_rougeL": 0.17770470477158784
+      "train_sum_loss": 5.219159950256348,
+      "train_class_loss": 1.099994719696045,
+      "train_class_acc": 0.3462,
+      "val_sum_loss": 5.416218318939209,
+      "val_class_loss": 1.0967657680511476,
+      "val_class_acc": 0.386,
+      "val_rouge1": 0.21779871449217764,
+      "val_rouge2": 0.028536134623968267,
+      "val_rougeL": 0.19329056631523484
     },
     {
       "epoch": 3,
-      "train_sum_loss": 4.5570032602945965,
-      "train_class_loss": 1.0329711267153423,
-      "train_class_acc": 0.4613333333333333,
-      "val_sum_loss": 5.637807807922363,
-      "val_class_loss": 1.338168493270874,
-      "val_class_acc": 0.358,
-      "val_rouge1": 0.21535871469554163,
-      "val_rouge2": 0.02978786749856711,
-      "val_rougeL": 0.1842483145510327
+      "train_sum_loss": 4.404316672515869,
+      "train_class_loss": 1.0899620735168456,
+      "train_class_acc": 0.361,
+      "val_sum_loss": 5.30168932723999,
+      "val_class_loss": 1.7727555074691772,
+      "val_class_acc": 0.342,
+      "val_rouge1": 0.22501668712961082,
+      "val_rouge2": 0.03326989668183761,
+      "val_rougeL": 0.20778061653444058
     }
   ],
   "config": {

model.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:30128df14f8ee585a1c686f17af6e9df803766e5ac6d55d8c59c5c93ba437131
+oid sha256:b2545d4f075dd3d1dfeadaa8b83777d84078f8373b4f4ee7c4418f5186bd3b97
 size 385512714

tokenizer/tokenizer_config.json

@@ -1,4 +1,5 @@
 {
+  "add_prefix_space": null,
   "added_tokens_decoder": {
     "0": {
       "content": "<pad>",
@@ -930,6 +931,7 @@
   "clean_up_tokenization_spaces": true,
   "eos_token": "</s>",
   "extra_ids": 100,
+  "extra_special_tokens": {},
   "model_max_length": 512,
   "pad_token": "<pad>",
   "tokenizer_class": "T5Tokenizer",