add model

README.md

@@ -16,9 +16,9 @@ should probably proofread and complete it, then remove this comment. -->
 
 # bart-mlm
 
-This model is a fine-tuned version of [](https://huggingface.co/) on the None dataset.
+This model is a fine-tuned version of [mwesner/bart-mlm](https://huggingface.co/mwesner/bart-mlm) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 7.0454
+- Loss: 7.5338
 
 ## Model description
 
@@ -37,43 +37,26 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 2e-05
-- train_batch_size: 8
-- eval_batch_size: 8
+- learning_rate: 0.001
+- train_batch_size: 2
+- eval_batch_size: 2
 - seed: 42
+- gradient_accumulation_steps: 8
+- total_train_batch_size: 16
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: cosine_with_restarts
 - lr_scheduler_warmup_steps: 500
-- num_epochs: 2
+- num_epochs: 5
 
 ### Training results
 
-| Training Loss | Epoch | Step   | Validation Loss |
-|:-------------:|:-----:|:------:|:---------------:|
-| 7.3964        | 0.08  | 10000  | 7.4260          |
-| 7.3303        | 0.16  | 20000  | 7.3293          |
-| 7.2823        | 0.25  | 30000  | 7.2939          |
-| 7.2526        | 0.33  | 40000  | 7.2624          |
-| 7.224         | 0.41  | 50000  | 7.2391          |
-| 7.204         | 0.49  | 60000  | 7.2192          |
-| 7.1693        | 0.57  | 70000  | 7.2053          |
-| 7.1845        | 0.66  | 80000  | 7.1815          |
-| 7.1538        | 0.74  | 90000  | 7.1610          |
-| 7.1278        | 0.82  | 100000 | 7.1459          |
-| 7.1158        | 0.9   | 110000 | 7.1332          |
-| 7.1126        | 0.98  | 120000 | 7.1200          |
-| 7.0827        | 1.07  | 130000 | 7.1105          |
-| 7.0857        | 1.15  | 140000 | 7.0978          |
-| 7.0646        | 1.23  | 150000 | 7.0784          |
-| 7.0606        | 1.31  | 160000 | 7.0716          |
-| 7.0539        | 1.39  | 170000 | 7.0659          |
-| 7.0282        | 1.48  | 180000 | 7.0602          |
-| 7.0427        | 1.56  | 190000 | 7.0553          |
-| 7.0366        | 1.64  | 200000 | 7.0486          |
-| 7.0133        | 1.72  | 210000 | 7.0529          |
-| 7.0235        | 1.8   | 220000 | 7.0458          |
-| 7.0167        | 1.89  | 230000 | 7.0434          |
-| 7.0208        | 1.97  | 240000 | 7.0435          |
+| Training Loss | Epoch | Step  | Validation Loss |
+|:-------------:|:-----:|:-----:|:---------------:|
+| 7.5202        | 1.0   | 15237 | 7.5964          |
+| 7.5151        | 2.0   | 30474 | 7.5400          |
+| 7.5157        | 3.0   | 45711 | 7.5351          |
+| 7.5172        | 4.0   | 60948 | 7.5317          |
+| 7.5108        | 5.0   | 76185 | 7.5338          |
 
 
 ### Framework versions

config.json

@@ -1,4 +1,5 @@
 {
+  "_name_or_path": "mwesner/bart-mlm",
   "activation_dropout": 0.0,
   "activation_function": "gelu_new",
   "architectures": [

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f2aa1bb3bf469d1b814e3e644507d983c3944ef45780b5bf28a926f8c3cc69ef
+oid sha256:6973d75bd1d5d885201d3212cefccd11ffa156e08d77c805c685b1936dbae085
 size 1625557313

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d6715133c83b5593f520595bfe4a43dfb0af1e2b1b0e07dcbff95159068b6e11
+oid sha256:1b5d042db71939915533f2461a0667e96ea3f46a809941bef5fec36939d77d84
 size 2607