CuATR-distilbert-LoRA

README.md

@@ -18,9 +18,9 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [distilbert-base-uncased](https://huggingface.co/distilbert-base-uncased) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.6930
-- Accuracy: 0.4348
-- F1: 0.0
+- Loss: 0.6865
+- Accuracy: 0.6087
+- F1: 0.7429
 
 ## Model description
 
@@ -52,11 +52,11 @@ The following hyperparameters were used during training:
 
 ### Training results
 
-| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1  |
-|:-------------:|:-----:|:----:|:---------------:|:--------:|:---:|
-| 0.6988        | 0.67  | 1    | 0.6933          | 0.4348   | 0.0 |
-| 0.6919        | 2.0   | 3    | 0.6931          | 0.4348   | 0.0 |
-| 0.708         | 2.67  | 4    | 0.6930          | 0.4348   | 0.0 |
+| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1     |
+|:-------------:|:-----:|:----:|:---------------:|:--------:|:------:|
+| 0.6894        | 0.67  | 1    | 0.6867          | 0.6087   | 0.7429 |
+| 0.6931        | 2.0   | 3    | 0.6865          | 0.6087   | 0.7429 |
+| 0.6873        | 2.67  | 4    | 0.6865          | 0.6087   | 0.7429 |
 
 
 ### Framework versions

adapter_config.json

@@ -12,13 +12,13 @@
   "lora_dropout": 0.1,
   "modules_to_save": null,
   "peft_type": "LORA",
-  "r": 32,
+  "r": 4,
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
+    "v_lin",
     "q_lin",
-    "k_lin",
-    "v_lin"
+    "k_lin"
   ],
   "task_type": "TOKEN_CLS"
 }

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cb1e15b0cdf0fbf5843b9a4fb29ba339fba175c5524f64d493856c454c480b60
-size 3544216
+oid sha256:a2303adc67ddf09fc04b437f94e611540e087e9104083d01ac994cf027044b73
+size 447536

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c26d240ce75d08d702c5cfea7817f2d4881a2de4521774ac22c7368cbe5c22cb
+oid sha256:28b83cee47f7fb3807d87d114642f6a24ecf08708f7e1deea6b3572f865871a8
 size 4600