Hugging Face's logo

cristian-untaru
/
lora-distilbert-medical-triage

Text Classification
Transformers
Safetensors
PEFT
English
distilbert
medical-triage
lora
healthcare
symptom-checker
natural-language-processing
academic-project
Eval Results (legacy)
text-embeddings-inference
Model card Files
xet
 Community
cristian-untaru commited on
Commit
6f074c9
·
verified ·
1 Parent(s): 20a8b0d

Add merged LoRA DistilBERT medical triage model

Browse files
Files changed (3) hide show
  1. README.md +9 -9
  2. model.safetensors +1 -1
  3. training_config.json +3 -3
README.md CHANGED
@@ -34,21 +34,21 @@ The classifier predicts one of three triage labels:
34
  - LoRA alpha: 32
35
  - LoRA dropout: 0.10
36
  - Max sequence length: 128
37
- - Epochs: 5
38
  - Best checkpoint selected by validation macro F1
39
 
40
  ## Test metrics
41
 
42
  ```json
43
  {
44
- "eval_loss": 0.7654246091842651,
45
- "eval_accuracy": 0.7453,
46
- "eval_precision": 0.7463,
47
- "eval_recall": 0.7503,
48
- "eval_f1": 0.7477,
49
- "eval_auc": 0.8684,
50
- "eval_specificity": 0.8719,
51
- "eval_iou": 0.6011
52
  }
53
  ```
54
 
 
34
  - LoRA alpha: 32
35
  - LoRA dropout: 0.10
36
  - Max sequence length: 128
37
+ - Epochs: 8
38
  - Best checkpoint selected by validation macro F1
39
 
40
  ## Test metrics
41
 
42
  ```json
43
  {
44
+ "eval_loss": 0.5353153944015503,
45
+ "eval_accuracy": 0.8491,
46
+ "eval_precision": 0.849,
47
+ "eval_recall": 0.849,
48
+ "eval_f1": 0.849,
49
+ "eval_auc": 0.9197,
50
+ "eval_specificity": 0.924,
51
+ "eval_iou": 0.7408
52
  }
53
  ```
54
 
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:25b05e2f0bc6386b3cacc21ef86075d8cb6da34fe8c15adac0a95092ee60a115
3
  size 267835644
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b21e87abb29a24ce2472f4a0a8528fd7f2b59a904462db957c8510dd666dd2d4
3
  size 267835644
training_config.json CHANGED
@@ -15,14 +15,14 @@
15
  ],
16
  "num_labels": 3,
17
  "max_length": 128,
18
- "num_train_epochs": 5,
19
- "learning_rate": 0.0001,
20
  "batch_size_train": 16,
21
  "batch_size_eval": 32,
22
  "weight_decay": 0.01,
23
  "warmup_ratio": 0.1,
24
  "fp16": true,
25
- "early_stopping_patience": 2,
26
  "metric_for_best_model": "f1",
27
  "train_size": 490,
28
  "validation_size": 105,
 
15
  ],
16
  "num_labels": 3,
17
  "max_length": 128,
18
+ "num_train_epochs": 8,
19
+ "learning_rate": 0.0002,
20
  "batch_size_train": 16,
21
  "batch_size_eval": 32,
22
  "weight_decay": 0.01,
23
  "warmup_ratio": 0.1,
24
  "fp16": true,
25
+ "early_stopping_patience": 3,
26
  "metric_for_best_model": "f1",
27
  "train_size": 490,
28
  "validation_size": 105,