Finetune on MentalChat16K - eval_loss: 0.7298

README.md

@@ -16,7 +16,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [microsoft/phi-2](https://huggingface.co/microsoft/phi-2) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.7296
+- Loss: 0.7112
 
 ## Model description
 

training_metrics.json

@@ -0,0 +1,43 @@
+{
+  "model": "phi2-mental-health",
+  "base_model": "microsoft/phi-2",
+  "dataset": "ShenLab/MentalChat16K",
+  "lora_config": {
+    "rank": 16,
+    "alpha": 32,
+    "target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "dense"
+    ],
+    "dropout": 0.1
+  },
+  "training": {
+    "final_train_loss": 0.7486542798042297,
+    "total_steps": 2500,
+    "epochs": 4,
+    "learning_rate": 0.0002,
+    "per_device_batch_size": 4,
+    "gradient_accumulation": 2
+  },
+  "evaluation": {
+    "eval_loss": 0.7297702431678772,
+    "eval_runtime": 4064.1661,
+    "eval_samples_per_second": 0.116,
+    "eval_steps_per_second": 0.029,
+    "epoch": 3.7397157816005984
+  },
+  "test_eval": {
+    "eval_loss": 0.7111775875091553,
+    "eval_runtime": 39.2705,
+    "eval_samples_per_second": 12.019,
+    "eval_steps_per_second": 3.005,
+    "epoch": 3.7397157816005984
+  },
+  "dataset_stats": {
+    "train_size": 5347,
+    "val_size": 472,
+    "test_size": 472
+  }
+}