Model save

README.md

@@ -27,7 +27,7 @@ print(output["generated_text"])
 
 ## Training procedure
 
-[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/mingzeli/PreThink_MemAgent/runs/f0f3qo3c) 
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/mingzeli/PreThink_MemAgent/runs/kprp9gb7) 
 
 
 This model was trained with SFT.

all_results.json

@@ -1,8 +1,8 @@
 {
-    "total_flos": 1446350766178304.0,
-    "train_loss": 8.238658274636032,
-    "train_runtime": 17767.8846,
+    "total_flos": 2.2203599102840668e+18,
+    "train_loss": 0.3945140190995656,
+    "train_runtime": 7938.1861,
     "train_samples": 79626,
-    "train_samples_per_second": 8.963,
-    "train_steps_per_second": 0.07
+    "train_samples_per_second": 20.062,
+    "train_steps_per_second": 0.079
 }

train_results.json

@@ -1,8 +1,8 @@
 {
-    "total_flos": 1446350766178304.0,
-    "train_loss": 8.238658274636032,
-    "train_runtime": 17767.8846,
+    "total_flos": 2.2203599102840668e+18,
+    "train_loss": 0.3945140190995656,
+    "train_runtime": 7938.1861,
     "train_samples": 79626,
-    "train_samples_per_second": 8.963,
-    "train_steps_per_second": 0.07
+    "train_samples_per_second": 20.062,
+    "train_steps_per_second": 0.079
 }