End of training

README.md

@@ -2,12 +2,13 @@
 license: gemma
 library_name: peft
 tags:
+- alignment-handbook
 - trl
 - sft
 - generated_from_trainer
 base_model: google/gemma-7b
 datasets:
-- generator
+- chansung/merged_ds_coding
 model-index:
 - name: coding_llamaduo_result3
   results: []
@@ -18,7 +19,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # coding_llamaduo_result3
 
-This model is a fine-tuned version of [google/gemma-7b](https://huggingface.co/google/gemma-7b) on the generator dataset.
+This model is a fine-tuned version of [google/gemma-7b](https://huggingface.co/google/gemma-7b) on the chansung/merged_ds_coding dataset.
 It achieves the following results on the evaluation set:
 - Loss: 1.7502
 

all_results.json

@@ -1,5 +1,10 @@
 {
     "epoch": 10.0,
+    "eval_loss": 1.750213384628296,
+    "eval_runtime": 1.1155,
+    "eval_samples": 16,
+    "eval_samples_per_second": 1.793,
+    "eval_steps_per_second": 0.896,
     "total_flos": 1.257706785410646e+18,
     "train_loss": 1.5742560074096772,
     "train_runtime": 5911.531,

eval_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 5.0,
-    "eval_loss": 1.248147964477539,
-    "eval_runtime": 1.085,
+    "epoch": 10.0,
+    "eval_loss": 1.750213384628296,
+    "eval_runtime": 1.1155,
     "eval_samples": 16,
-    "eval_samples_per_second": 1.843,
-    "eval_steps_per_second": 0.922
+    "eval_samples_per_second": 1.793,
+    "eval_steps_per_second": 0.896
 }

runs/Apr21_23-51-41_060407ee4835/events.out.tfevents.1713764138.060407ee4835.3304.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3bb047cb09b767c6c9ff1aa13fc6b98ee2d09ac22a44e6f3a68e88f288d7e323
+size 359