End of training

README.md

@@ -1,10 +1,15 @@
 ---
+base_model: data/gemma-2b
 tags:
+- alignment-handbook
+- trl
+- sft
+- generated_from_trainer
 - trl
 - sft
 - generated_from_trainer
 datasets:
-- generator
+- argilla/dpo-mix-7k
 model-index:
 - name: zephyr-2b-gemma-dft
   results: []
@@ -16,7 +21,7 @@ should probably proofread and complete it, then remove this comment. -->
 [<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://zebra.wandb.io/cto/distillm/runs/kje9xugl)
 # zephyr-2b-gemma-dft
 
-This model was trained from scratch on the generator dataset.
+This model is a fine-tuned version of [data/gemma-2b](https://huggingface.co/data/gemma-2b) on the argilla/dpo-mix-7k dataset.
 It achieves the following results on the evaluation set:
 - Loss: 0.0000
 

all_results.json

@@ -1,5 +1,10 @@
 {
     "epoch": 0.9981515711645101,
+    "eval_loss": 1.3726342331210617e-05,
+    "eval_runtime": 50.9693,
+    "eval_samples": 750,
+    "eval_samples_per_second": 19.051,
+    "eval_steps_per_second": 2.394,
     "total_flos": 5.260333472022528e+16,
     "train_loss": 1.8035727003330572e-05,
     "train_runtime": 2689.0507,

config.json

@@ -24,6 +24,6 @@
   "rope_theta": 10000.0,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.42.4",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 256000
 }

eval_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 0.9981515711645101,
+    "eval_loss": 1.3726342331210617e-05,
+    "eval_runtime": 50.9693,
+    "eval_samples": 750,
+    "eval_samples_per_second": 19.051,
+    "eval_steps_per_second": 2.394
+}

runs/Jul23_14-10-34_ale-distillm-8-0-0/events.out.tfevents.1721743269.ale-distillm-8-0-0.46360.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dea757015fb058d92f113d7f7d08820ca7406431658d26f566eb8bdeefafee1a
+size 359