End of training

README.md

@@ -1,8 +1,15 @@
 ---
+base_model: models/zephyr-gemma-2-9b-sft-4k
 tags:
+- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
+- trl
+- dpo
+- generated_from_trainer
+datasets:
+- HuggingFaceH4/ultrafeedback_binarized
 model-index:
 - name: zephyr-gemma-2-9b-dpo-4k
   results: []
@@ -13,17 +20,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # zephyr-gemma-2-9b-dpo-4k
 
-This model was trained from scratch on an unknown dataset.
+This model is a fine-tuned version of [models/zephyr-gemma-2-9b-sft-4k](https://huggingface.co/models/zephyr-gemma-2-9b-sft-4k) on the HuggingFaceH4/ultrafeedback_binarized dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.5449
-- Rewards/chosen: -0.9048
-- Rewards/rejected: -1.4075
-- Rewards/accuracies: 0.6760
-- Rewards/margins: 0.5028
-- Logps/rejected: -487.0408
-- Logps/chosen: -459.3047
-- Logits/rejected: -11.1294
-- Logits/chosen: -11.3851
+- Loss: 0.5439
+- Rewards/chosen: -0.9090
+- Rewards/rejected: -1.4174
+- Rewards/accuracies: 0.6720
+- Rewards/margins: 0.5084
+- Logps/rejected: -488.0301
+- Logps/chosen: -459.7270
+- Logits/rejected: -11.0859
+- Logits/chosen: -11.3601
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 0.998691442030882,
+    "eval_logits/chosen": -11.360100746154785,
+    "eval_logits/rejected": -11.085914611816406,
+    "eval_logps/chosen": -459.7269592285156,
+    "eval_logps/rejected": -488.03009033203125,
+    "eval_loss": 0.5439372658729553,
+    "eval_rewards/accuracies": 0.671999990940094,
+    "eval_rewards/chosen": -0.9089793562889099,
+    "eval_rewards/margins": 0.5084334015846252,
+    "eval_rewards/rejected": -1.4174127578735352,
+    "eval_runtime": 223.5785,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 8.945,
+    "eval_steps_per_second": 1.118,
     "total_flos": 0.0,
     "train_loss": 0.5785154289169632,
     "train_runtime": 17436.3201,

config.json

@@ -29,6 +29,6 @@
   "sliding_window_size": 4096,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.45.0.dev0",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 256000
 }

eval_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 0.998691442030882,
+    "eval_logits/chosen": -11.360100746154785,
+    "eval_logits/rejected": -11.085914611816406,
+    "eval_logps/chosen": -459.7269592285156,
+    "eval_logps/rejected": -488.03009033203125,
+    "eval_loss": 0.5439372658729553,
+    "eval_rewards/accuracies": 0.671999990940094,
+    "eval_rewards/chosen": -0.9089793562889099,
+    "eval_rewards/margins": 0.5084334015846252,
+    "eval_rewards/rejected": -1.4174127578735352,
+    "eval_runtime": 223.5785,
+    "eval_samples": 2000,
+    "eval_samples_per_second": 8.945,
+    "eval_steps_per_second": 1.118
+}

runs/Aug11_15-24-00_action-graph-trainer/events.out.tfevents.1723408324.action-graph-trainer.2255898.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed10c075192166c6d31b6d373a21955ce6fd56415ad0c7c6f06d27737950f335
+size 828