End of training

README.md

@@ -2,9 +2,16 @@
 license: mit
 base_model: BramVanroy/fietje-2b-sft
 tags:
+- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
+- trl
+- dpo
+- generated_from_trainer
+datasets:
+- BramVanroy/ultra_feedback_dutch_cleaned
+- BramVanroy/orca_dpo_pairs_dutch_cleaned
 model-index:
 - name: fietje-2b-dpo-lr2.0e-6-beta0.2-gradaccum2-v6
   results: []
@@ -15,7 +22,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # fietje-2b-dpo-lr2.0e-6-beta0.2-gradaccum2-v6
 
-This model is a fine-tuned version of [BramVanroy/fietje-2b-sft](https://huggingface.co/BramVanroy/fietje-2b-sft) on the None dataset.
+This model is a fine-tuned version of [BramVanroy/fietje-2b-sft](https://huggingface.co/BramVanroy/fietje-2b-sft) on the BramVanroy/ultra_feedback_dutch_cleaned and the BramVanroy/orca_dpo_pairs_dutch_cleaned datasets.
 It achieves the following results on the evaluation set:
 - Loss: 0.2842
 - Rewards/chosen: -1.1549

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 1.0,
+    "eval_logits/chosen": -1.352773666381836,
+    "eval_logits/rejected": -1.2867770195007324,
+    "eval_logps/chosen": -451.33642578125,
+    "eval_logps/rejected": -657.6813354492188,
+    "eval_loss": 0.2841675579547882,
+    "eval_rewards/accuracies": 0.88671875,
+    "eval_rewards/chosen": -1.1548646688461304,
+    "eval_rewards/margins": 2.4814703464508057,
+    "eval_rewards/rejected": -3.6363346576690674,
+    "eval_runtime": 527.3224,
+    "eval_samples": 1536,
+    "eval_samples_per_second": 2.913,
+    "eval_steps_per_second": 0.728,
     "train_loss": 0.3634731802597361,
     "train_runtime": 16591.2723,
     "train_samples": 18653,

config.json

@@ -29,6 +29,6 @@
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.39.1",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 50297
 }

eval_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 1.0,
+    "eval_logits/chosen": -1.352773666381836,
+    "eval_logits/rejected": -1.2867770195007324,
+    "eval_logps/chosen": -451.33642578125,
+    "eval_logps/rejected": -657.6813354492188,
+    "eval_loss": 0.2841675579547882,
+    "eval_rewards/accuracies": 0.88671875,
+    "eval_rewards/chosen": -1.1548646688461304,
+    "eval_rewards/margins": 2.4814703464508057,
+    "eval_rewards/rejected": -3.6363346576690674,
+    "eval_runtime": 527.3224,
+    "eval_samples": 1536,
+    "eval_samples_per_second": 2.913,
+    "eval_steps_per_second": 0.728
+}