End of training

README.md

@@ -2,9 +2,15 @@
 license: apache-2.0
 base_model: martimfasantos/tinyllama-1.1b-sum-sft-full_old
 tags:
+- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
+- trl
+- dpo
+- generated_from_trainer
+datasets:
+- openai/summarize_from_feedback
 model-index:
 - name: tinyllama-1.1b-sum-dpo-full_LR5e-8_2epochs_old
   results: []
@@ -15,7 +21,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # tinyllama-1.1b-sum-dpo-full_LR5e-8_2epochs_old
 
-This model is a fine-tuned version of [martimfasantos/tinyllama-1.1b-sum-sft-full_old](https://huggingface.co/martimfasantos/tinyllama-1.1b-sum-sft-full_old) on an unknown dataset.
+This model is a fine-tuned version of [martimfasantos/tinyllama-1.1b-sum-sft-full_old](https://huggingface.co/martimfasantos/tinyllama-1.1b-sum-sft-full_old) on the openai/summarize_from_feedback dataset.
 It achieves the following results on the evaluation set:
 - Loss: 0.6808
 - Rewards/chosen: -0.1214

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 2.0,
+    "eval_logits/chosen": -2.962392568588257,
+    "eval_logits/rejected": -2.956583023071289,
+    "eval_logps/chosen": -70.84989929199219,
+    "eval_logps/rejected": -78.1531753540039,
+    "eval_loss": 0.6807694435119629,
+    "eval_rewards/accuracies": 0.6089683771133423,
+    "eval_rewards/chosen": -0.12137996405363083,
+    "eval_rewards/margins": 0.028350669890642166,
+    "eval_rewards/rejected": -0.1497306227684021,
+    "eval_runtime": 382.3469,
+    "eval_samples": 4304,
+    "eval_samples_per_second": 11.257,
+    "eval_steps_per_second": 1.407,
     "total_flos": 0.0,
     "train_loss": 0.6777890619945723,
     "train_runtime": 94641.0715,

config.json

@@ -24,6 +24,6 @@
   "tie_word_embeddings": false,
   "torch_dtype": "float32",
   "transformers_version": "4.41.2",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 32000
 }

eval_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 2.0,
+    "eval_logits/chosen": -2.962392568588257,
+    "eval_logits/rejected": -2.956583023071289,
+    "eval_logps/chosen": -70.84989929199219,
+    "eval_logps/rejected": -78.1531753540039,
+    "eval_loss": 0.6807694435119629,
+    "eval_rewards/accuracies": 0.6089683771133423,
+    "eval_rewards/chosen": -0.12137996405363083,
+    "eval_rewards/margins": 0.028350669890642166,
+    "eval_rewards/rejected": -0.1497306227684021,
+    "eval_runtime": 382.3469,
+    "eval_samples": 4304,
+    "eval_samples_per_second": 11.257,
+    "eval_steps_per_second": 1.407
+}

runs/Jun17_08-38-05_poseidon/events.out.tfevents.1718709106.poseidon.4064829.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1953d949a92ff272d2db6a0038e76b4ba08bda7bccbf77e01f49a998ffccaf0b
+size 828