End of training

README.md

@@ -2,9 +2,15 @@
 license: apache-2.0
 base_model: martimfasantos/tinyllama-1.1b-sum-sft-full
 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
   results: []
@@ -15,7 +21,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # tinyllama-1.1b-sum-dpo-full
 
-This model is a fine-tuned version of [martimfasantos/tinyllama-1.1b-sum-sft-full](https://huggingface.co/martimfasantos/tinyllama-1.1b-sum-sft-full) on an unknown dataset.
+This model is a fine-tuned version of [martimfasantos/tinyllama-1.1b-sum-sft-full](https://huggingface.co/martimfasantos/tinyllama-1.1b-sum-sft-full) on the openai/summarize_from_feedback dataset.
 It achieves the following results on the evaluation set:
 - Loss: 0.6549
 - Rewards/chosen: -0.4976

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 2.0,
+    "eval_logits/chosen": -2.5584018230438232,
+    "eval_logits/rejected": -2.5515997409820557,
+    "eval_logps/chosen": -108.46733856201172,
+    "eval_logps/rejected": -123.28102111816406,
+    "eval_loss": 0.6549462676048279,
+    "eval_rewards/accuracies": 0.6194238066673279,
+    "eval_rewards/chosen": -0.49755439162254333,
+    "eval_rewards/margins": 0.10345453023910522,
+    "eval_rewards/rejected": -0.6010088920593262,
+    "eval_runtime": 359.6814,
+    "eval_samples": 4304,
+    "eval_samples_per_second": 11.966,
+    "eval_steps_per_second": 1.496,
     "total_flos": 0.0,
     "train_loss": 0.6539983197297005,
     "train_runtime": 91180.3592,

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

@@ -1,16 +1,16 @@
 {
     "epoch": 2.0,
-    "eval_logits/chosen": -1.4189956188201904,
-    "eval_logits/rejected": -1.3920025825500488,
-    "eval_logps/chosen": -244.38746643066406,
-    "eval_logps/rejected": -295.19293212890625,
-    "eval_loss": 0.6342188119888306,
-    "eval_rewards/accuracies": 0.6579925417900085,
-    "eval_rewards/chosen": -1.8568360805511475,
-    "eval_rewards/margins": 0.4635196328163147,
-    "eval_rewards/rejected": -2.3203558921813965,
-    "eval_runtime": 354.8785,
+    "eval_logits/chosen": -2.5584018230438232,
+    "eval_logits/rejected": -2.5515997409820557,
+    "eval_logps/chosen": -108.46733856201172,
+    "eval_logps/rejected": -123.28102111816406,
+    "eval_loss": 0.6549462676048279,
+    "eval_rewards/accuracies": 0.6194238066673279,
+    "eval_rewards/chosen": -0.49755439162254333,
+    "eval_rewards/margins": 0.10345453023910522,
+    "eval_rewards/rejected": -0.6010088920593262,
+    "eval_runtime": 359.6814,
     "eval_samples": 4304,
-    "eval_samples_per_second": 12.128,
-    "eval_steps_per_second": 1.516
+    "eval_samples_per_second": 11.966,
+    "eval_steps_per_second": 1.496
 }

runs/Jun05_01-08-42_poseidon/events.out.tfevents.1717641866.poseidon.2879567.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3dd10146b2a8ae603d74ba6f00f02bff077ad61786447629b8e26139225e0b8
+size 828