End of training

README.md

@@ -3,9 +3,15 @@ library_name: transformers
 license: apache-2.0
 base_model: TinyLlama/TinyLlama-1.1B-Chat-v1.0
 tags:
+- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
+- trl
+- dpo
+- generated_from_trainer
+datasets:
+- data/ui_math
 model-index:
 - name: tinyllama-1.1b-chat-v1.0-ui-math-dpo
   results: []
@@ -16,7 +22,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # tinyllama-1.1b-chat-v1.0-ui-math-dpo
 
-This model is a fine-tuned version of [TinyLlama/TinyLlama-1.1B-Chat-v1.0](https://huggingface.co/TinyLlama/TinyLlama-1.1B-Chat-v1.0) on an unknown dataset.
+This model is a fine-tuned version of [TinyLlama/TinyLlama-1.1B-Chat-v1.0](https://huggingface.co/TinyLlama/TinyLlama-1.1B-Chat-v1.0) on the data/ui_math dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.6894
+- Rewards/chosen: -0.8659
+- Rewards/rejected: -0.8634
+- Rewards/accuracies: 0.625
+- Rewards/margins: -0.0025
+- Logps/rejected: -327.4781
+- Logps/chosen: -342.2401
+- Logits/rejected: -1.8881
+- Logits/chosen: -2.0022
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 0.9966024915062288,
+    "eval_logits/chosen": -2.0022222995758057,
+    "eval_logits/rejected": -1.888062834739685,
+    "eval_logps/chosen": -342.24005126953125,
+    "eval_logps/rejected": -327.4780578613281,
+    "eval_loss": 0.6894429922103882,
+    "eval_rewards/accuracies": 0.625,
+    "eval_rewards/chosen": -0.8659123182296753,
+    "eval_rewards/margins": -0.0024991072714328766,
+    "eval_rewards/rejected": -0.8634131550788879,
+    "eval_runtime": 1.2317,
+    "eval_samples": 100,
+    "eval_samples_per_second": 81.189,
+    "eval_steps_per_second": 0.812,
     "total_flos": 0.0,
     "train_loss": 0.5847668994556774,
     "train_runtime": 2901.4735,

config.json

@@ -24,6 +24,6 @@
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.44.1",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 32000
 }

eval_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 0.9966024915062288,
+    "eval_logits/chosen": -2.0022222995758057,
+    "eval_logits/rejected": -1.888062834739685,
+    "eval_logps/chosen": -342.24005126953125,
+    "eval_logps/rejected": -327.4780578613281,
+    "eval_loss": 0.6894429922103882,
+    "eval_rewards/accuracies": 0.625,
+    "eval_rewards/chosen": -0.8659123182296753,
+    "eval_rewards/margins": -0.0024991072714328766,
+    "eval_rewards/rejected": -0.8634131550788879,
+    "eval_runtime": 1.2317,
+    "eval_samples": 100,
+    "eval_samples_per_second": 81.189,
+    "eval_steps_per_second": 0.812
+}