End of training

README.md

@@ -2,10 +2,16 @@
 license: apache-2.0
 base_model: mistralai/Mistral-7B-v0.1
 tags:
+- alignment-handbook
+- trl
+- orpo
+- generated_from_trainer
 - trl
 - orpo
 - alignment-handbook
 - generated_from_trainer
+datasets:
+- HuggingFaceH4/ultrafeedback_binarized
 model-index:
 - name: zephyr-7b-sft-full-orpo
   results: []
@@ -17,20 +23,20 @@ should probably proofread and complete it, then remove this comment. -->
 [<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/statking/huggingface/runs/90a8kp39)
 # zephyr-7b-sft-full-orpo
 
-This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on an unknown dataset.
+This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on the HuggingFaceH4/ultrafeedback_binarized dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.4714
-- Rewards/chosen: -0.0357
-- Rewards/rejected: -0.0466
-- Rewards/accuracies: 0.6528
-- Rewards/margins: 0.0109
-- Logps/rejected: -0.9324
-- Logps/chosen: -0.7143
-- Logits/rejected: -2.9543
-- Logits/chosen: -2.9692
-- Nll Loss: 0.4361
-- Log Odds Ratio: -0.6245
-- Log Odds Chosen: 0.3669
+- Loss: 0.4701
+- Rewards/chosen: -0.0364
+- Rewards/rejected: -0.0499
+- Rewards/accuracies: 0.6587
+- Rewards/margins: 0.0135
+- Logps/rejected: -0.9978
+- Logps/chosen: -0.7282
+- Logits/rejected: -2.9263
+- Logits/chosen: -2.9434
+- Nll Loss: 0.4357
+- Log Odds Ratio: -0.6093
+- Log Odds Chosen: 0.4456
 
 ## Model description
 

all_results.json

@@ -1,21 +1,21 @@
 {
     "epoch": 0.9994756161510225,
-    "eval_log_odds_chosen": 1.220078706741333,
-    "eval_log_odds_ratio": -1.047989010810852,
-    "eval_logits/chosen": -2.849764585494995,
-    "eval_logits/rejected": -2.8669533729553223,
-    "eval_logps/chosen": -2.7812891006469727,
-    "eval_logps/rejected": -3.8601787090301514,
-    "eval_loss": 1.3770909309387207,
-    "eval_nll_loss": 1.3531930446624756,
-    "eval_rewards/accuracies": 0.6527777910232544,
-    "eval_rewards/chosen": -0.13906444609165192,
-    "eval_rewards/margins": 0.05394447594881058,
-    "eval_rewards/rejected": -0.1930089294910431,
-    "eval_runtime": 135.9849,
+    "eval_log_odds_chosen": 0.4455896019935608,
+    "eval_log_odds_ratio": -0.6093403697013855,
+    "eval_logits/chosen": -2.943448543548584,
+    "eval_logits/rejected": -2.9263010025024414,
+    "eval_logps/chosen": -0.7282419204711914,
+    "eval_logps/rejected": -0.9978163242340088,
+    "eval_loss": 0.4700748324394226,
+    "eval_nll_loss": 0.4356931149959564,
+    "eval_rewards/accuracies": 0.658730149269104,
+    "eval_rewards/chosen": -0.03641209378838539,
+    "eval_rewards/margins": 0.013478721491992474,
+    "eval_rewards/rejected": -0.04989081248641014,
+    "eval_runtime": 137.5022,
     "eval_samples": 1994,
-    "eval_samples_per_second": 14.663,
-    "eval_steps_per_second": 0.463,
+    "eval_samples_per_second": 14.502,
+    "eval_steps_per_second": 0.458,
     "total_flos": 0.0,
     "train_loss": 0.5301580581685054,
     "train_runtime": 20737.8205,

config.json

@@ -21,6 +21,6 @@
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.41.0.dev0",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 32000
 }

eval_results.json

@@ -1,19 +1,19 @@
 {
-    "epoch": 2.9984268484530676,
-    "eval_log_odds_chosen": 1.220078706741333,
-    "eval_log_odds_ratio": -1.047989010810852,
-    "eval_logits/chosen": -2.849764585494995,
-    "eval_logits/rejected": -2.8669533729553223,
-    "eval_logps/chosen": -2.7812891006469727,
-    "eval_logps/rejected": -3.8601787090301514,
-    "eval_loss": 1.3770909309387207,
-    "eval_nll_loss": 1.3531930446624756,
-    "eval_rewards/accuracies": 0.6527777910232544,
-    "eval_rewards/chosen": -0.13906444609165192,
-    "eval_rewards/margins": 0.05394447594881058,
-    "eval_rewards/rejected": -0.1930089294910431,
-    "eval_runtime": 135.9849,
+    "epoch": 0.9994756161510225,
+    "eval_log_odds_chosen": 0.4455896019935608,
+    "eval_log_odds_ratio": -0.6093403697013855,
+    "eval_logits/chosen": -2.943448543548584,
+    "eval_logits/rejected": -2.9263010025024414,
+    "eval_logps/chosen": -0.7282419204711914,
+    "eval_logps/rejected": -0.9978163242340088,
+    "eval_loss": 0.4700748324394226,
+    "eval_nll_loss": 0.4356931149959564,
+    "eval_rewards/accuracies": 0.658730149269104,
+    "eval_rewards/chosen": -0.03641209378838539,
+    "eval_rewards/margins": 0.013478721491992474,
+    "eval_rewards/rejected": -0.04989081248641014,
+    "eval_runtime": 137.5022,
     "eval_samples": 1994,
-    "eval_samples_per_second": 14.663,
-    "eval_steps_per_second": 0.463
+    "eval_samples_per_second": 14.502,
+    "eval_steps_per_second": 0.458
 }