End of training

README.md

@@ -1,11 +1,13 @@
 ---
 library_name: peft
 tags:
+- alignment-handbook
 - trl
 - dpo
-- alignment-handbook
 - generated_from_trainer
 base_model: NbAiLab/nb-gpt-j-6B-v2
+datasets:
+- hugodk-sch/aftonposten_title_prefs
 model-index:
 - name: aftonposten-6b-align-scan
   results: []
@@ -16,17 +18,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # aftonposten-6b-align-scan
 
-This model is a fine-tuned version of [NbAiLab/nb-gpt-j-6B-v2](https://huggingface.co/NbAiLab/nb-gpt-j-6B-v2) on the None dataset.
+This model is a fine-tuned version of [data/ap-gpt-j-6b-sft-qlora-04-08](https://huggingface.co/data/ap-gpt-j-6b-sft-qlora-04-08) on the hugodk-sch/aftonposten_title_prefs dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.4747
-- Rewards/chosen: 0.2822
-- Rewards/rejected: 0.1614
-- Rewards/accuracies: 0.5569
-- Rewards/margins: 0.1208
-- Logps/rejected: -37.2860
-- Logps/chosen: -33.6314
-- Logits/rejected: -2.1560
-- Logits/chosen: -2.1607
+- Loss: 0.4763
+- Rewards/chosen: 0.2774
+- Rewards/rejected: 0.1670
+- Rewards/accuracies: 0.5685
+- Rewards/margins: 0.1104
+- Logps/rejected: -37.2781
+- Logps/chosen: -33.6382
+- Logits/rejected: -2.1561
+- Logits/chosen: -2.1608
 
 ## Model description
 

all_results.json

@@ -1,18 +1,18 @@
 {
     "epoch": 4.0,
-    "eval_logits/chosen": -2.2319846153259277,
-    "eval_logits/rejected": -2.227167844772339,
-    "eval_logps/chosen": -33.76113510131836,
-    "eval_logps/rejected": -37.29436492919922,
-    "eval_loss": 0.49125906825065613,
-    "eval_rewards/accuracies": 0.545265793800354,
-    "eval_rewards/chosen": 0.19139321148395538,
-    "eval_rewards/margins": 0.03581271693110466,
-    "eval_rewards/rejected": 0.15558050572872162,
-    "eval_runtime": 145.6689,
+    "eval_logits/chosen": -2.1607675552368164,
+    "eval_logits/rejected": -2.156055450439453,
+    "eval_logps/chosen": -33.63822555541992,
+    "eval_logps/rejected": -37.278076171875,
+    "eval_loss": 0.4762561619281769,
+    "eval_rewards/accuracies": 0.5685215592384338,
+    "eval_rewards/chosen": 0.27742961049079895,
+    "eval_rewards/margins": 0.11044850200414658,
+    "eval_rewards/rejected": 0.16698110103607178,
+    "eval_runtime": 145.2249,
     "eval_samples": 343,
-    "eval_samples_per_second": 2.355,
-    "eval_steps_per_second": 0.295,
+    "eval_samples_per_second": 2.362,
+    "eval_steps_per_second": 0.296,
     "train_loss": 0.17319794591371115,
     "train_runtime": 10768.0107,
     "train_samples": 3079,

config.json

@@ -18,6 +18,8 @@
   "n_layer": 28,
   "n_positions": 2048,
   "quantization_config": {
+    "_load_in_4bit": true,
+    "_load_in_8bit": false,
     "bnb_4bit_compute_dtype": "float16",
     "bnb_4bit_quant_type": "nf4",
     "bnb_4bit_use_double_quant": false,
@@ -45,7 +47,7 @@
   },
   "tie_word_embeddings": false,
   "tokenizer_class": "GPT2Tokenizer",
-  "transformers_version": "4.37.2",
+  "transformers_version": "4.39.0.dev0",
   "use_cache": true,
   "vocab_size": 50400
 }

eval_results.json

@@ -1,16 +1,16 @@
 {
-    "epoch": 1.0,
-    "eval_logits/chosen": -2.2319846153259277,
-    "eval_logits/rejected": -2.227167844772339,
-    "eval_logps/chosen": -33.76113510131836,
-    "eval_logps/rejected": -37.29436492919922,
-    "eval_loss": 0.49125906825065613,
-    "eval_rewards/accuracies": 0.545265793800354,
-    "eval_rewards/chosen": 0.19139321148395538,
-    "eval_rewards/margins": 0.03581271693110466,
-    "eval_rewards/rejected": 0.15558050572872162,
-    "eval_runtime": 145.6689,
+    "epoch": 4.0,
+    "eval_logits/chosen": -2.1607675552368164,
+    "eval_logits/rejected": -2.156055450439453,
+    "eval_logps/chosen": -33.63822555541992,
+    "eval_logps/rejected": -37.278076171875,
+    "eval_loss": 0.4762561619281769,
+    "eval_rewards/accuracies": 0.5685215592384338,
+    "eval_rewards/chosen": 0.27742961049079895,
+    "eval_rewards/margins": 0.11044850200414658,
+    "eval_rewards/rejected": 0.16698110103607178,
+    "eval_runtime": 145.2249,
     "eval_samples": 343,
-    "eval_samples_per_second": 2.355,
-    "eval_steps_per_second": 0.295
+    "eval_samples_per_second": 2.362,
+    "eval_steps_per_second": 0.296
 }