Model save

README.md

@@ -2,15 +2,9 @@
 license: mit
 base_model: HuggingFaceH4/mistral-7b-sft-beta
 tags:
-- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
-- trl
-- dpo
-- generated_from_trainer
-datasets:
-- HuggingFaceH4/ultrafeedback_binarized
 model-index:
 - name: 0.01_version_debug_iter_1
   results: []
@@ -21,7 +15,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # 0.01_version_debug_iter_1
 
-This model is a fine-tuned version of [HuggingFaceH4/mistral-7b-sft-beta](https://huggingface.co/HuggingFaceH4/mistral-7b-sft-beta) on the HuggingFaceH4/ultrafeedback_binarized dataset.
+This model is a fine-tuned version of [HuggingFaceH4/mistral-7b-sft-beta](https://huggingface.co/HuggingFaceH4/mistral-7b-sft-beta) on the None dataset.
 
 ## Model description
 

all_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 0.9874476987447699,
     "total_flos": 0.0,
-    "train_loss": 0.6490306530968618,
-    "train_runtime": 1625.865,
+    "train_loss": 0.649040699005127,
+    "train_runtime": 1650.5799,
     "train_samples": 15283,
-    "train_samples_per_second": 9.4,
+    "train_samples_per_second": 9.259,
     "train_steps_per_second": 0.036
 }

config.json

@@ -21,6 +21,6 @@
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.40.2",
-  "use_cache": true,
+  "use_cache": false,
   "vocab_size": 32000
 }

model-00001-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6d7e545fb262eaf15fc7f3a6a82ad32c4cd811da1994fbcfb735140575123a64
+oid sha256:7d69774d27b6ed3d430faf88c75361b5d47ea11f8bf0de9458eab10b5ff4718f
 size 4943162336

model-00002-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d937ee04a9b76a0b664b470c7f14f998531b48d727897042ce341cac7de6f363
+oid sha256:0e76569195bfebe00c70ab21110ce8611f3b2d4ee945279a8f9b4921e7104758
 size 4999819336

model-00003-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:33805bbe468c389a0c9e65fc0e6f57f40fa04e262857550ac843a8cbfb13f1bd
+oid sha256:e8cdd165c611cc4a0fc6eadb83e41ae18bece4f08bc5fe5e96ea06ef31d2813a
 size 4540516344

train_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 0.9874476987447699,
     "total_flos": 0.0,
-    "train_loss": 0.6490306530968618,
-    "train_runtime": 1625.865,
+    "train_loss": 0.649040699005127,
+    "train_runtime": 1650.5799,
     "train_samples": 15283,
-    "train_samples_per_second": 9.4,
+    "train_samples_per_second": 9.259,
     "train_steps_per_second": 0.036
 }

trainer_state.json

@@ -10,7 +10,7 @@
   "log_history": [
     {
       "epoch": 0.016736401673640166,
-      "grad_norm": 7.773013789330872,
+      "grad_norm": 7.7724267324991665,
       "learning_rate": 8.333333333333333e-08,
       "logits/chosen": -2.807276487350464,
       "logits/rejected": -2.7759768962860107,
@@ -25,86 +25,86 @@
     },
     {
       "epoch": 0.16736401673640167,
-      "grad_norm": 7.444333838896045,
+      "grad_norm": 7.45652629723192,
       "learning_rate": 4.930057285201027e-07,
-      "logits/chosen": -2.755530834197998,
-      "logits/rejected": -2.7461330890655518,
-      "logps/chosen": -271.9606628417969,
-      "logps/rejected": -260.7101745605469,
-      "loss": 0.6915,
-      "rewards/accuracies": 0.5208333134651184,
-      "rewards/chosen": 0.005092799663543701,
-      "rewards/margins": 0.003585505299270153,
-      "rewards/rejected": 0.00150729448068887,
+      "logits/chosen": -2.7555408477783203,
+      "logits/rejected": -2.746152400970459,
+      "logps/chosen": -271.9476013183594,
+      "logps/rejected": -260.6995849609375,
+      "loss": 0.6914,
+      "rewards/accuracies": 0.5104166865348816,
+      "rewards/chosen": 0.0052236998453736305,
+      "rewards/margins": 0.003610293846577406,
+      "rewards/rejected": 0.0016134059987962246,
       "step": 10
     },
     {
       "epoch": 0.33472803347280333,
-      "grad_norm": 6.714535781496835,
+      "grad_norm": 6.711189239787838,
       "learning_rate": 4.187457503795526e-07,
-      "logits/chosen": -2.787006378173828,
-      "logits/rejected": -2.768144130706787,
-      "logps/chosen": -261.48565673828125,
-      "logps/rejected": -250.9136199951172,
-      "loss": 0.6745,
-      "rewards/accuracies": 0.703125,
-      "rewards/chosen": 0.031960152089595795,
-      "rewards/margins": 0.04067179188132286,
-      "rewards/rejected": -0.008711638860404491,
+      "logits/chosen": -2.786916971206665,
+      "logits/rejected": -2.768092632293701,
+      "logps/chosen": -261.447265625,
+      "logps/rejected": -250.86181640625,
+      "loss": 0.6746,
+      "rewards/accuracies": 0.7093750238418579,
+      "rewards/chosen": 0.032344214618206024,
+      "rewards/margins": 0.04053739458322525,
+      "rewards/rejected": -0.008193179033696651,
       "step": 20
     },
     {
       "epoch": 0.502092050209205,
-      "grad_norm": 7.372509129721545,
+      "grad_norm": 7.388933346143141,
       "learning_rate": 2.8691164100062034e-07,
-      "logits/chosen": -2.7877068519592285,
-      "logits/rejected": -2.768649101257324,
-      "logps/chosen": -293.81915283203125,
-      "logps/rejected": -254.2086639404297,
-      "loss": 0.6487,
-      "rewards/accuracies": 0.6781250238418579,
-      "rewards/chosen": -0.023630935698747635,
-      "rewards/margins": 0.12687507271766663,
-      "rewards/rejected": -0.15050600469112396,
+      "logits/chosen": -2.787904739379883,
+      "logits/rejected": -2.768831729888916,
+      "logps/chosen": -293.8032531738281,
+      "logps/rejected": -254.18838500976562,
+      "loss": 0.6486,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.02347174473106861,
+      "rewards/margins": 0.12683136761188507,
+      "rewards/rejected": -0.15030309557914734,
       "step": 30
     },
     {
       "epoch": 0.6694560669456067,
-      "grad_norm": 8.124696915257744,
+      "grad_norm": 8.184184168957682,
       "learning_rate": 1.4248369943086995e-07,
-      "logits/chosen": -2.760054111480713,
-      "logits/rejected": -2.7374792098999023,
-      "logps/chosen": -264.1077575683594,
-      "logps/rejected": -255.5675048828125,
-      "loss": 0.6354,
+      "logits/chosen": -2.7605953216552734,
+      "logits/rejected": -2.738043785095215,
+      "logps/chosen": -264.13555908203125,
+      "logps/rejected": -255.6201629638672,
+      "loss": 0.6353,
       "rewards/accuracies": 0.6875,
-      "rewards/chosen": -0.11188302934169769,
-      "rewards/margins": 0.1606856733560562,
-      "rewards/rejected": -0.2725687325000763,
+      "rewards/chosen": -0.11216118186712265,
+      "rewards/margins": 0.16093404591083527,
+      "rewards/rejected": -0.2730952203273773,
       "step": 40
     },
     {
       "epoch": 0.8368200836820083,
-      "grad_norm": 8.573481060543557,
+      "grad_norm": 8.554193147799957,
       "learning_rate": 3.473909705816111e-08,
-      "logits/chosen": -2.769763946533203,
-      "logits/rejected": -2.74979305267334,
-      "logps/chosen": -279.1478576660156,
-      "logps/rejected": -287.2943420410156,
-      "loss": 0.6202,
-      "rewards/accuracies": 0.628125011920929,
-      "rewards/chosen": -0.22139184176921844,
-      "rewards/margins": 0.13642652332782745,
-      "rewards/rejected": -0.3578183948993683,
+      "logits/chosen": -2.7702176570892334,
+      "logits/rejected": -2.7502920627593994,
+      "logps/chosen": -279.1240234375,
+      "logps/rejected": -287.3242492675781,
+      "loss": 0.6201,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.22115378081798553,
+      "rewards/margins": 0.1369638293981552,
+      "rewards/rejected": -0.35811761021614075,
       "step": 50
     },
     {
       "epoch": 0.9874476987447699,
       "step": 59,
       "total_flos": 0.0,
-      "train_loss": 0.6490306530968618,
-      "train_runtime": 1625.865,
-      "train_samples_per_second": 9.4,
+      "train_loss": 0.649040699005127,
+      "train_runtime": 1650.5799,
+      "train_samples_per_second": 9.259,
       "train_steps_per_second": 0.036
     }
   ],

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:94f057f319a1eb7cb3f08b8445e8151e5adc867fa1455dcc764d89ba7a5f948e
+oid sha256:85d59930b4abf212a8998254052a9cb7a5e88274d98802ae2794df3a3039d3f8
 size 6328