Model save

README.md

@@ -1,15 +1,10 @@
 ---
-base_model: data/gemma-2b
 tags:
-- alignment-handbook
-- trl
-- sft
-- generated_from_trainer
 - trl
 - sft
 - generated_from_trainer
 datasets:
-- argilla/dpo-mix-7k
+- generator
 model-index:
 - name: zephyr-2b-gemma-dft-debug
   results: []
@@ -20,9 +15,9 @@ should probably proofread and complete it, then remove this comment. -->
 
 # zephyr-2b-gemma-dft-debug
 
-This model is a fine-tuned version of [data/gemma-2b](https://huggingface.co/data/gemma-2b) on the argilla/dpo-mix-7k dataset.
+This model was trained from scratch on the generator dataset.
 It achieves the following results on the evaluation set:
-- Loss: 5.7530
+- Loss: 5.7648
 
 ## Model description
 
@@ -56,7 +51,7 @@ The following hyperparameters were used during training:
 
 | Training Loss | Epoch  | Step | Validation Loss |
 |:-------------:|:------:|:----:|:---------------:|
-| 5.7909        | 0.9982 | 270  | 5.7530          |
+| 5.8055        | 0.9982 | 270  | 5.7648          |
 
 
 ### Framework versions

all_results.json

@@ -1,14 +1,9 @@
 {
     "epoch": 0.9981515711645101,
-    "eval_loss": 5.75302791595459,
-    "eval_runtime": 422.4972,
-    "eval_samples": 750,
-    "eval_samples_per_second": 2.298,
-    "eval_steps_per_second": 0.289,
     "total_flos": 5.260333472022528e+16,
-    "train_loss": 7.3803352285314485,
-    "train_runtime": 2516.7229,
+    "train_loss": 7.369897298459653,
+    "train_runtime": 2195.5957,
     "train_samples": 6750,
-    "train_samples_per_second": 3.438,
-    "train_steps_per_second": 0.107
+    "train_samples_per_second": 3.941,
+    "train_steps_per_second": 0.123
 }

config.json

@@ -24,6 +24,6 @@
   "rope_theta": 10000.0,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.43.3",
-  "use_cache": true,
+  "use_cache": false,
   "vocab_size": 256000
 }

model-00001-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3cc245fefab97d950d4759fe13ab557dd93524fd58102bb040e57fbe13709f50
+oid sha256:302478d12c555b4bbed9c86317b38cad6aee12e98984b44a35510ea75c5df6d7
 size 4945242264

model-00002-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4c9ca3299801e9176ca770f65c661ccc72a5b519bf063d407a2606c996c2efcd
+oid sha256:a5919bd0e78fe39414c69b98ae088404ebc6c7007c7c4d18b927edbe5d5c217f
 size 67121608

runs/Aug02_11-52-50_ale-distillm-8-0-0/events.out.tfevents.1722595982.ale-distillm-8-0-0.9078.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69e554244db879b5ab4de86d2113d3b1ae9168e943b825257008aff373713372
+size 7183

runs/Aug02_11-59-21_ale-distillm-8-0-0/events.out.tfevents.1722596374.ale-distillm-8-0-0.10851.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c059071fa523a9a278afcf726b9909363182d2461df2f6cd359627e03a57f072
+size 17239

train_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 0.9981515711645101,
     "total_flos": 5.260333472022528e+16,
-    "train_loss": 7.3803352285314485,
-    "train_runtime": 2516.7229,
+    "train_loss": 7.369897298459653,
+    "train_runtime": 2195.5957,
     "train_samples": 6750,
-    "train_samples_per_second": 3.438,
-    "train_steps_per_second": 0.107
+    "train_samples_per_second": 3.941,
+    "train_steps_per_second": 0.123
 }

trainer_state.json

@@ -10,398 +10,398 @@
   "log_history": [
     {
       "epoch": 0.018484288354898338,
-      "grad_norm": 422.0,
+      "grad_norm": 314.0,
       "learning_rate": 3.7037037037037037e-06,
-      "loss": 20.7086,
+      "loss": 20.7105,
       "step": 5
     },
     {
       "epoch": 0.036968576709796676,
-      "grad_norm": 185.0,
+      "grad_norm": 177.0,
       "learning_rate": 7.4074074074074075e-06,
-      "loss": 19.253,
+      "loss": 19.245,
       "step": 10
     },
     {
       "epoch": 0.05545286506469501,
-      "grad_norm": 147.0,
+      "grad_norm": 189.0,
       "learning_rate": 1.1111111111111113e-05,
-      "loss": 16.7138,
+      "loss": 16.7313,
       "step": 15
     },
     {
       "epoch": 0.07393715341959335,
-      "grad_norm": 62.75,
+      "grad_norm": 85.0,
       "learning_rate": 1.4814814814814815e-05,
-      "loss": 14.0999,
+      "loss": 14.1818,
       "step": 20
     },
     {
       "epoch": 0.09242144177449169,
-      "grad_norm": 46.25,
+      "grad_norm": 54.25,
       "learning_rate": 1.851851851851852e-05,
-      "loss": 11.9456,
+      "loss": 11.905,
       "step": 25
     },
     {
       "epoch": 0.11090573012939002,
-      "grad_norm": 49.75,
+      "grad_norm": 91.0,
       "learning_rate": 1.9992479525042305e-05,
-      "loss": 10.3044,
+      "loss": 10.2358,
       "step": 30
     },
     {
       "epoch": 0.12939001848428835,
-      "grad_norm": 88.5,
+      "grad_norm": 60.0,
       "learning_rate": 1.9946562024066018e-05,
-      "loss": 9.602,
+      "loss": 9.522,
       "step": 35
     },
     {
       "epoch": 0.1478743068391867,
-      "grad_norm": 49.25,
+      "grad_norm": 22.125,
       "learning_rate": 1.9859096633447965e-05,
-      "loss": 9.1612,
+      "loss": 9.0956,
       "step": 40
     },
     {
       "epoch": 0.16635859519408502,
-      "grad_norm": 114.0,
+      "grad_norm": 98.5,
       "learning_rate": 1.973044870579824e-05,
-      "loss": 8.5549,
+      "loss": 8.5045,
       "step": 45
     },
     {
       "epoch": 0.18484288354898337,
-      "grad_norm": 54.75,
+      "grad_norm": 22.5,
       "learning_rate": 1.95611556177388e-05,
-      "loss": 8.3168,
+      "loss": 8.2548,
       "step": 50
     },
     {
       "epoch": 0.2033271719038817,
-      "grad_norm": 16.875,
+      "grad_norm": 46.5,
       "learning_rate": 1.93519245252219e-05,
-      "loss": 7.8042,
+      "loss": 7.7123,
       "step": 55
     },
     {
       "epoch": 0.22181146025878004,
-      "grad_norm": 55.75,
+      "grad_norm": 34.75,
       "learning_rate": 1.9103629409661468e-05,
-      "loss": 7.4335,
+      "loss": 7.3372,
       "step": 60
     },
     {
       "epoch": 0.24029574861367836,
-      "grad_norm": 17.5,
+      "grad_norm": 45.75,
       "learning_rate": 1.881730742721608e-05,
-      "loss": 7.4047,
+      "loss": 7.2639,
       "step": 65
     },
     {
       "epoch": 0.2587800369685767,
-      "grad_norm": 52.5,
+      "grad_norm": 99.5,
       "learning_rate": 1.8494154576472976e-05,
-      "loss": 7.0682,
+      "loss": 6.9563,
       "step": 70
     },
     {
       "epoch": 0.27726432532347506,
-      "grad_norm": 69.5,
+      "grad_norm": 58.75,
       "learning_rate": 1.8135520702629677e-05,
-      "loss": 7.1519,
+      "loss": 6.8706,
       "step": 75
     },
     {
       "epoch": 0.2957486136783734,
-      "grad_norm": 23.125,
+      "grad_norm": 189.0,
       "learning_rate": 1.7742903859041324e-05,
-      "loss": 6.9284,
+      "loss": 6.8386,
       "step": 80
     },
     {
       "epoch": 0.3142329020332717,
-      "grad_norm": 17.25,
+      "grad_norm": 31.0,
       "learning_rate": 1.7317944049686125e-05,
-      "loss": 6.723,
+      "loss": 6.7403,
       "step": 85
     },
     {
       "epoch": 0.33271719038817005,
-      "grad_norm": 103.5,
+      "grad_norm": 55.25,
       "learning_rate": 1.686241637868734e-05,
-      "loss": 6.6468,
+      "loss": 6.554,
       "step": 90
     },
     {
       "epoch": 0.3512014787430684,
-      "grad_norm": 17.625,
+      "grad_norm": 58.5,
       "learning_rate": 1.637822363550706e-05,
-      "loss": 6.5427,
+      "loss": 6.5162,
       "step": 95
     },
     {
       "epoch": 0.36968576709796674,
-      "grad_norm": 15.6875,
+      "grad_norm": 18.875,
       "learning_rate": 1.586738834678418e-05,
-      "loss": 6.3795,
+      "loss": 6.3662,
       "step": 100
     },
     {
       "epoch": 0.38817005545286504,
-      "grad_norm": 17.875,
+      "grad_norm": 25.125,
       "learning_rate": 1.5332044328016916e-05,
-      "loss": 6.4565,
+      "loss": 6.4888,
       "step": 105
     },
     {
       "epoch": 0.4066543438077634,
-      "grad_norm": 68.0,
+      "grad_norm": 17.0,
       "learning_rate": 1.4774427770379492e-05,
-      "loss": 6.416,
+      "loss": 6.5204,
       "step": 110
     },
     {
       "epoch": 0.42513863216266173,
-      "grad_norm": 71.5,
+      "grad_norm": 125.5,
       "learning_rate": 1.4196867899904292e-05,
-      "loss": 6.2522,
+      "loss": 6.3338,
       "step": 115
     },
     {
       "epoch": 0.4436229205175601,
-      "grad_norm": 28.875,
+      "grad_norm": 17.5,
       "learning_rate": 1.3601777248047105e-05,
-      "loss": 6.0918,
+      "loss": 6.0951,
       "step": 120
     },
     {
       "epoch": 0.46210720887245843,
-      "grad_norm": 25.75,
+      "grad_norm": 10.0,
       "learning_rate": 1.2991641574276419e-05,
-      "loss": 6.0943,
+      "loss": 6.1339,
       "step": 125
     },
     {
       "epoch": 0.4805914972273567,
-      "grad_norm": 24.375,
+      "grad_norm": 24.0,
       "learning_rate": 1.2369009482781191e-05,
-      "loss": 6.0615,
+      "loss": 6.1337,
       "step": 130
     },
     {
       "epoch": 0.49907578558225507,
-      "grad_norm": 53.0,
+      "grad_norm": 40.75,
       "learning_rate": 1.1736481776669307e-05,
-      "loss": 5.9591,
+      "loss": 5.997,
       "step": 135
     },
     {
       "epoch": 0.5175600739371534,
-      "grad_norm": 93.5,
+      "grad_norm": 61.5,
       "learning_rate": 1.1096700594125318e-05,
-      "loss": 6.0178,
+      "loss": 6.0472,
       "step": 140
     },
     {
       "epoch": 0.5360443622920518,
-      "grad_norm": 14.0625,
+      "grad_norm": 18.125,
       "learning_rate": 1.0452338371907065e-05,
-      "loss": 5.8597,
+      "loss": 5.8697,
       "step": 145
     },
     {
       "epoch": 0.5545286506469501,
-      "grad_norm": 102.0,
+      "grad_norm": 85.0,
       "learning_rate": 9.806086682281759e-06,
-      "loss": 5.8942,
+      "loss": 5.9026,
       "step": 150
     },
     {
       "epoch": 0.5730129390018485,
-      "grad_norm": 30.125,
+      "grad_norm": 28.0,
       "learning_rate": 9.160644990030932e-06,
-      "loss": 5.9999,
+      "loss": 6.016,
       "step": 155
     },
     {
       "epoch": 0.5914972273567468,
-      "grad_norm": 23.25,
+      "grad_norm": 12.75,
       "learning_rate": 8.518709376487515e-06,
-      "loss": 5.9019,
+      "loss": 5.9233,
       "step": 160
     },
     {
       "epoch": 0.609981515711645,
-      "grad_norm": 56.0,
+      "grad_norm": 77.5,
       "learning_rate": 7.882961277705897e-06,
-      "loss": 5.8274,
+      "loss": 5.8483,
       "step": 165
     },
     {
       "epoch": 0.6284658040665434,
-      "grad_norm": 10.1875,
+      "grad_norm": 27.25,
       "learning_rate": 7.256056283806987e-06,
-      "loss": 5.9167,
+      "loss": 5.9308,
       "step": 170
     },
     {
       "epoch": 0.6469500924214417,
-      "grad_norm": 26.75,
+      "grad_norm": 58.0,
       "learning_rate": 6.640613046284581e-06,
-      "loss": 5.9097,
+      "loss": 5.9355,
       "step": 175
     },
     {
       "epoch": 0.6654343807763401,
-      "grad_norm": 11.5625,
+      "grad_norm": 31.125,
       "learning_rate": 6.039202339608432e-06,
-      "loss": 5.6327,
+      "loss": 5.6434,
       "step": 180
     },
     {
       "epoch": 0.6839186691312384,
-      "grad_norm": 26.75,
+      "grad_norm": 25.0,
       "learning_rate": 5.454336322814995e-06,
-      "loss": 5.6804,
+      "loss": 5.6888,
       "step": 185
     },
     {
       "epoch": 0.7024029574861368,
-      "grad_norm": 30.375,
+      "grad_norm": 34.75,
       "learning_rate": 4.888458045941269e-06,
-      "loss": 5.7316,
+      "loss": 5.7455,
       "step": 190
     },
     {
       "epoch": 0.7208872458410351,
-      "grad_norm": 39.5,
+      "grad_norm": 23.375,
       "learning_rate": 4.343931245134616e-06,
-      "loss": 5.7299,
+      "loss": 5.7362,
       "step": 195
     },
     {
       "epoch": 0.7393715341959335,
-      "grad_norm": 11.75,
+      "grad_norm": 12.8125,
       "learning_rate": 3.823030469065431e-06,
-      "loss": 5.6982,
+      "loss": 5.6849,
       "step": 200
     },
     {
       "epoch": 0.7578558225508318,
-      "grad_norm": 20.625,
+      "grad_norm": 24.875,
       "learning_rate": 3.3279315778858034e-06,
       "loss": 5.862,
       "step": 205
     },
     {
       "epoch": 0.7763401109057301,
-      "grad_norm": 50.75,
+      "grad_norm": 36.0,
       "learning_rate": 2.8607026544210115e-06,
-      "loss": 5.7507,
+      "loss": 5.7541,
       "step": 210
     },
     {
       "epoch": 0.7948243992606284,
-      "grad_norm": 49.75,
+      "grad_norm": 38.25,
       "learning_rate": 2.423295365558821e-06,
-      "loss": 5.7101,
+      "loss": 5.7035,
       "step": 215
     },
     {
       "epoch": 0.8133086876155268,
-      "grad_norm": 14.4375,
+      "grad_norm": 19.5,
       "learning_rate": 2.01753680992107e-06,
-      "loss": 5.8428,
+      "loss": 5.8542,
       "step": 220
     },
     {
       "epoch": 0.8317929759704251,
-      "grad_norm": 34.75,
+      "grad_norm": 33.25,
       "learning_rate": 1.6451218858706374e-06,
-      "loss": 5.7162,
+      "loss": 5.7262,
       "step": 225
     },
     {
       "epoch": 0.8502772643253235,
       "grad_norm": 47.0,
       "learning_rate": 1.307606211733522e-06,
-      "loss": 5.6621,
+      "loss": 5.6701,
       "step": 230
     },
     {
       "epoch": 0.8687615526802218,
-      "grad_norm": 12.875,
+      "grad_norm": 15.3125,
       "learning_rate": 1.0063996278090704e-06,
-      "loss": 5.6616,
+      "loss": 5.6769,
       "step": 235
     },
     {
       "epoch": 0.8872458410351202,
-      "grad_norm": 13.1875,
+      "grad_norm": 16.0,
       "learning_rate": 7.427603073110967e-07,
-      "loss": 5.8559,
+      "loss": 5.868,
       "step": 240
     },
     {
       "epoch": 0.9057301293900185,
-      "grad_norm": 7.96875,
+      "grad_norm": 7.71875,
       "learning_rate": 5.177895008392353e-07,
-      "loss": 5.7149,
+      "loss": 5.7182,
       "step": 245
     },
     {
       "epoch": 0.9242144177449169,
-      "grad_norm": 61.75,
+      "grad_norm": 62.5,
       "learning_rate": 3.3242693633337986e-07,
-      "loss": 5.9011,
+      "loss": 5.9037,
       "step": 250
     },
     {
       "epoch": 0.9426987060998152,
-      "grad_norm": 13.625,
+      "grad_norm": 12.875,
       "learning_rate": 1.874468937261531e-07,
-      "loss": 5.685,
+      "loss": 5.6952,
       "step": 255
     },
     {
       "epoch": 0.9611829944547134,
-      "grad_norm": 15.6875,
+      "grad_norm": 16.875,
       "learning_rate": 8.345497068998897e-08,
-      "loss": 5.7765,
+      "loss": 5.7802,
       "step": 260
     },
     {
       "epoch": 0.9796672828096118,
-      "grad_norm": 18.5,
+      "grad_norm": 13.75,
       "learning_rate": 2.088555298867978e-08,
-      "loss": 5.734,
+      "loss": 5.7397,
       "step": 265
     },
     {
       "epoch": 0.9981515711645101,
-      "grad_norm": 28.25,
+      "grad_norm": 26.625,
       "learning_rate": 0.0,
-      "loss": 5.7909,
+      "loss": 5.8055,
       "step": 270
     },
     {
       "epoch": 0.9981515711645101,
-      "eval_loss": 5.75302791595459,
-      "eval_runtime": 376.2306,
-      "eval_samples_per_second": 2.581,
-      "eval_steps_per_second": 0.324,
+      "eval_loss": 5.764816761016846,
+      "eval_runtime": 319.0279,
+      "eval_samples_per_second": 3.044,
+      "eval_steps_per_second": 0.382,
       "step": 270
     },
     {
       "epoch": 0.9981515711645101,
       "step": 270,
       "total_flos": 5.260333472022528e+16,
-      "train_loss": 7.3803352285314485,
-      "train_runtime": 2516.7229,
-      "train_samples_per_second": 3.438,
-      "train_steps_per_second": 0.107
+      "train_loss": 7.369897298459653,
+      "train_runtime": 2195.5957,
+      "train_samples_per_second": 3.941,
+      "train_steps_per_second": 0.123
     }
   ],
   "logging_steps": 5,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:aa1e4059799b3fd2bc19d9b9bdbb3a0911a6dc0077207bdb4e02f055fcee2618
-size 5368
+oid sha256:94f28c170a957c643092ce3557ce1fd1db0ab541b074c9aab5a4e5efa0f58622
+size 5496