Model save

README.md

@@ -1,11 +1,8 @@
 ---
 base_model: mistralai/Mistral-7B-Instruct-v0.1
-datasets:
-- generator
 library_name: peft
 license: apache-2.0
 tags:
-- alignment-handbook
 - trl
 - sft
 - generated_from_trainer
@@ -19,9 +16,9 @@ should probably proofread and complete it, then remove this comment. -->
 
 # mistral_cot_true_simple_lora
 
-This model is a fine-tuned version of [mistralai/Mistral-7B-Instruct-v0.1](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1) on the generator dataset.
+This model is a fine-tuned version of [mistralai/Mistral-7B-Instruct-v0.1](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.9395
+- Loss: 0.6070
 
 ## Model description
 
@@ -50,25 +47,16 @@ The following hyperparameters were used during training:
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: cosine
 - lr_scheduler_warmup_ratio: 0.1
-- num_epochs: 15
+- training_steps: 20
 
 ### Training results
 
-| Training Loss | Epoch   | Step | Validation Loss |
-|:-------------:|:-------:|:----:|:---------------:|
-| 1.0104        | 0.8571  | 3    | 0.8812          |
-| 0.8795        | 2.0     | 7    | 0.7078          |
-| 0.6536        | 2.8571  | 10   | 0.6557          |
-| 0.6536        | 4.0     | 14   | 0.6372          |
-| 0.472         | 4.8571  | 17   | 0.6561          |
-| 0.3263        | 6.0     | 21   | 0.6875          |
-| 0.3263        | 6.8571  | 24   | 0.7510          |
-| 0.1683        | 8.0     | 28   | 0.8222          |
-| 0.0699        | 8.8571  | 31   | 0.8685          |
-| 0.0369        | 10.0    | 35   | 0.9062          |
-| 0.0369        | 10.8571 | 38   | 0.9268          |
-| 0.0248        | 12.0    | 42   | 0.9377          |
-| 0.0212        | 12.8571 | 45   | 0.9395          |
+| Training Loss | Epoch | Step | Validation Loss |
+|:-------------:|:-----:|:----:|:---------------:|
+| 0.9142        | 0.4   | 5    | 0.7228          |
+| 0.6834        | 0.8   | 10   | 0.6278          |
+| 0.5454        | 1.2   | 15   | 0.6100          |
+| 0.5243        | 1.6   | 20   | 0.6070          |
 
 
 ### Framework versions

all_results.json

@@ -1,14 +1,9 @@
 {
-    "epoch": 12.857142857142858,
-    "eval_loss": 0.939548134803772,
-    "eval_runtime": 1.1735,
-    "eval_samples": 20,
-    "eval_samples_per_second": 4.261,
-    "eval_steps_per_second": 1.704,
-    "total_flos": 10453480636416.0,
-    "train_loss": 0.2976339568694433,
-    "train_runtime": 339.6314,
+    "epoch": 1.6,
+    "total_flos": 1591906369536.0,
+    "train_loss": 0.6768446683883667,
+    "train_runtime": 82.5503,
     "train_samples": 100,
-    "train_samples_per_second": 1.148,
-    "train_steps_per_second": 0.132
+    "train_samples_per_second": 1.938,
+    "train_steps_per_second": 0.242
 }

train_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 12.857142857142858,
-    "total_flos": 10453480636416.0,
-    "train_loss": 0.2976339568694433,
-    "train_runtime": 339.6314,
+    "epoch": 1.6,
+    "total_flos": 1591906369536.0,
+    "train_loss": 0.6768446683883667,
+    "train_runtime": 82.5503,
     "train_samples": 100,
-    "train_samples_per_second": 1.148,
-    "train_steps_per_second": 0.132
+    "train_samples_per_second": 1.938,
+    "train_steps_per_second": 0.242
 }

trainer_state.json

@@ -1,201 +1,94 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 12.857142857142858,
-  "eval_steps": 500,
-  "global_step": 45,
+  "epoch": 1.6,
+  "eval_steps": 5,
+  "global_step": 20,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.2857142857142857,
-      "grad_norm": 0.7675084033789445,
-      "learning_rate": 4e-05,
-      "loss": 1.0104,
+      "epoch": 0.08,
+      "grad_norm": 0.7334153552614141,
+      "learning_rate": 0.0001,
+      "loss": 1.115,
       "step": 1
     },
     {
-      "epoch": 0.8571428571428571,
-      "eval_loss": 0.8811686635017395,
-      "eval_runtime": 2.3203,
-      "eval_samples_per_second": 2.155,
-      "eval_steps_per_second": 0.862,
-      "step": 3
-    },
-    {
-      "epoch": 1.4285714285714286,
-      "grad_norm": 0.3794777121113468,
-      "learning_rate": 0.0002,
-      "loss": 0.8795,
+      "epoch": 0.4,
+      "grad_norm": 0.5975179564042915,
+      "learning_rate": 0.00018660254037844388,
+      "loss": 0.9142,
       "step": 5
     },
     {
-      "epoch": 2.0,
-      "eval_loss": 0.7077744007110596,
-      "eval_runtime": 1.1217,
-      "eval_samples_per_second": 4.458,
-      "eval_steps_per_second": 1.783,
-      "step": 7
+      "epoch": 0.4,
+      "eval_loss": 0.722756564617157,
+      "eval_runtime": 3.0808,
+      "eval_samples_per_second": 6.492,
+      "eval_steps_per_second": 1.623,
+      "step": 5
     },
     {
-      "epoch": 2.857142857142857,
-      "grad_norm": 0.3530496345436479,
-      "learning_rate": 0.0001923879532511287,
-      "loss": 0.6536,
+      "epoch": 0.8,
+      "grad_norm": 0.4159151923493577,
+      "learning_rate": 0.00011736481776669306,
+      "loss": 0.6834,
       "step": 10
     },
     {
-      "epoch": 2.857142857142857,
-      "eval_loss": 0.6557201743125916,
-      "eval_runtime": 1.1188,
-      "eval_samples_per_second": 4.469,
-      "eval_steps_per_second": 1.788,
+      "epoch": 0.8,
+      "eval_loss": 0.6278184056282043,
+      "eval_runtime": 1.8619,
+      "eval_samples_per_second": 10.742,
+      "eval_steps_per_second": 2.685,
       "step": 10
     },
     {
-      "epoch": 4.0,
-      "eval_loss": 0.6372172236442566,
-      "eval_runtime": 1.1233,
-      "eval_samples_per_second": 4.451,
-      "eval_steps_per_second": 1.781,
-      "step": 14
-    },
-    {
-      "epoch": 4.285714285714286,
-      "grad_norm": 0.26413927917508156,
-      "learning_rate": 0.00017071067811865476,
-      "loss": 0.472,
+      "epoch": 1.2,
+      "grad_norm": 0.40842948448715766,
+      "learning_rate": 3.5721239031346066e-05,
+      "loss": 0.5454,
       "step": 15
     },
     {
-      "epoch": 4.857142857142857,
-      "eval_loss": 0.6561022400856018,
-      "eval_runtime": 1.1268,
-      "eval_samples_per_second": 4.437,
-      "eval_steps_per_second": 1.775,
-      "step": 17
-    },
-    {
-      "epoch": 5.714285714285714,
-      "grad_norm": 0.3618434995437738,
-      "learning_rate": 0.000138268343236509,
-      "loss": 0.3263,
-      "step": 20
-    },
-    {
-      "epoch": 6.0,
-      "eval_loss": 0.687515914440155,
-      "eval_runtime": 1.1222,
-      "eval_samples_per_second": 4.456,
-      "eval_steps_per_second": 1.782,
-      "step": 21
-    },
-    {
-      "epoch": 6.857142857142857,
-      "eval_loss": 0.7509881854057312,
-      "eval_runtime": 1.1228,
-      "eval_samples_per_second": 4.453,
-      "eval_steps_per_second": 1.781,
-      "step": 24
-    },
-    {
-      "epoch": 7.142857142857143,
-      "grad_norm": 0.3781011248380884,
-      "learning_rate": 0.0001,
-      "loss": 0.1683,
-      "step": 25
-    },
-    {
-      "epoch": 8.0,
-      "eval_loss": 0.8222058415412903,
-      "eval_runtime": 1.1218,
-      "eval_samples_per_second": 4.457,
-      "eval_steps_per_second": 1.783,
-      "step": 28
-    },
-    {
-      "epoch": 8.571428571428571,
-      "grad_norm": 0.264128562506645,
-      "learning_rate": 6.173165676349103e-05,
-      "loss": 0.0699,
-      "step": 30
-    },
-    {
-      "epoch": 8.857142857142858,
-      "eval_loss": 0.8684995770454407,
-      "eval_runtime": 1.1229,
-      "eval_samples_per_second": 4.453,
-      "eval_steps_per_second": 1.781,
-      "step": 31
-    },
-    {
-      "epoch": 10.0,
-      "grad_norm": 0.24436285012704104,
-      "learning_rate": 2.9289321881345254e-05,
-      "loss": 0.0369,
-      "step": 35
-    },
-    {
-      "epoch": 10.0,
-      "eval_loss": 0.906228244304657,
-      "eval_runtime": 1.122,
-      "eval_samples_per_second": 4.456,
-      "eval_steps_per_second": 1.782,
-      "step": 35
-    },
-    {
-      "epoch": 10.857142857142858,
-      "eval_loss": 0.9267812967300415,
-      "eval_runtime": 1.1231,
-      "eval_samples_per_second": 4.452,
-      "eval_steps_per_second": 1.781,
-      "step": 38
-    },
-    {
-      "epoch": 11.428571428571429,
-      "grad_norm": 0.13496685015110904,
-      "learning_rate": 7.612046748871327e-06,
-      "loss": 0.0248,
-      "step": 40
-    },
-    {
-      "epoch": 12.0,
-      "eval_loss": 0.9376745223999023,
-      "eval_runtime": 1.1206,
-      "eval_samples_per_second": 4.462,
-      "eval_steps_per_second": 1.785,
-      "step": 42
+      "epoch": 1.2,
+      "eval_loss": 0.6100303530693054,
+      "eval_runtime": 1.8606,
+      "eval_samples_per_second": 10.749,
+      "eval_steps_per_second": 2.687,
+      "step": 15
     },
     {
-      "epoch": 12.857142857142858,
-      "grad_norm": 0.13014521278503757,
+      "epoch": 1.6,
+      "grad_norm": 0.3737806540766406,
       "learning_rate": 0.0,
-      "loss": 0.0212,
-      "step": 45
+      "loss": 0.5243,
+      "step": 20
     },
     {
-      "epoch": 12.857142857142858,
-      "eval_loss": 0.939548134803772,
-      "eval_runtime": 1.1886,
-      "eval_samples_per_second": 4.207,
-      "eval_steps_per_second": 1.683,
-      "step": 45
+      "epoch": 1.6,
+      "eval_loss": 0.6069896221160889,
+      "eval_runtime": 1.8906,
+      "eval_samples_per_second": 10.579,
+      "eval_steps_per_second": 2.645,
+      "step": 20
     },
     {
-      "epoch": 12.857142857142858,
-      "step": 45,
-      "total_flos": 10453480636416.0,
-      "train_loss": 0.2976339568694433,
-      "train_runtime": 339.6314,
-      "train_samples_per_second": 1.148,
-      "train_steps_per_second": 0.132
+      "epoch": 1.6,
+      "step": 20,
+      "total_flos": 1591906369536.0,
+      "train_loss": 0.6768446683883667,
+      "train_runtime": 82.5503,
+      "train_samples_per_second": 1.938,
+      "train_steps_per_second": 0.242
     }
   ],
   "logging_steps": 5,
-  "max_steps": 45,
+  "max_steps": 20,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 15,
+  "num_train_epochs": 2,
   "save_steps": 1000,
   "stateful_callbacks": {
     "TrainerControl": {
@@ -209,7 +102,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 10453480636416.0,
+  "total_flos": 1591906369536.0,
   "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null