Model save

README.md

@@ -3,7 +3,6 @@ base_model: meta-llama/Llama-2-7b-chat-hf
 library_name: peft
 license: llama2
 tags:
-- alignment-handbook
 - trl
 - sft
 - generated_from_trainer
@@ -19,7 +18,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [meta-llama/Llama-2-7b-chat-hf](https://huggingface.co/meta-llama/Llama-2-7b-chat-hf) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 1.2316
+- Loss: 1.1053
 
 ## Model description
 
@@ -50,16 +49,22 @@ 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
-- training_steps: 20
+- training_steps: 50
 
 ### Training results
 
 | Training Loss | Epoch  | Step | Validation Loss |
 |:-------------:|:------:|:----:|:---------------:|
-| 2.4151        | 0.7692 | 5    | 1.7343          |
-| 1.5453        | 1.5385 | 10   | 1.3924          |
-| 1.3352        | 2.3077 | 15   | 1.2640          |
-| 1.1897        | 3.0769 | 20   | 1.2316          |
+| 2.622         | 0.7692 | 5    | 2.1044          |
+| 1.7003        | 1.5385 | 10   | 1.4252          |
+| 1.3273        | 2.3077 | 15   | 1.2309          |
+| 1.1213        | 3.0769 | 20   | 1.1700          |
+| 1.004         | 3.8462 | 25   | 1.1185          |
+| 0.9185        | 4.6154 | 30   | 1.0977          |
+| 0.8917        | 5.3846 | 35   | 1.0975          |
+| 0.8018        | 6.1538 | 40   | 1.1012          |
+| 0.7711        | 6.9231 | 45   | 1.1045          |
+| 0.7648        | 7.6923 | 50   | 1.1053          |
 
 
 ### Framework versions

all_results.json

@@ -1,14 +1,9 @@
 {
-    "epoch": 3.076923076923077,
-    "eval_loss": 1.2315905094146729,
-    "eval_runtime": 1.4037,
-    "eval_samples": 20,
-    "eval_samples_per_second": 14.248,
-    "eval_steps_per_second": 2.137,
-    "total_flos": 207917580288.0,
-    "train_loss": 1.6456967353820802,
-    "train_runtime": 83.291,
+    "epoch": 7.6923076923076925,
+    "total_flos": 516835442688.0,
+    "train_loss": 1.1978940773010254,
+    "train_runtime": 176.0775,
     "train_samples": 100,
-    "train_samples_per_second": 3.842,
-    "train_steps_per_second": 0.24
+    "train_samples_per_second": 4.543,
+    "train_steps_per_second": 0.284
 }

train_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 3.076923076923077,
-    "total_flos": 207917580288.0,
-    "train_loss": 1.6456967353820802,
-    "train_runtime": 83.291,
+    "epoch": 7.6923076923076925,
+    "total_flos": 516835442688.0,
+    "train_loss": 1.1978940773010254,
+    "train_runtime": 176.0775,
     "train_samples": 100,
-    "train_samples_per_second": 3.842,
-    "train_steps_per_second": 0.24
+    "train_samples_per_second": 4.543,
+    "train_steps_per_second": 0.284
 }

trainer_state.json

@@ -1,9 +1,9 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 3.076923076923077,
+  "epoch": 7.6923076923076925,
   "eval_steps": 5,
-  "global_step": 20,
+  "global_step": 50,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -11,84 +11,174 @@
     {
       "epoch": 0.15384615384615385,
       "grad_norm": 1.9916846117848972,
-      "learning_rate": 0.0001,
+      "learning_rate": 4e-05,
       "loss": 2.9025,
       "step": 1
     },
     {
       "epoch": 0.7692307692307693,
-      "grad_norm": 1.2833575797244527,
-      "learning_rate": 0.00018660254037844388,
-      "loss": 2.4151,
+      "grad_norm": 1.603915199386799,
+      "learning_rate": 0.0002,
+      "loss": 2.622,
       "step": 5
     },
     {
       "epoch": 0.7692307692307693,
-      "eval_loss": 1.7343145608901978,
-      "eval_runtime": 2.6395,
-      "eval_samples_per_second": 7.577,
-      "eval_steps_per_second": 1.137,
+      "eval_loss": 2.1043648719787598,
+      "eval_runtime": 2.618,
+      "eval_samples_per_second": 7.639,
+      "eval_steps_per_second": 1.146,
       "step": 5
     },
     {
       "epoch": 1.5384615384615383,
-      "grad_norm": 0.5231603461231863,
-      "learning_rate": 0.00011736481776669306,
-      "loss": 1.5453,
+      "grad_norm": 0.5619542049385124,
+      "learning_rate": 0.00019396926207859084,
+      "loss": 1.7003,
       "step": 10
     },
     {
       "epoch": 1.5384615384615383,
-      "eval_loss": 1.3923780918121338,
-      "eval_runtime": 1.3672,
-      "eval_samples_per_second": 14.628,
-      "eval_steps_per_second": 2.194,
+      "eval_loss": 1.4251809120178223,
+      "eval_runtime": 1.3576,
+      "eval_samples_per_second": 14.732,
+      "eval_steps_per_second": 2.21,
       "step": 10
     },
     {
       "epoch": 2.3076923076923075,
-      "grad_norm": 0.8142750502442172,
-      "learning_rate": 3.5721239031346066e-05,
-      "loss": 1.3352,
+      "grad_norm": 0.459259181992223,
+      "learning_rate": 0.0001766044443118978,
+      "loss": 1.3273,
       "step": 15
     },
     {
       "epoch": 2.3076923076923075,
-      "eval_loss": 1.263976812362671,
-      "eval_runtime": 1.3788,
-      "eval_samples_per_second": 14.506,
-      "eval_steps_per_second": 2.176,
+      "eval_loss": 1.2308980226516724,
+      "eval_runtime": 1.3431,
+      "eval_samples_per_second": 14.891,
+      "eval_steps_per_second": 2.234,
       "step": 15
     },
     {
       "epoch": 3.076923076923077,
-      "grad_norm": 0.6311551234598356,
-      "learning_rate": 0.0,
-      "loss": 1.1897,
+      "grad_norm": 0.38172071531869,
+      "learning_rate": 0.00015000000000000001,
+      "loss": 1.1213,
       "step": 20
     },
     {
       "epoch": 3.076923076923077,
-      "eval_loss": 1.2315905094146729,
-      "eval_runtime": 1.3669,
-      "eval_samples_per_second": 14.632,
-      "eval_steps_per_second": 2.195,
+      "eval_loss": 1.1700019836425781,
+      "eval_runtime": 1.348,
+      "eval_samples_per_second": 14.836,
+      "eval_steps_per_second": 2.225,
       "step": 20
     },
     {
-      "epoch": 3.076923076923077,
-      "step": 20,
-      "total_flos": 207917580288.0,
-      "train_loss": 1.6456967353820802,
-      "train_runtime": 83.291,
-      "train_samples_per_second": 3.842,
-      "train_steps_per_second": 0.24
+      "epoch": 3.8461538461538463,
+      "grad_norm": 0.6756597684966411,
+      "learning_rate": 0.00011736481776669306,
+      "loss": 1.004,
+      "step": 25
+    },
+    {
+      "epoch": 3.8461538461538463,
+      "eval_loss": 1.1185358762741089,
+      "eval_runtime": 1.3436,
+      "eval_samples_per_second": 14.885,
+      "eval_steps_per_second": 2.233,
+      "step": 25
+    },
+    {
+      "epoch": 4.615384615384615,
+      "grad_norm": 1.092815432478573,
+      "learning_rate": 8.263518223330697e-05,
+      "loss": 0.9185,
+      "step": 30
+    },
+    {
+      "epoch": 4.615384615384615,
+      "eval_loss": 1.0976699590682983,
+      "eval_runtime": 1.3469,
+      "eval_samples_per_second": 14.849,
+      "eval_steps_per_second": 2.227,
+      "step": 30
+    },
+    {
+      "epoch": 5.384615384615385,
+      "grad_norm": 0.3415440309552443,
+      "learning_rate": 5.000000000000002e-05,
+      "loss": 0.8917,
+      "step": 35
+    },
+    {
+      "epoch": 5.384615384615385,
+      "eval_loss": 1.0975412130355835,
+      "eval_runtime": 1.3424,
+      "eval_samples_per_second": 14.899,
+      "eval_steps_per_second": 2.235,
+      "step": 35
+    },
+    {
+      "epoch": 6.153846153846154,
+      "grad_norm": 0.3137972502013654,
+      "learning_rate": 2.339555568810221e-05,
+      "loss": 0.8018,
+      "step": 40
+    },
+    {
+      "epoch": 6.153846153846154,
+      "eval_loss": 1.1011784076690674,
+      "eval_runtime": 1.341,
+      "eval_samples_per_second": 14.915,
+      "eval_steps_per_second": 2.237,
+      "step": 40
+    },
+    {
+      "epoch": 6.923076923076923,
+      "grad_norm": 0.5726296599342907,
+      "learning_rate": 6.030737921409169e-06,
+      "loss": 0.7711,
+      "step": 45
+    },
+    {
+      "epoch": 6.923076923076923,
+      "eval_loss": 1.1044588088989258,
+      "eval_runtime": 1.3483,
+      "eval_samples_per_second": 14.833,
+      "eval_steps_per_second": 2.225,
+      "step": 45
+    },
+    {
+      "epoch": 7.6923076923076925,
+      "grad_norm": 0.36148605864973277,
+      "learning_rate": 0.0,
+      "loss": 0.7648,
+      "step": 50
+    },
+    {
+      "epoch": 7.6923076923076925,
+      "eval_loss": 1.105344295501709,
+      "eval_runtime": 1.3427,
+      "eval_samples_per_second": 14.896,
+      "eval_steps_per_second": 2.234,
+      "step": 50
+    },
+    {
+      "epoch": 7.6923076923076925,
+      "step": 50,
+      "total_flos": 516835442688.0,
+      "train_loss": 1.1978940773010254,
+      "train_runtime": 176.0775,
+      "train_samples_per_second": 4.543,
+      "train_steps_per_second": 0.284
     }
   ],
   "logging_steps": 5,
-  "max_steps": 20,
+  "max_steps": 50,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 4,
+  "num_train_epochs": 9,
   "save_steps": 1000,
   "stateful_callbacks": {
     "TrainerControl": {
@@ -102,7 +192,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 207917580288.0,
+  "total_flos": 516835442688.0,
   "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null