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: 0.7346
+- Loss: 0.6129
 
 ## 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 |
 |:-------------:|:------:|:----:|:---------------:|
-| 1.3704        | 0.7692 | 5    | 0.9995          |
-| 0.8977        | 1.5385 | 10   | 0.8262          |
-| 0.7547        | 2.3077 | 15   | 0.7461          |
-| 0.6981        | 3.0769 | 20   | 0.7346          |
+| 1.4804        | 0.7692 | 5    | 1.1583          |
+| 0.9708        | 1.5385 | 10   | 0.8724          |
+| 0.7685        | 2.3077 | 15   | 0.7143          |
+| 0.6454        | 3.0769 | 20   | 0.6602          |
+| 0.5826        | 3.8462 | 25   | 0.6317          |
+| 0.5458        | 4.6154 | 30   | 0.6193          |
+| 0.4844        | 5.3846 | 35   | 0.6151          |
+| 0.506         | 6.1538 | 40   | 0.6133          |
+| 0.4766        | 6.9231 | 45   | 0.6126          |
+| 0.4639        | 7.6923 | 50   | 0.6129          |
 
 
 ### Framework versions

all_results.json

@@ -1,14 +1,9 @@
 {
-    "epoch": 3.076923076923077,
-    "eval_loss": 0.7345929145812988,
-    "eval_runtime": 1.4047,
-    "eval_samples": 20,
-    "eval_samples_per_second": 14.237,
-    "eval_steps_per_second": 2.136,
-    "total_flos": 469873164288.0,
-    "train_loss": 0.9361693680286407,
-    "train_runtime": 84.6814,
+    "epoch": 7.6923076923076925,
+    "total_flos": 1205290795008.0,
+    "train_loss": 0.692623188495636,
+    "train_runtime": 177.9895,
     "train_samples": 100,
-    "train_samples_per_second": 3.779,
-    "train_steps_per_second": 0.236
+    "train_samples_per_second": 4.495,
+    "train_steps_per_second": 0.281
 }

train_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 3.076923076923077,
-    "total_flos": 469873164288.0,
-    "train_loss": 0.9361693680286407,
-    "train_runtime": 84.6814,
+    "epoch": 7.6923076923076925,
+    "total_flos": 1205290795008.0,
+    "train_loss": 0.692623188495636,
+    "train_runtime": 177.9895,
     "train_samples": 100,
-    "train_samples_per_second": 3.779,
-    "train_steps_per_second": 0.236
+    "train_samples_per_second": 4.495,
+    "train_steps_per_second": 0.281
 }

trainer_state.json

@@ -1,94 +1,184 @@
 {
   "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,
   "log_history": [
     {
       "epoch": 0.15384615384615385,
-      "grad_norm": 0.7335724302830133,
-      "learning_rate": 0.0001,
+      "grad_norm": 0.7337189811685318,
+      "learning_rate": 4e-05,
       "loss": 1.4896,
       "step": 1
     },
     {
       "epoch": 0.7692307692307693,
-      "grad_norm": 0.5632629853760255,
-      "learning_rate": 0.00018660254037844388,
-      "loss": 1.3704,
+      "grad_norm": 0.788344053257058,
+      "learning_rate": 0.0002,
+      "loss": 1.4804,
       "step": 5
     },
     {
       "epoch": 0.7692307692307693,
-      "eval_loss": 0.9995014071464539,
-      "eval_runtime": 2.648,
-      "eval_samples_per_second": 7.553,
-      "eval_steps_per_second": 1.133,
+      "eval_loss": 1.1582854986190796,
+      "eval_runtime": 2.6274,
+      "eval_samples_per_second": 7.612,
+      "eval_steps_per_second": 1.142,
       "step": 5
     },
     {
       "epoch": 1.5384615384615383,
-      "grad_norm": 0.35539514680445433,
-      "learning_rate": 0.00011736481776669306,
-      "loss": 0.8977,
+      "grad_norm": 0.4371059352187654,
+      "learning_rate": 0.00019396926207859084,
+      "loss": 0.9708,
       "step": 10
     },
     {
       "epoch": 1.5384615384615383,
-      "eval_loss": 0.8261862993240356,
-      "eval_runtime": 1.4047,
-      "eval_samples_per_second": 14.238,
-      "eval_steps_per_second": 2.136,
+      "eval_loss": 0.8723791241645813,
+      "eval_runtime": 1.3731,
+      "eval_samples_per_second": 14.565,
+      "eval_steps_per_second": 2.185,
       "step": 10
     },
     {
       "epoch": 2.3076923076923075,
-      "grad_norm": 0.23338309624627512,
-      "learning_rate": 3.5721239031346066e-05,
-      "loss": 0.7547,
+      "grad_norm": 0.25625240789847736,
+      "learning_rate": 0.0001766044443118978,
+      "loss": 0.7685,
       "step": 15
     },
     {
       "epoch": 2.3076923076923075,
-      "eval_loss": 0.7461313605308533,
-      "eval_runtime": 1.4009,
-      "eval_samples_per_second": 14.277,
-      "eval_steps_per_second": 2.141,
+      "eval_loss": 0.7142642140388489,
+      "eval_runtime": 1.3722,
+      "eval_samples_per_second": 14.575,
+      "eval_steps_per_second": 2.186,
       "step": 15
     },
     {
       "epoch": 3.076923076923077,
-      "grad_norm": 0.22429457553641297,
-      "learning_rate": 0.0,
-      "loss": 0.6981,
+      "grad_norm": 0.37095540648653375,
+      "learning_rate": 0.00015000000000000001,
+      "loss": 0.6454,
       "step": 20
     },
     {
       "epoch": 3.076923076923077,
-      "eval_loss": 0.7345929145812988,
-      "eval_runtime": 1.3788,
-      "eval_samples_per_second": 14.506,
-      "eval_steps_per_second": 2.176,
+      "eval_loss": 0.660171389579773,
+      "eval_runtime": 1.3569,
+      "eval_samples_per_second": 14.739,
+      "eval_steps_per_second": 2.211,
       "step": 20
     },
     {
-      "epoch": 3.076923076923077,
-      "step": 20,
-      "total_flos": 469873164288.0,
-      "train_loss": 0.9361693680286407,
-      "train_runtime": 84.6814,
-      "train_samples_per_second": 3.779,
-      "train_steps_per_second": 0.236
+      "epoch": 3.8461538461538463,
+      "grad_norm": 0.269490348868028,
+      "learning_rate": 0.00011736481776669306,
+      "loss": 0.5826,
+      "step": 25
+    },
+    {
+      "epoch": 3.8461538461538463,
+      "eval_loss": 0.6317092776298523,
+      "eval_runtime": 1.3548,
+      "eval_samples_per_second": 14.762,
+      "eval_steps_per_second": 2.214,
+      "step": 25
+    },
+    {
+      "epoch": 4.615384615384615,
+      "grad_norm": 0.1367860640067106,
+      "learning_rate": 8.263518223330697e-05,
+      "loss": 0.5458,
+      "step": 30
+    },
+    {
+      "epoch": 4.615384615384615,
+      "eval_loss": 0.6193430423736572,
+      "eval_runtime": 1.3508,
+      "eval_samples_per_second": 14.806,
+      "eval_steps_per_second": 2.221,
+      "step": 30
+    },
+    {
+      "epoch": 5.384615384615385,
+      "grad_norm": 0.14497012519202485,
+      "learning_rate": 5.000000000000002e-05,
+      "loss": 0.4844,
+      "step": 35
+    },
+    {
+      "epoch": 5.384615384615385,
+      "eval_loss": 0.6151253581047058,
+      "eval_runtime": 1.3512,
+      "eval_samples_per_second": 14.802,
+      "eval_steps_per_second": 2.22,
+      "step": 35
+    },
+    {
+      "epoch": 6.153846153846154,
+      "grad_norm": 0.1438702628497647,
+      "learning_rate": 2.339555568810221e-05,
+      "loss": 0.506,
+      "step": 40
+    },
+    {
+      "epoch": 6.153846153846154,
+      "eval_loss": 0.6133347749710083,
+      "eval_runtime": 1.358,
+      "eval_samples_per_second": 14.728,
+      "eval_steps_per_second": 2.209,
+      "step": 40
+    },
+    {
+      "epoch": 6.923076923076923,
+      "grad_norm": 0.15810290565530724,
+      "learning_rate": 6.030737921409169e-06,
+      "loss": 0.4766,
+      "step": 45
+    },
+    {
+      "epoch": 6.923076923076923,
+      "eval_loss": 0.6125746965408325,
+      "eval_runtime": 1.3498,
+      "eval_samples_per_second": 14.817,
+      "eval_steps_per_second": 2.223,
+      "step": 45
+    },
+    {
+      "epoch": 7.6923076923076925,
+      "grad_norm": 0.16553701241746885,
+      "learning_rate": 0.0,
+      "loss": 0.4639,
+      "step": 50
+    },
+    {
+      "epoch": 7.6923076923076925,
+      "eval_loss": 0.6129434108734131,
+      "eval_runtime": 1.3475,
+      "eval_samples_per_second": 14.842,
+      "eval_steps_per_second": 2.226,
+      "step": 50
+    },
+    {
+      "epoch": 7.6923076923076925,
+      "step": 50,
+      "total_flos": 1205290795008.0,
+      "train_loss": 0.692623188495636,
+      "train_runtime": 177.9895,
+      "train_samples_per_second": 4.495,
+      "train_steps_per_second": 0.281
     }
   ],
   "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": 469873164288.0,
+  "total_flos": 1205290795008.0,
   "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null