Model save

README.md

@@ -3,7 +3,6 @@ base_model: meta-llama/Llama-3.1-8B-Instruct
 library_name: peft
 license: llama3.1
 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-3.1-8B-Instruct](https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.6177
+- Loss: 0.5915
 
 ## Model description
 
@@ -50,16 +49,18 @@ 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: 30
 
 ### Training results
 
 | Training Loss | Epoch  | Step | Validation Loss |
 |:-------------:|:------:|:----:|:---------------:|
-| 0.9053        | 0.7692 | 5    | 0.6993          |
-| 0.623         | 1.5385 | 10   | 0.6391          |
-| 0.5832        | 2.3077 | 15   | 0.6230          |
-| 0.5339        | 3.0769 | 20   | 0.6177          |
+| 0.9274        | 0.7692 | 5    | 0.7078          |
+| 0.6265        | 1.5385 | 10   | 0.6385          |
+| 0.5732        | 2.3077 | 15   | 0.6119          |
+| 0.4985        | 3.0769 | 20   | 0.5948          |
+| 0.4635        | 3.8462 | 25   | 0.5914          |
+| 0.455         | 4.6154 | 30   | 0.5915          |
 
 
 ### Framework versions

all_results.json

@@ -1,14 +1,9 @@
 {
-    "epoch": 3.076923076923077,
-    "eval_loss": 0.6177161335945129,
-    "eval_runtime": 1.3449,
-    "eval_samples": 20,
-    "eval_samples_per_second": 14.871,
-    "eval_steps_per_second": 2.231,
-    "total_flos": 2380446105600.0,
-    "train_loss": 0.6673399448394776,
-    "train_runtime": 85.3981,
+    "epoch": 4.615384615384615,
+    "total_flos": 3573950251008.0,
+    "train_loss": 0.5939286708831787,
+    "train_runtime": 114.5202,
     "train_samples": 100,
-    "train_samples_per_second": 3.747,
-    "train_steps_per_second": 0.234
+    "train_samples_per_second": 4.191,
+    "train_steps_per_second": 0.262
 }

train_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 3.076923076923077,
-    "total_flos": 2380446105600.0,
-    "train_loss": 0.6673399448394776,
-    "train_runtime": 85.3981,
+    "epoch": 4.615384615384615,
+    "total_flos": 3573950251008.0,
+    "train_loss": 0.5939286708831787,
+    "train_runtime": 114.5202,
     "train_samples": 100,
-    "train_samples_per_second": 3.747,
-    "train_steps_per_second": 0.234
+    "train_samples_per_second": 4.191,
+    "train_steps_per_second": 0.262
 }

trainer_state.json

@@ -1,94 +1,124 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 3.076923076923077,
+  "epoch": 4.615384615384615,
   "eval_steps": 5,
-  "global_step": 20,
+  "global_step": 30,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.15384615384615385,
-      "grad_norm": 0.872483858294593,
-      "learning_rate": 0.0001,
+      "grad_norm": 0.8729932578899202,
+      "learning_rate": 6.666666666666667e-05,
       "loss": 1.0249,
       "step": 1
     },
     {
       "epoch": 0.7692307692307693,
-      "grad_norm": 0.41582677007417446,
-      "learning_rate": 0.00018660254037844388,
-      "loss": 0.9053,
+      "grad_norm": 0.39702351772749944,
+      "learning_rate": 0.00019730448705798239,
+      "loss": 0.9274,
       "step": 5
     },
     {
       "epoch": 0.7692307692307693,
-      "eval_loss": 0.6993206143379211,
-      "eval_runtime": 2.4983,
-      "eval_samples_per_second": 8.005,
-      "eval_steps_per_second": 1.201,
+      "eval_loss": 0.7077673077583313,
+      "eval_runtime": 2.52,
+      "eval_samples_per_second": 7.936,
+      "eval_steps_per_second": 1.19,
       "step": 5
     },
     {
       "epoch": 1.5384615384615383,
-      "grad_norm": 0.23072988157616092,
-      "learning_rate": 0.00011736481776669306,
-      "loss": 0.623,
+      "grad_norm": 0.2396442839420031,
+      "learning_rate": 0.0001686241637868734,
+      "loss": 0.6265,
       "step": 10
     },
     {
       "epoch": 1.5384615384615383,
-      "eval_loss": 0.6391023993492126,
-      "eval_runtime": 1.304,
-      "eval_samples_per_second": 15.337,
-      "eval_steps_per_second": 2.301,
+      "eval_loss": 0.6385108232498169,
+      "eval_runtime": 1.2976,
+      "eval_samples_per_second": 15.413,
+      "eval_steps_per_second": 2.312,
       "step": 10
     },
     {
       "epoch": 2.3076923076923075,
-      "grad_norm": 0.26693843382261334,
-      "learning_rate": 3.5721239031346066e-05,
-      "loss": 0.5832,
+      "grad_norm": 0.310264308105529,
+      "learning_rate": 0.00011736481776669306,
+      "loss": 0.5732,
       "step": 15
     },
     {
       "epoch": 2.3076923076923075,
-      "eval_loss": 0.6230154037475586,
-      "eval_runtime": 1.3111,
-      "eval_samples_per_second": 15.254,
-      "eval_steps_per_second": 2.288,
+      "eval_loss": 0.6119084358215332,
+      "eval_runtime": 1.2993,
+      "eval_samples_per_second": 15.392,
+      "eval_steps_per_second": 2.309,
       "step": 15
     },
     {
       "epoch": 3.076923076923077,
-      "grad_norm": 0.22743307044169467,
-      "learning_rate": 0.0,
-      "loss": 0.5339,
+      "grad_norm": 0.24471165996330677,
+      "learning_rate": 6.039202339608432e-05,
+      "loss": 0.4985,
       "step": 20
     },
     {
       "epoch": 3.076923076923077,
-      "eval_loss": 0.6177161335945129,
-      "eval_runtime": 1.3038,
-      "eval_samples_per_second": 15.34,
-      "eval_steps_per_second": 2.301,
+      "eval_loss": 0.5947881937026978,
+      "eval_runtime": 1.2973,
+      "eval_samples_per_second": 15.416,
+      "eval_steps_per_second": 2.312,
       "step": 20
     },
     {
-      "epoch": 3.076923076923077,
-      "step": 20,
-      "total_flos": 2380446105600.0,
-      "train_loss": 0.6673399448394776,
-      "train_runtime": 85.3981,
-      "train_samples_per_second": 3.747,
-      "train_steps_per_second": 0.234
+      "epoch": 3.8461538461538463,
+      "grad_norm": 0.30409093115670077,
+      "learning_rate": 1.6451218858706374e-05,
+      "loss": 0.4635,
+      "step": 25
+    },
+    {
+      "epoch": 3.8461538461538463,
+      "eval_loss": 0.5913671851158142,
+      "eval_runtime": 1.3034,
+      "eval_samples_per_second": 15.345,
+      "eval_steps_per_second": 2.302,
+      "step": 25
+    },
+    {
+      "epoch": 4.615384615384615,
+      "grad_norm": 0.28415223207533613,
+      "learning_rate": 0.0,
+      "loss": 0.455,
+      "step": 30
+    },
+    {
+      "epoch": 4.615384615384615,
+      "eval_loss": 0.5914875268936157,
+      "eval_runtime": 1.2992,
+      "eval_samples_per_second": 15.394,
+      "eval_steps_per_second": 2.309,
+      "step": 30
+    },
+    {
+      "epoch": 4.615384615384615,
+      "step": 30,
+      "total_flos": 3573950251008.0,
+      "train_loss": 0.5939286708831787,
+      "train_runtime": 114.5202,
+      "train_samples_per_second": 4.191,
+      "train_steps_per_second": 0.262
     }
   ],
   "logging_steps": 5,
-  "max_steps": 20,
+  "max_steps": 30,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 4,
+  "num_train_epochs": 5,
   "save_steps": 1000,
   "stateful_callbacks": {
     "TrainerControl": {
@@ -102,7 +132,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 2380446105600.0,
+  "total_flos": 3573950251008.0,
   "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null