End of training

README.md

@@ -18,11 +18,11 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [riotu-lab/ArabianGPT-01B](https://huggingface.co/riotu-lab/ArabianGPT-01B) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 2.0654
-- Bleu: 0.3119
-- Rouge1: 0.5862
-- Rouge2: 0.3489
-- Rougel: 0.5479
+- Loss: 2.1411
+- Bleu: 0.2987
+- Rouge1: 0.5831
+- Rouge2: 0.3405
+- Rougel: 0.5413
 
 ## Model description
 

all_results.json

@@ -1,19 +1,19 @@
 {
-    "epoch": 18.0,
-    "eval_bleu": 0.31303204778954513,
-    "eval_loss": 1.9793106317520142,
-    "eval_rouge1": 0.5832240471437244,
-    "eval_rouge2": 0.34612838297153165,
-    "eval_rougeL": 0.545438606875621,
-    "eval_runtime": 29.6378,
-    "eval_samples": 925,
-    "eval_samples_per_second": 31.21,
-    "eval_steps_per_second": 3.914,
-    "perplexity": 7.237751788702562,
-    "total_flos": 2.5969031184384e+16,
-    "train_loss": 1.6592627337077177,
-    "train_runtime": 3469.9556,
-    "train_samples": 3681,
-    "train_samples_per_second": 21.216,
-    "train_steps_per_second": 2.657
+    "epoch": 20.0,
+    "eval_bleu": 0.2986862166187925,
+    "eval_loss": 2.1411283016204834,
+    "eval_rouge1": 0.5830931277059221,
+    "eval_rouge2": 0.3405124611999597,
+    "eval_rougeL": 0.5412563338799627,
+    "eval_runtime": 29.7624,
+    "eval_samples": 884,
+    "eval_samples_per_second": 29.702,
+    "eval_steps_per_second": 3.73,
+    "perplexity": 8.50903297033769,
+    "total_flos": 2.7664555180032e+16,
+    "train_loss": 0.09870077279897836,
+    "train_runtime": 270.9621,
+    "train_samples": 3531,
+    "train_samples_per_second": 260.627,
+    "train_steps_per_second": 32.624
 }

egy_training_log.txt

@@ -472,3 +472,5 @@ WARNING:root:Epoch 19.0: No losses recorded yet.
 INFO:absl:Using default tokenizer.
 INFO:root:Epoch 20.0: Train Loss = 1.6189, Eval Loss = 2.0766849517822266
 INFO:absl:Using default tokenizer.
+INFO:__main__:*** Evaluate ***
+INFO:absl:Using default tokenizer.

eval_results.json

@@ -1,13 +1,13 @@
 {
-    "epoch": 18.0,
-    "eval_bleu": 0.31303204778954513,
-    "eval_loss": 1.9793106317520142,
-    "eval_rouge1": 0.5832240471437244,
-    "eval_rouge2": 0.34612838297153165,
-    "eval_rougeL": 0.545438606875621,
-    "eval_runtime": 29.6378,
-    "eval_samples": 925,
-    "eval_samples_per_second": 31.21,
-    "eval_steps_per_second": 3.914,
-    "perplexity": 7.237751788702562
+    "epoch": 20.0,
+    "eval_bleu": 0.2986862166187925,
+    "eval_loss": 2.1411283016204834,
+    "eval_rouge1": 0.5830931277059221,
+    "eval_rouge2": 0.3405124611999597,
+    "eval_rougeL": 0.5412563338799627,
+    "eval_runtime": 29.7624,
+    "eval_samples": 884,
+    "eval_samples_per_second": 29.702,
+    "eval_steps_per_second": 3.73,
+    "perplexity": 8.50903297033769
 }

train_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 18.0,
-    "total_flos": 2.5969031184384e+16,
-    "train_loss": 1.6592627337077177,
-    "train_runtime": 3469.9556,
-    "train_samples": 3681,
-    "train_samples_per_second": 21.216,
-    "train_steps_per_second": 2.657
+    "epoch": 20.0,
+    "total_flos": 2.7664555180032e+16,
+    "train_loss": 0.09870077279897836,
+    "train_runtime": 270.9621,
+    "train_samples": 3531,
+    "train_samples_per_second": 260.627,
+    "train_steps_per_second": 32.624
 }

trainer_state.json

@@ -1,9 +1,9 @@
 {
   "best_metric": 1.9793106317520142,
   "best_model_checkpoint": "/home/iais_marenpielka/Bouthaina/results_fixed/checkpoint-6915",
-  "epoch": 18.0,
+  "epoch": 20.0,
   "eval_steps": 500,
-  "global_step": 8298,
+  "global_step": 8840,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -337,17 +337,48 @@
       "step": 8298
     },
     {
-      "epoch": 18.0,
-      "step": 8298,
-      "total_flos": 2.5969031184384e+16,
-      "train_loss": 1.6592627337077177,
-      "train_runtime": 3469.9556,
-      "train_samples_per_second": 21.216,
-      "train_steps_per_second": 2.657
+      "epoch": 19.0,
+      "eval_bleu": 0.3098880521487277,
+      "eval_loss": 2.0766849517822266,
+      "eval_rouge1": 0.5857850781975715,
+      "eval_rouge2": 0.3476439342876294,
+      "eval_rougeL": 0.5470857900583828,
+      "eval_runtime": 29.9167,
+      "eval_samples_per_second": 29.549,
+      "eval_steps_per_second": 3.71,
+      "step": 8398
+    },
+    {
+      "epoch": 19.23076923076923,
+      "grad_norm": 1.234198808670044,
+      "learning_rate": 2.0383693045563552e-06,
+      "loss": 1.6189,
+      "step": 8500
+    },
+    {
+      "epoch": 20.0,
+      "eval_bleu": 0.31187064199371684,
+      "eval_loss": 2.06535267829895,
+      "eval_rouge1": 0.5861980524100046,
+      "eval_rouge2": 0.3488620525891558,
+      "eval_rougeL": 0.5478910457529194,
+      "eval_runtime": 29.8775,
+      "eval_samples_per_second": 29.588,
+      "eval_steps_per_second": 3.715,
+      "step": 8840
+    },
+    {
+      "epoch": 20.0,
+      "step": 8840,
+      "total_flos": 2.7664555180032e+16,
+      "train_loss": 0.09870077279897836,
+      "train_runtime": 270.9621,
+      "train_samples_per_second": 260.627,
+      "train_steps_per_second": 32.624
     }
   ],
   "logging_steps": 500,
-  "max_steps": 9220,
+  "max_steps": 8840,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 20,
   "save_steps": 500,
@@ -372,7 +403,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 2.5969031184384e+16,
+  "total_flos": 2.7664555180032e+16,
   "train_batch_size": 8,
   "trial_name": null,
   "trial_params": null