End of training

README.md

@@ -18,8 +18,8 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [google-bert/bert-base-multilingual-cased](https://huggingface.co/google-bert/bert-base-multilingual-cased) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.7142
-- Accuracy: 0.7941
+- Loss: 0.7135
+- Accuracy: 0.7981
 
 ## Model description
 

all_results.json

@@ -1,15 +1,15 @@
 {
     "epoch": 5.0,
-    "eval_accuracy": 0.7772225450319582,
-    "eval_loss": 0.7830244302749634,
-    "eval_runtime": 9.0522,
+    "eval_accuracy": 0.7981406159209762,
+    "eval_loss": 0.713493287563324,
+    "eval_runtime": 8.9062,
     "eval_samples": 8605,
-    "eval_samples_per_second": 950.598,
-    "eval_steps_per_second": 118.866,
+    "eval_samples_per_second": 966.185,
+    "eval_steps_per_second": 7.635,
     "total_flos": 4.52871045169152e+16,
-    "train_loss": 0.6343187750493727,
-    "train_runtime": 497.1667,
+    "train_loss": 0.846687467213457,
+    "train_runtime": 657.9188,
     "train_samples": 34420,
-    "train_samples_per_second": 346.162,
-    "train_steps_per_second": 2.705
+    "train_samples_per_second": 261.582,
+    "train_steps_per_second": 2.044
 }

config.json

@@ -61,7 +61,7 @@
   "pooler_type": "first_token_transform",
   "position_embedding_type": "absolute",
   "problem_type": "single_label_classification",
-  "torch_dtype": "float32",
+  "torch_dtype": "float16",
   "transformers_version": "4.49.0.dev0",
   "type_vocab_size": 2,
   "use_cache": true,

eval_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 5.0,
-    "eval_accuracy": 0.7772225450319582,
-    "eval_loss": 0.7830244302749634,
-    "eval_runtime": 9.0522,
+    "eval_accuracy": 0.7981406159209762,
+    "eval_loss": 0.713493287563324,
+    "eval_runtime": 8.9062,
     "eval_samples": 8605,
-    "eval_samples_per_second": 950.598,
-    "eval_steps_per_second": 118.866
+    "eval_samples_per_second": 966.185,
+    "eval_steps_per_second": 7.635
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0b9a4546ae5611e3cfafad236495e1e3c7d480d46933737b8acb17807c5f337e
-size 711486528
+oid sha256:9125bf178a92f532d8d8b4d53ad9c709a765e3daff0466d4e9873e7dd8bf53d8
+size 355755032

train_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 5.0,
     "total_flos": 4.52871045169152e+16,
-    "train_loss": 0.6343187750493727,
-    "train_runtime": 497.1667,
+    "train_loss": 0.846687467213457,
+    "train_runtime": 657.9188,
     "train_samples": 34420,
-    "train_samples_per_second": 346.162,
-    "train_steps_per_second": 2.705
+    "train_samples_per_second": 261.582,
+    "train_steps_per_second": 2.044
 }

trainer_state.json

@@ -1,6 +1,6 @@
 {
-  "best_metric": 0.7830244302749634,
-  "best_model_checkpoint": "./bert-base-multilingual-cased/checkpoint-1345",
+  "best_metric": 0.7983730389308542,
+  "best_model_checkpoint": "./sdg-classifier/checkpoint-1076",
   "epoch": 5.0,
   "eval_steps": 500,
   "global_step": 1345,
@@ -10,148 +10,148 @@
   "log_history": [
     {
       "epoch": 0.37174721189591076,
-      "grad_norm": 3.511324882507324,
-      "learning_rate": 9.25650557620818e-06,
-      "loss": 2.1825,
+      "grad_norm": 3.068047046661377,
+      "learning_rate": 1.4370370370370372e-05,
+      "loss": 2.5931,
       "step": 100
     },
     {
       "epoch": 0.7434944237918215,
-      "grad_norm": 6.582064628601074,
-      "learning_rate": 8.513011152416357e-06,
-      "loss": 1.2704,
+      "grad_norm": 4.889872074127197,
+      "learning_rate": 1.897520661157025e-05,
+      "loss": 1.2927,
       "step": 200
     },
     {
       "epoch": 1.0,
-      "eval_accuracy": 0.7317838466008135,
-      "eval_loss": 0.9832090139389038,
-      "eval_runtime": 9.2351,
-      "eval_samples_per_second": 931.767,
-      "eval_steps_per_second": 116.511,
+      "eval_accuracy": 0.7515398024404416,
+      "eval_loss": 0.8947181105613708,
+      "eval_runtime": 9.4625,
+      "eval_samples_per_second": 909.378,
+      "eval_steps_per_second": 7.186,
       "step": 269
     },
     {
       "epoch": 1.1152416356877324,
-      "grad_norm": 4.232427597045898,
-      "learning_rate": 7.769516728624536e-06,
-      "loss": 0.9943,
+      "grad_norm": 4.267650604248047,
+      "learning_rate": 1.732231404958678e-05,
+      "loss": 0.9586,
       "step": 300
     },
     {
       "epoch": 1.486988847583643,
-      "grad_norm": 4.7747673988342285,
-      "learning_rate": 7.0260223048327144e-06,
-      "loss": 0.9604,
+      "grad_norm": 5.622753620147705,
+      "learning_rate": 1.566942148760331e-05,
+      "loss": 0.8529,
       "step": 400
     },
     {
       "epoch": 1.858736059479554,
-      "grad_norm": 4.983441352844238,
-      "learning_rate": 6.282527881040893e-06,
-      "loss": 0.8977,
+      "grad_norm": 4.57409143447876,
+      "learning_rate": 1.4016528925619836e-05,
+      "loss": 0.7953,
       "step": 500
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.7608367228355607,
-      "eval_loss": 0.8561291098594666,
-      "eval_runtime": 9.0456,
-      "eval_samples_per_second": 951.291,
-      "eval_steps_per_second": 118.953,
+      "eval_accuracy": 0.7795467751307379,
+      "eval_loss": 0.7700016498565674,
+      "eval_runtime": 9.1997,
+      "eval_samples_per_second": 935.353,
+      "eval_steps_per_second": 7.392,
       "step": 538
     },
     {
       "epoch": 2.2304832713754648,
-      "grad_norm": 4.7981953620910645,
-      "learning_rate": 5.539033457249071e-06,
-      "loss": 0.8473,
+      "grad_norm": 5.621212005615234,
+      "learning_rate": 1.2363636363636364e-05,
+      "loss": 0.7169,
       "step": 600
     },
     {
       "epoch": 2.6022304832713754,
-      "grad_norm": 6.707701206207275,
-      "learning_rate": 4.795539033457249e-06,
-      "loss": 0.8154,
+      "grad_norm": 7.005183696746826,
+      "learning_rate": 1.0710743801652894e-05,
+      "loss": 0.666,
       "step": 700
     },
     {
       "epoch": 2.973977695167286,
-      "grad_norm": 5.124625205993652,
-      "learning_rate": 4.052044609665427e-06,
-      "loss": 0.7954,
+      "grad_norm": 5.196578025817871,
+      "learning_rate": 9.057851239669422e-06,
+      "loss": 0.6549,
       "step": 800
     },
     {
       "epoch": 3.0,
-      "eval_accuracy": 0.7712957582800697,
-      "eval_loss": 0.8103732466697693,
-      "eval_runtime": 9.0145,
-      "eval_samples_per_second": 954.573,
-      "eval_steps_per_second": 119.363,
+      "eval_accuracy": 0.7937245787332946,
+      "eval_loss": 0.7241168022155762,
+      "eval_runtime": 9.2798,
+      "eval_samples_per_second": 927.284,
+      "eval_steps_per_second": 7.328,
       "step": 807
     },
     {
       "epoch": 3.345724907063197,
-      "grad_norm": 5.56990909576416,
-      "learning_rate": 3.3085501858736065e-06,
-      "loss": 0.7345,
+      "grad_norm": 4.715315341949463,
+      "learning_rate": 7.40495867768595e-06,
+      "loss": 0.5591,
       "step": 900
     },
     {
       "epoch": 3.717472118959108,
-      "grad_norm": 7.059451580047607,
-      "learning_rate": 2.5650557620817845e-06,
-      "loss": 0.7413,
+      "grad_norm": 7.226228713989258,
+      "learning_rate": 5.7520661157024795e-06,
+      "loss": 0.5658,
       "step": 1000
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.7769901220220802,
-      "eval_loss": 0.7903115749359131,
-      "eval_runtime": 8.9773,
-      "eval_samples_per_second": 958.534,
-      "eval_steps_per_second": 119.858,
+      "eval_accuracy": 0.7983730389308542,
+      "eval_loss": 0.7134895920753479,
+      "eval_runtime": 9.1183,
+      "eval_samples_per_second": 943.703,
+      "eval_steps_per_second": 7.458,
       "step": 1076
     },
     {
       "epoch": 4.089219330855019,
-      "grad_norm": 7.092191219329834,
-      "learning_rate": 1.821561338289963e-06,
-      "loss": 0.7269,
+      "grad_norm": 7.13743782043457,
+      "learning_rate": 4.099173553719009e-06,
+      "loss": 0.542,
       "step": 1100
     },
     {
       "epoch": 4.4609665427509295,
-      "grad_norm": 6.146620750427246,
-      "learning_rate": 1.0780669144981413e-06,
-      "loss": 0.7001,
+      "grad_norm": 6.263808727264404,
+      "learning_rate": 2.4462809917355375e-06,
+      "loss": 0.4922,
       "step": 1200
     },
     {
       "epoch": 4.83271375464684,
-      "grad_norm": 6.915247917175293,
-      "learning_rate": 3.345724907063197e-07,
-      "loss": 0.6915,
+      "grad_norm": 7.6533989906311035,
+      "learning_rate": 7.933884297520662e-07,
+      "loss": 0.4799,
       "step": 1300
     },
     {
       "epoch": 5.0,
-      "eval_accuracy": 0.7772225450319582,
-      "eval_loss": 0.7830244302749634,
-      "eval_runtime": 9.0703,
-      "eval_samples_per_second": 948.705,
-      "eval_steps_per_second": 118.63,
+      "eval_accuracy": 0.7940732132481115,
+      "eval_loss": 0.7142194509506226,
+      "eval_runtime": 9.2321,
+      "eval_samples_per_second": 932.07,
+      "eval_steps_per_second": 7.366,
       "step": 1345
     },
     {
       "epoch": 5.0,
       "step": 1345,
       "total_flos": 4.52871045169152e+16,
-      "train_loss": 0.6343187750493727,
-      "train_runtime": 497.1667,
-      "train_samples_per_second": 346.162,
-      "train_steps_per_second": 2.705
+      "train_loss": 0.846687467213457,
+      "train_runtime": 657.9188,
+      "train_samples_per_second": 261.582,
+      "train_steps_per_second": 2.044
     }
   ],
   "logging_steps": 100,
@@ -172,7 +172,7 @@
     }
   },
   "total_flos": 4.52871045169152e+16,
-  "train_batch_size": 64,
+  "train_batch_size": 128,
   "trial_name": null,
   "trial_params": null
 }