hpc model

README.md

@@ -1,45 +1,97 @@
 ---
+datasets:
+- BramVanroy/hebban-reviews
 language:
 - nl
 license: mit
-tags:
-- sentiment-analysis
-- dutch
-- text
-datasets:
-- BramVanroy/hebban-reviews
 metrics:
 - accuracy
 - f1
 - precision
+- qwk
 - recall
-widget:
-- text: "Wauw, wat een leuk boek! Ik heb me er er goed mee vermaakt."
-- text: "Nee, deze vond ik niet goed. De auteur doet zijn best om je als lezer mee te trekken in het verhaal maar mij overtuigt het alleszins niet."
-- text: "Ik vind het niet slecht maar de schrijfstijl trekt me ook niet echt aan. Het wordt een beetje saai vanaf het vijfde hoofdstuk"
-
 model-index:
 - name: bert-base-dutch-cased-hebban-reviews
   results:
-  - task:
-      type: text-classification
-      name: sentiment analysis
-    dataset:
-      type: BramVanroy/hebban-reviews     
-      name: Hebban Reviews
+  - dataset:
+      config: filtered_sentiment
+      name: BramVanroy/hebban-reviews - filtered_sentiment - 2.0.0
+      revision: 2.0.0
       split: test
-      revision: 1.1.0
+      type: BramVanroy/hebban-reviews
     metrics:
-      - type: accuracy
-        value: 0.8219
-        name: Test accuracy
-      - type: f1
-        value: 0.8227
-        name: Test f1
-      - type: precision
-        value: 0.8245
-        name: Test precision
-      - type: recall
-        value: 0.8219
-        name: Test recall
----
+    - name: Test accuracy
+      type: accuracy
+      value: 0.8042406311637081
+    - name: Test f1
+      type: f1
+      value: 0.8125977499178383
+    - name: Test precision
+      type: precision
+      value: 0.8283602308368182
+    - name: Test qwk
+      type: qwk
+      value: 0.7301452890386257
+    - name: Test recall
+      type: recall
+      value: 0.8042406311637081
+tags:
+- sentiment-analysis
+- dutch
+- text
+widget:
+- text: Wauw, wat een leuk boek! Ik heb me er er goed mee vermaakt.
+- text: Nee, deze vond ik niet goed. De auteur doet zijn best om je als lezer mee
+    te trekken in het verhaal maar mij overtuigt het alleszins niet.
+- text: Ik vind het niet slecht maar de schrijfstijl trekt me ook niet echt aan. Het
+    wordt een beetje saai vanaf het vijfde hoofdstuk
+---
+
+# bert-base-dutch-cased-hebban-reviews
+
+# Dataset
+- dataset_name: BramVanroy/hebban-reviews
+- dataset_config: filtered_sentiment
+- dataset_revision: 2.0.0
+- labelcolumn: review_sentiment
+- textcolumn: review_text_without_quotes
+
+# Training
+- optim: adamw_hf
+- learning_rate: 5e-05
+- per_device_train_batch_size: 64
+- per_device_eval_batch_size: 64
+- gradient_accumulation_steps: 1
+- max_steps: 5001
+- save_steps: 500
+- metric_for_best_model: qwk
+
+# Best checkedpoint based on validation
+- best_metric: 0.732569302631819
+- best_model_checkpoint: trained/hebban-reviews/bert-base-dutch-cased/checkpoint-3000
+
+# Test results of best checkpoint
+- accuracy: 0.8042406311637081
+- f1: 0.8125977499178383
+- precision: 0.8283602308368182
+- qwk: 0.7301452890386257
+- recall: 0.8042406311637081
+
+## Confusion matric
+
+![cfm](fig/test_confusion_matrix.png)
+
+## Normalized confusion matrix
+
+![norm cfm](fig/test_confusion_matrix_norm.png)
+
+# Environment
+- cuda_capabilities: 8.0; 8.0
+- cuda_device_count: 2
+- cuda_devices: NVIDIA A100-SXM4-80GB; NVIDIA A100-SXM4-80GB
+- finetuner_commit: 48bb3434fa8bbfc9b2d0061ca6c8feb87f78a7ef
+- platform: Linux-4.18.0-305.49.1.el8_4.x86_64-x86_64-with-glibc2.28
+- python_version: 3.9.5
+- toch_version: 1.10.0
+- transformers_version: 4.21.0
+

all_results.json

@@ -0,0 +1,12 @@
+{
+    "accuracy": 0.8042406311637081,
+    "epoch": 4.38,
+    "f1": 0.8125977499178383,
+    "precision": 0.8283602308368182,
+    "qwk": 0.7301452890386257,
+    "recall": 0.8042406311637081,
+    "train_loss": 0.3670836862052257,
+    "train_runtime": 2628.6927,
+    "train_samples_per_second": 243.516,
+    "train_steps_per_second": 1.902
+}

args.json

@@ -0,0 +1,134 @@
+{
+    "_n_gpu": 1,
+    "adafactor": false,
+    "adam_beta1": 0.9,
+    "adam_beta2": 0.999,
+    "adam_epsilon": 1e-08,
+    "auto_find_batch_size": false,
+    "bf16": false,
+    "bf16_full_eval": false,
+    "calculate_qwk": true,
+    "data_seed": 42,
+    "dataloader_drop_last": false,
+    "dataloader_num_workers": 0,
+    "dataloader_pin_memory": true,
+    "dataset_config": "filtered_sentiment",
+    "dataset_name": "BramVanroy/hebban-reviews",
+    "dataset_revision": "2.0.0",
+    "ddp_bucket_cap_mb": null,
+    "ddp_find_unused_parameters": null,
+    "debug": [],
+    "deepspeed": null,
+    "disable_tqdm": false,
+    "do_early_stopping": false,
+    "do_eval": true,
+    "do_optimize": false,
+    "do_predict": true,
+    "do_train": true,
+    "early_stopping_patience": 1,
+    "early_stopping_threshold": 0.0,
+    "eval_accumulation_steps": null,
+    "eval_delay": 0,
+    "eval_steps": 500,
+    "evaluation_strategy": "steps",
+    "fp16": true,
+    "fp16_backend": "auto",
+    "fp16_full_eval": false,
+    "fp16_opt_level": "O1",
+    "fsdp": [],
+    "fsdp_min_num_params": 0,
+    "fsdp_transformer_layer_cls_to_wrap": null,
+    "full_determinism": false,
+    "gradient_accumulation_steps": 1,
+    "gradient_checkpointing": false,
+    "greater_is_better": true,
+    "group_by_length": false,
+    "half_precision_backend": "cuda_amp",
+    "hub_model_id": null,
+    "hub_private_repo": false,
+    "hub_strategy": "every_save",
+    "hub_token": null,
+    "ignore_data_skip": false,
+    "include_inputs_for_metrics": false,
+    "jit_mode_eval": false,
+    "label_names": null,
+    "label_smoothing_factor": 0.0,
+    "labelcolumn": "review_sentiment",
+    "labelnames": [
+        "negative",
+        "neutral",
+        "positive"
+    ],
+    "learning_rate": 5e-05,
+    "length_column_name": "length",
+    "load_best_model_at_end": true,
+    "local_rank": 0,
+    "log_level": -1,
+    "log_level_replica": -1,
+    "log_on_each_node": true,
+    "logging_dir": "trained/hebban-reviews/bert-base-dutch-cased/runs/Jul28_16-04-25_node3900.accelgor.os",
+    "logging_first_step": false,
+    "logging_nan_inf_filter": true,
+    "logging_steps": 500,
+    "logging_strategy": "steps",
+    "lr_scheduler_type": "linear",
+    "max_grad_norm": 1.0,
+    "max_seq_length": null,
+    "max_steps": 5001,
+    "max_test_samples": null,
+    "max_train_samples": null,
+    "max_validation_samples": null,
+    "metric_for_best_model": "qwk",
+    "model_name_or_path": "GroNLP/bert-base-dutch-cased",
+    "model_revision": "main",
+    "mp_parameters": "",
+    "n_trials": 8,
+    "no_cuda": false,
+    "num_train_epochs": 3.0,
+    "optim": "adamw_hf",
+    "output_dir": "trained/hebban-reviews/bert-base-dutch-cased",
+    "overwrite_cache": false,
+    "overwrite_output_dir": true,
+    "past_index": -1,
+    "per_device_eval_batch_size": 64,
+    "per_device_train_batch_size": 64,
+    "per_gpu_eval_batch_size": null,
+    "per_gpu_train_batch_size": null,
+    "prediction_loss_only": false,
+    "push_to_hub": false,
+    "push_to_hub_model_id": null,
+    "push_to_hub_organization": null,
+    "push_to_hub_token": null,
+    "ray_scope": "all",
+    "remove_unused_columns": true,
+    "report_to": [
+        "tensorboard"
+    ],
+    "resume_from_checkpoint": null,
+    "run_name": "trained/hebban-reviews/bert-base-dutch-cased",
+    "save_on_each_node": false,
+    "save_steps": 500,
+    "save_strategy": "steps",
+    "save_total_limit": null,
+    "scheduler_type": null,
+    "seed": 42,
+    "sharded_ddp": [],
+    "skip_memory_metrics": true,
+    "split_seed": 42,
+    "testsplit_name": "test",
+    "textcolumn": "review_text_without_quotes",
+    "tf32": null,
+    "torchdynamo": null,
+    "tpu_metrics_debug": false,
+    "tpu_num_cores": null,
+    "trainsplit_name": "train",
+    "use_class_weights": true,
+    "use_ipex": false,
+    "use_legacy_prediction_loop": false,
+    "validation_size": 0.1,
+    "validationsplit_name": "validation",
+    "warmup_ratio": 0.0,
+    "warmup_steps": 0,
+    "weight_decay": 0.0,
+    "xpu_backend": null
+}

config.json

@@ -29,7 +29,7 @@
   "pad_token_id": 3,
   "position_embedding_type": "absolute",
   "torch_dtype": "float32",
-  "transformers_version": "4.21.0.dev0",
+  "transformers_version": "4.21.0",
   "type_vocab_size": 2,
   "use_cache": true,
   "vocab_size": 30073

env.json

@@ -0,0 +1,10 @@
+{
+    "cuda_capabilities": "8.0; 8.0",
+    "cuda_device_count": 2,
+    "cuda_devices": "NVIDIA A100-SXM4-80GB; NVIDIA A100-SXM4-80GB",
+    "finetuner_commit": "48bb3434fa8bbfc9b2d0061ca6c8feb87f78a7ef",
+    "platform": "Linux-4.18.0-305.49.1.el8_4.x86_64-x86_64-with-glibc2.28",
+    "python_version": "3.9.5",
+    "toch_version": "1.10.0",
+    "transformers_version": "4.21.0"
+}

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c24b6e7600731f4f827d6ebf7c530b3919a7fcdc81f0ec28aab9b152f169cb40
+oid sha256:59583fc84aa34c92c26d33c8977e280946522eb61277a2cb5b308fd0125d9a83
 size 436629869

test_predictions.txt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92fd7af07a226b247dfc5c998b14bbdf422497447eedd8467b76697dde323ae8
+size 65400677

test_results.json

@@ -0,0 +1,7 @@
+{
+    "accuracy": 0.8042406311637081,
+    "f1": 0.8125977499178383,
+    "precision": 0.8283602308368182,
+    "qwk": 0.7301452890386257,
+    "recall": 0.8042406311637081
+}

tokenizer_config.json

@@ -8,7 +8,7 @@
   "never_split": null,
   "pad_token": "[PAD]",
   "sep_token": "[SEP]",
-  "special_tokens_map_file": "/home/bram/.cache/huggingface/transformers/adb82a117c09b0f8768357de8e836a9e0610730782f82edc49dd0020c48f1d03.dd8bd9bfd3664b530ea4e645105f557769387b3da9f79bdb55ed556bdd80611d",
+  "special_tokens_map_file": "/data/gent/vo/000/gvo00042/vsc42515/cache/transformers/adb82a117c09b0f8768357de8e836a9e0610730782f82edc49dd0020c48f1d03.dd8bd9bfd3664b530ea4e645105f557769387b3da9f79bdb55ed556bdd80611d",
   "strip_accents": null,
   "tokenize_chinese_chars": true,
   "tokenizer_class": "BertTokenizer",

train_results.json

@@ -0,0 +1,7 @@
+{
+    "epoch": 4.38,
+    "train_loss": 0.3670836862052257,
+    "train_runtime": 2628.6927,
+    "train_samples_per_second": 243.516,
+    "train_steps_per_second": 1.902
+}

trainer_state.json

@@ -1,457 +1,215 @@
 {
-  "best_metric": 0.8199536621312183,
-  "best_model_checkpoint": "/home/bram/shares/predict/trained/dutch/hebban-reviews/bert-base-dutch-cased/checkpoint-11000",
-  "epoch": 3.9447731755424065,
-  "global_step": 12000,
+  "best_metric": 0.732569302631819,
+  "best_model_checkpoint": "trained/hebban-reviews/bert-base-dutch-cased/checkpoint-3000",
+  "epoch": 4.382997370727432,
+  "global_step": 5001,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.16,
-      "learning_rate": 4.792083333333334e-05,
-      "loss": 0.7371,
+      "epoch": 0.44,
+      "learning_rate": 4.5000999800039995e-05,
+      "loss": 0.6824,
       "step": 500
     },
     {
-      "epoch": 0.16,
-      "eval_accuracy": 0.7546844181459567,
-      "eval_f1": 0.7629915603619231,
-      "eval_loss": 0.6527217626571655,
-      "eval_precision": 0.7780177612862748,
-      "eval_recall": 0.7546844181459567,
-      "eval_runtime": 24.4866,
-      "eval_samples_per_second": 662.566,
-      "eval_steps_per_second": 27.607,
+      "epoch": 0.44,
+      "eval_accuracy": 0.7763806706114399,
+      "eval_f1": 0.7835745137864604,
+      "eval_loss": 0.602776288986206,
+      "eval_precision": 0.8014359034966511,
+      "eval_qwk": 0.6748032105430382,
+      "eval_recall": 0.7763806706114399,
+      "eval_runtime": 23.7871,
+      "eval_samples_per_second": 682.05,
+      "eval_steps_per_second": 5.339,
       "step": 500
     },
     {
-      "epoch": 0.33,
-      "learning_rate": 4.58375e-05,
-      "loss": 0.6457,
+      "epoch": 0.88,
+      "learning_rate": 4.000199960007999e-05,
+      "loss": 0.5903,
       "step": 1000
     },
     {
-      "epoch": 0.33,
-      "eval_accuracy": 0.7643614398422091,
-      "eval_f1": 0.7769010779814548,
-      "eval_loss": 0.6216204166412354,
-      "eval_precision": 0.8017754901402682,
-      "eval_recall": 0.7643614398422091,
-      "eval_runtime": 24.6912,
-      "eval_samples_per_second": 657.077,
-      "eval_steps_per_second": 27.378,
+      "epoch": 0.88,
+      "eval_accuracy": 0.7441444773175543,
+      "eval_f1": 0.7623993009582866,
+      "eval_loss": 0.5676863789558411,
+      "eval_precision": 0.8079269153410689,
+      "eval_qwk": 0.6673976411458487,
+      "eval_recall": 0.7441444773175543,
+      "eval_runtime": 23.3353,
+      "eval_samples_per_second": 695.257,
+      "eval_steps_per_second": 5.442,
       "step": 1000
     },
     {
-      "epoch": 0.49,
-      "learning_rate": 4.375416666666667e-05,
-      "loss": 0.6184,
+      "epoch": 1.31,
+      "learning_rate": 3.50129974005199e-05,
+      "loss": 0.487,
       "step": 1500
     },
     {
-      "epoch": 0.49,
-      "eval_accuracy": 0.7790310650887574,
-      "eval_f1": 0.789661025567087,
-      "eval_loss": 0.6005069017410278,
-      "eval_precision": 0.8094066060995802,
-      "eval_recall": 0.7790310650887574,
-      "eval_runtime": 24.5988,
-      "eval_samples_per_second": 659.544,
-      "eval_steps_per_second": 27.481,
+      "epoch": 1.31,
+      "eval_accuracy": 0.7938239644970414,
+      "eval_f1": 0.8033347539691899,
+      "eval_loss": 0.5801703333854675,
+      "eval_precision": 0.8206089090362688,
+      "eval_qwk": 0.7239594229437711,
+      "eval_recall": 0.7938239644970414,
+      "eval_runtime": 23.5912,
+      "eval_samples_per_second": 687.714,
+      "eval_steps_per_second": 5.383,
       "step": 1500
     },
     {
-      "epoch": 0.66,
-      "learning_rate": 4.1670833333333334e-05,
-      "loss": 0.6042,
+      "epoch": 1.75,
+      "learning_rate": 3.001399720055989e-05,
+      "loss": 0.4465,
       "step": 2000
     },
     {
-      "epoch": 0.66,
-      "eval_accuracy": 0.8077539447731755,
-      "eval_f1": 0.8118714417209679,
-      "eval_loss": 0.5989466309547424,
-      "eval_precision": 0.8220143498287384,
-      "eval_recall": 0.8077539447731755,
-      "eval_runtime": 24.5976,
-      "eval_samples_per_second": 659.577,
-      "eval_steps_per_second": 27.482,
+      "epoch": 1.75,
+      "eval_accuracy": 0.7809418145956607,
+      "eval_f1": 0.7939963208456309,
+      "eval_loss": 0.5808519721031189,
+      "eval_precision": 0.8222816466721266,
+      "eval_qwk": 0.7118880899721043,
+      "eval_recall": 0.7809418145956607,
+      "eval_runtime": 23.7528,
+      "eval_samples_per_second": 683.034,
+      "eval_steps_per_second": 5.347,
       "step": 2000
     },
     {
-      "epoch": 0.82,
-      "learning_rate": 3.959166666666667e-05,
-      "loss": 0.5744,
+      "epoch": 2.19,
+      "learning_rate": 2.501499700059988e-05,
+      "loss": 0.3808,
       "step": 2500
     },
     {
-      "epoch": 0.82,
-      "eval_accuracy": 0.7972140039447732,
-      "eval_f1": 0.8043419281165946,
-      "eval_loss": 0.5735336542129517,
-      "eval_precision": 0.8156333669948874,
-      "eval_recall": 0.7972140039447732,
-      "eval_runtime": 24.5948,
-      "eval_samples_per_second": 659.651,
-      "eval_steps_per_second": 27.485,
+      "epoch": 2.19,
+      "eval_accuracy": 0.8011587771203156,
+      "eval_f1": 0.8108361051205994,
+      "eval_loss": 0.7650117874145508,
+      "eval_precision": 0.8317278803823261,
+      "eval_qwk": 0.7224991347098964,
+      "eval_recall": 0.8011587771203156,
+      "eval_runtime": 23.668,
+      "eval_samples_per_second": 685.482,
+      "eval_steps_per_second": 5.366,
       "step": 2500
     },
     {
-      "epoch": 0.99,
-      "learning_rate": 3.7512500000000004e-05,
-      "loss": 0.5833,
+      "epoch": 2.63,
+      "learning_rate": 2.001599680063987e-05,
+      "loss": 0.2909,
       "step": 3000
     },
     {
-      "epoch": 0.99,
-      "eval_accuracy": 0.8060281065088757,
-      "eval_f1": 0.8074379985793598,
-      "eval_loss": 0.5610195994377136,
-      "eval_precision": 0.8123421315309852,
-      "eval_recall": 0.8060281065088757,
-      "eval_runtime": 24.5935,
-      "eval_samples_per_second": 659.686,
-      "eval_steps_per_second": 27.487,
+      "epoch": 2.63,
+      "eval_accuracy": 0.8086168639053254,
+      "eval_f1": 0.8166494576368075,
+      "eval_loss": 0.7910537719726562,
+      "eval_precision": 0.8320947395056137,
+      "eval_qwk": 0.732569302631819,
+      "eval_recall": 0.8086168639053254,
+      "eval_runtime": 23.5554,
+      "eval_samples_per_second": 688.761,
+      "eval_steps_per_second": 5.392,
       "step": 3000
     },
     {
-      "epoch": 1.15,
-      "learning_rate": 3.542916666666667e-05,
-      "loss": 0.4633,
+      "epoch": 3.07,
+      "learning_rate": 1.5026994601079786e-05,
+      "loss": 0.2746,
       "step": 3500
     },
     {
-      "epoch": 1.15,
-      "eval_accuracy": 0.8055966469428008,
-      "eval_f1": 0.8112577676458685,
-      "eval_loss": 0.5967000126838684,
-      "eval_precision": 0.8196531984723394,
-      "eval_recall": 0.8055966469428008,
-      "eval_runtime": 24.7636,
-      "eval_samples_per_second": 655.155,
-      "eval_steps_per_second": 27.298,
+      "epoch": 3.07,
+      "eval_accuracy": 0.7985083826429981,
+      "eval_f1": 0.8084609549617363,
+      "eval_loss": 0.9503954648971558,
+      "eval_precision": 0.8285545185236465,
+      "eval_qwk": 0.7236034727222331,
+      "eval_recall": 0.7985083826429981,
+      "eval_runtime": 23.5785,
+      "eval_samples_per_second": 688.084,
+      "eval_steps_per_second": 5.386,
       "step": 3500
     },
     {
-      "epoch": 1.31,
-      "learning_rate": 3.3345833333333335e-05,
-      "loss": 0.4473,
+      "epoch": 3.51,
+      "learning_rate": 1.0027994401119777e-05,
+      "loss": 0.1939,
       "step": 4000
     },
     {
-      "epoch": 1.31,
-      "eval_accuracy": 0.8108357988165681,
-      "eval_f1": 0.8144157036046853,
-      "eval_loss": 0.6390397548675537,
-      "eval_precision": 0.8190905659244092,
-      "eval_recall": 0.8108357988165681,
-      "eval_runtime": 24.7631,
-      "eval_samples_per_second": 655.169,
-      "eval_steps_per_second": 27.299,
+      "epoch": 3.51,
+      "eval_accuracy": 0.7966592702169625,
+      "eval_f1": 0.8064656300991808,
+      "eval_loss": 0.9597578048706055,
+      "eval_precision": 0.8249666716932728,
+      "eval_qwk": 0.7250412688400546,
+      "eval_recall": 0.7966592702169625,
+      "eval_runtime": 23.6525,
+      "eval_samples_per_second": 685.931,
+      "eval_steps_per_second": 5.369,
       "step": 4000
     },
     {
-      "epoch": 1.48,
-      "learning_rate": 3.12625e-05,
-      "loss": 0.4493,
+      "epoch": 3.94,
+      "learning_rate": 5.028994201159768e-06,
+      "loss": 0.1824,
       "step": 4500
     },
     {
-      "epoch": 1.48,
-      "eval_accuracy": 0.7642998027613412,
-      "eval_f1": 0.780188565362406,
-      "eval_loss": 0.6074336171150208,
-      "eval_precision": 0.818173356753982,
-      "eval_recall": 0.7642998027613412,
-      "eval_runtime": 24.6961,
-      "eval_samples_per_second": 656.945,
-      "eval_steps_per_second": 27.373,
+      "epoch": 3.94,
+      "eval_accuracy": 0.8023298816568047,
+      "eval_f1": 0.810968480495596,
+      "eval_loss": 1.060992956161499,
+      "eval_precision": 0.8277405554545381,
+      "eval_qwk": 0.7252407399333649,
+      "eval_recall": 0.8023298816568047,
+      "eval_runtime": 23.7166,
+      "eval_samples_per_second": 684.077,
+      "eval_steps_per_second": 5.355,
       "step": 4500
     },
     {
-      "epoch": 1.64,
-      "learning_rate": 2.917916666666667e-05,
-      "loss": 0.4402,
+      "epoch": 4.38,
+      "learning_rate": 2.999400119976005e-08,
+      "loss": 0.1426,
       "step": 5000
     },
     {
-      "epoch": 1.64,
-      "eval_accuracy": 0.817430966469428,
-      "eval_f1": 0.818650395547064,
-      "eval_loss": 0.6175166368484497,
-      "eval_precision": 0.8200081119899918,
-      "eval_recall": 0.817430966469428,
-      "eval_runtime": 24.5978,
-      "eval_samples_per_second": 659.57,
-      "eval_steps_per_second": 27.482,
+      "epoch": 4.38,
+      "eval_accuracy": 0.8090483234714004,
+      "eval_f1": 0.8157426407187547,
+      "eval_loss": 1.1207919120788574,
+      "eval_precision": 0.8273912560003507,
+      "eval_qwk": 0.7316045159099166,
+      "eval_recall": 0.8090483234714004,
+      "eval_runtime": 23.6323,
+      "eval_samples_per_second": 686.518,
+      "eval_steps_per_second": 5.374,
       "step": 5000
     },
     {
-      "epoch": 1.81,
-      "learning_rate": 2.7100000000000005e-05,
-      "loss": 0.4406,
-      "step": 5500
-    },
-    {
-      "epoch": 1.81,
-      "eval_accuracy": 0.8167529585798816,
-      "eval_f1": 0.8194587497642177,
-      "eval_loss": 0.6328938603401184,
-      "eval_precision": 0.8229261899048395,
-      "eval_recall": 0.8167529585798816,
-      "eval_runtime": 24.611,
-      "eval_samples_per_second": 659.218,
-      "eval_steps_per_second": 27.467,
-      "step": 5500
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 2.5016666666666667e-05,
-      "loss": 0.4457,
-      "step": 6000
-    },
-    {
-      "epoch": 1.97,
-      "eval_accuracy": 0.8178007889546351,
-      "eval_f1": 0.8196437981366307,
-      "eval_loss": 0.6221566796302795,
-      "eval_precision": 0.8219923397562409,
-      "eval_recall": 0.8178007889546351,
-      "eval_runtime": 24.6955,
-      "eval_samples_per_second": 656.961,
-      "eval_steps_per_second": 27.373,
-      "step": 6000
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 2.2933333333333333e-05,
-      "loss": 0.3013,
-      "step": 6500
-    },
-    {
-      "epoch": 2.14,
-      "eval_accuracy": 0.8099112426035503,
-      "eval_f1": 0.8154916160650456,
-      "eval_loss": 0.7676593661308289,
-      "eval_precision": 0.8240038245839082,
-      "eval_recall": 0.8099112426035503,
-      "eval_runtime": 24.7339,
-      "eval_samples_per_second": 655.941,
-      "eval_steps_per_second": 27.331,
-      "step": 6500
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 2.0854166666666668e-05,
-      "loss": 0.2875,
-      "step": 7000
-    },
-    {
-      "epoch": 2.3,
-      "eval_accuracy": 0.8025764299802761,
-      "eval_f1": 0.8097813881432507,
-      "eval_loss": 0.7549890279769897,
-      "eval_precision": 0.8215043597038734,
-      "eval_recall": 0.8025764299802761,
-      "eval_runtime": 24.7599,
-      "eval_samples_per_second": 655.252,
-      "eval_steps_per_second": 27.302,
-      "step": 7000
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 1.8770833333333333e-05,
-      "loss": 0.2804,
-      "step": 7500
-    },
-    {
-      "epoch": 2.47,
-      "eval_accuracy": 0.8102194280078896,
-      "eval_f1": 0.8148870229811181,
-      "eval_loss": 0.804236114025116,
-      "eval_precision": 0.8221239914552869,
-      "eval_recall": 0.8102194280078896,
-      "eval_runtime": 24.6532,
-      "eval_samples_per_second": 658.09,
-      "eval_steps_per_second": 27.42,
-      "step": 7500
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 1.66875e-05,
-      "loss": 0.2784,
-      "step": 8000
-    },
-    {
-      "epoch": 2.63,
-      "eval_accuracy": 0.8136711045364892,
-      "eval_f1": 0.8173932526970648,
-      "eval_loss": 0.8103494644165039,
-      "eval_precision": 0.8223803033268108,
-      "eval_recall": 0.8136711045364892,
-      "eval_runtime": 24.5862,
-      "eval_samples_per_second": 659.883,
-      "eval_steps_per_second": 27.495,
-      "step": 8000
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.4604166666666666e-05,
-      "loss": 0.275,
-      "step": 8500
-    },
-    {
-      "epoch": 2.79,
-      "eval_accuracy": 0.8132396449704142,
-      "eval_f1": 0.8172151251977423,
-      "eval_loss": 0.8132022619247437,
-      "eval_precision": 0.8227088066747076,
-      "eval_recall": 0.8132396449704142,
-      "eval_runtime": 24.5279,
-      "eval_samples_per_second": 661.452,
-      "eval_steps_per_second": 27.56,
-      "step": 8500
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 1.2520833333333334e-05,
-      "loss": 0.2651,
-      "step": 9000
-    },
-    {
-      "epoch": 2.96,
-      "eval_accuracy": 0.8065212031558185,
-      "eval_f1": 0.8120431442321697,
-      "eval_loss": 0.7826108336448669,
-      "eval_precision": 0.8202960014382004,
-      "eval_recall": 0.8065212031558185,
-      "eval_runtime": 24.7019,
-      "eval_samples_per_second": 656.791,
-      "eval_steps_per_second": 27.366,
-      "step": 9000
-    },
-    {
-      "epoch": 3.12,
-      "learning_rate": 1.0441666666666667e-05,
-      "loss": 0.1855,
-      "step": 9500
-    },
-    {
-      "epoch": 3.12,
-      "eval_accuracy": 0.8140409270216963,
-      "eval_f1": 0.8173959290765784,
-      "eval_loss": 1.0370513200759888,
-      "eval_precision": 0.8233538681349268,
-      "eval_recall": 0.8140409270216963,
-      "eval_runtime": 24.6749,
-      "eval_samples_per_second": 657.51,
-      "eval_steps_per_second": 27.396,
-      "step": 9500
-    },
-    {
-      "epoch": 3.29,
-      "learning_rate": 8.358333333333333e-06,
-      "loss": 0.1571,
-      "step": 10000
-    },
-    {
-      "epoch": 3.29,
-      "eval_accuracy": 0.8128081854043393,
-      "eval_f1": 0.8173020798805183,
-      "eval_loss": 1.036438226699829,
-      "eval_precision": 0.8246320290318612,
-      "eval_recall": 0.8128081854043393,
-      "eval_runtime": 24.6314,
-      "eval_samples_per_second": 658.672,
-      "eval_steps_per_second": 27.445,
-      "step": 10000
-    },
-    {
-      "epoch": 3.45,
-      "learning_rate": 6.275e-06,
-      "loss": 0.1608,
-      "step": 10500
-    },
-    {
-      "epoch": 3.45,
-      "eval_accuracy": 0.8108357988165681,
-      "eval_f1": 0.8154023680396078,
-      "eval_loss": 1.0521161556243896,
-      "eval_precision": 0.8221458981635473,
-      "eval_recall": 0.8108357988165681,
-      "eval_runtime": 24.6401,
-      "eval_samples_per_second": 658.44,
-      "eval_steps_per_second": 27.435,
-      "step": 10500
-    },
-    {
-      "epoch": 3.62,
-      "learning_rate": 4.191666666666667e-06,
-      "loss": 0.1542,
-      "step": 11000
-    },
-    {
-      "epoch": 3.62,
-      "eval_accuracy": 0.8197115384615384,
-      "eval_f1": 0.8199536621312183,
-      "eval_loss": 1.108322262763977,
-      "eval_precision": 0.8205846609719522,
-      "eval_recall": 0.8197115384615384,
-      "eval_runtime": 24.5935,
-      "eval_samples_per_second": 659.687,
-      "eval_steps_per_second": 27.487,
-      "step": 11000
-    },
-    {
-      "epoch": 3.78,
-      "learning_rate": 2.108333333333333e-06,
-      "loss": 0.1561,
-      "step": 11500
-    },
-    {
-      "epoch": 3.78,
-      "eval_accuracy": 0.8138560157790927,
-      "eval_f1": 0.8166825272884544,
-      "eval_loss": 1.0635404586791992,
-      "eval_precision": 0.8205938348020483,
-      "eval_recall": 0.8138560157790927,
-      "eval_runtime": 24.8393,
-      "eval_samples_per_second": 653.159,
-      "eval_steps_per_second": 27.215,
-      "step": 11500
-    },
-    {
-      "epoch": 3.94,
-      "learning_rate": 2.916666666666667e-08,
-      "loss": 0.1415,
-      "step": 12000
-    },
-    {
-      "epoch": 3.94,
-      "eval_accuracy": 0.815396942800789,
-      "eval_f1": 0.8176978993348879,
-      "eval_loss": 1.0898783206939697,
-      "eval_precision": 0.8208248888308868,
-      "eval_recall": 0.815396942800789,
-      "eval_runtime": 24.6997,
-      "eval_samples_per_second": 656.85,
-      "eval_steps_per_second": 27.369,
-      "step": 12000
-    },
-    {
-      "epoch": 3.94,
-      "step": 12000,
-      "total_flos": 1.4832731619013427e+17,
-      "train_loss": 0.37884856541951495,
-      "train_runtime": 3446.9333,
-      "train_samples_per_second": 167.105,
-      "train_steps_per_second": 3.481
+      "epoch": 4.38,
+      "step": 5001,
+      "total_flos": 1.6838837781764506e+17,
+      "train_loss": 0.3670836862052257,
+      "train_runtime": 2628.6927,
+      "train_samples_per_second": 243.516,
+      "train_steps_per_second": 1.902
     }
   ],
-  "max_steps": 12000,
-  "num_train_epochs": 4,
-  "total_flos": 1.4832731619013427e+17,
+  "max_steps": 5001,
+  "num_train_epochs": 5,
+  "total_flos": 1.6838837781764506e+17,
   "trial_name": null,
   "trial_params": null
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e629a85d11014bed4fb4099bbcedecc04daead523e24b38fb4e2c394f89df7a2
+oid sha256:ff9696a0fda64c2065dfd77fec96118ef39d58e7575a456552455a8001fdc944
 size 3375