hpc model

README.md

@@ -1,45 +1,100 @@
 ---
+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-multilingual-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: 0.7845
-        name: Test accuracy
-      - type: f1
-        value: 0.7911
-        name: Test f1
-      - type: precision
-        value: 0.8013
-        name: Test precision
-      - type: recall
-        value: 0.7845
-        name: Test recall
----
+    - name: Test accuracy
+      type: accuracy
+      value: 0.7764792899408284
+    - name: Test f1
+      type: f1
+      value: 0.7821329848271866
+    - name: Test precision
+      type: precision
+      value: 0.7907660190770787
+    - name: Test qwk
+      type: qwk
+      value: 0.6813121109021326
+    - name: Test recall
+      type: recall
+      value: 0.7764792899408284
+    task:
+      name: sentiment analysis
+      type: text-classification
+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-multilingual-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: 0.6828581526810108
+- best_model_checkpoint: trained/hebban-reviews/bert-base-multilingual-cased/checkpoint-1500
+
+# Test results of best checkpoint
+- accuracy: 0.7764792899408284
+- f1: 0.7821329848271866
+- precision: 0.7907660190770787
+- qwk: 0.6813121109021326
+- recall: 0.7764792899408284
+
+## 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: 66294c815326c93682003119534cb72009f558c2
+- 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": 0.7764792899408284,
+    "epoch": 4.38,
+    "f1": 0.7821329848271866,
+    "precision": 0.7907660190770787,
+    "qwk": 0.6813121109021326,
+    "recall": 0.7764792899408284,
+    "train_loss": 0.5313903366928219,
+    "train_runtime": 2667.6757,
+    "train_samples_per_second": 239.957,
+    "train_steps_per_second": 1.875
+}

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-multilingual-cased/runs/Jul28_16-55-25_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": "bert-base-multilingual-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-multilingual-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-multilingual-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

@@ -34,7 +34,7 @@
   "pooler_type": "first_token_transform",
   "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": 119547

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": "66294c815326c93682003119534cb72009f558c2",
+    "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 sha256:8ed77b30f2e050c50b9b4c6041ea2f4e10e9f41413b5565af17f7fb5b86f4fb6
+oid sha256:2ff04f2b96bacd77224f0a8a0ad534c4b54e436c98b11016ba592265153a672b
 size 711493997

test_predictions.txt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b0bbeeab622e2b2ede61d8dcc4642168cb49c6cfc5bd32605aba9da0e7b9ad0
+size 65402577

test_results.json

@@ -0,0 +1,7 @@
+{
+    "accuracy": 0.7764792899408284,
+    "f1": 0.7821329848271866,
+    "precision": 0.7907660190770787,
+    "qwk": 0.6813121109021326,
+    "recall": 0.7764792899408284
+}

train_results.json

@@ -0,0 +1,7 @@
+{
+    "epoch": 4.38,
+    "train_loss": 0.5313903366928219,
+    "train_runtime": 2667.6757,
+    "train_samples_per_second": 239.957,
+    "train_steps_per_second": 1.875
+}

trainer_state.json

@@ -1,457 +1,215 @@
 {
-  "best_metric": 0.7886478688599108,
-  "best_model_checkpoint": "/home/bram/shares/predict/trained/dutch/hebban-reviews/bert-base-multilingual-cased/checkpoint-11000",
-  "epoch": 3.9447731755424065,
-  "global_step": 12000,
+  "best_metric": 0.6828581526810108,
+  "best_model_checkpoint": "trained/hebban-reviews/bert-base-multilingual-cased/checkpoint-1500",
+  "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": 4.793333333333334e-05,
-      "loss": 0.9625,
+      "epoch": 0.44,
+      "learning_rate": 4.504099180163968e-05,
+      "loss": 0.8498,
       "step": 500
     },
     {
-      "epoch": 0.16,
-      "eval_accuracy": 0.6706114398422091,
-      "eval_f1": 0.6931944548479266,
-      "eval_loss": 0.8083919882774353,
-      "eval_precision": 0.7343487735413422,
-      "eval_recall": 0.6706114398422091,
-      "eval_runtime": 24.6562,
-      "eval_samples_per_second": 658.009,
-      "eval_steps_per_second": 27.417,
+      "epoch": 0.44,
+      "eval_accuracy": 0.6671597633136095,
+      "eval_f1": 0.6926558253677099,
+      "eval_loss": 0.7364796996116638,
+      "eval_precision": 0.7520529030378761,
+      "eval_qwk": 0.5224288418460314,
+      "eval_recall": 0.6671597633136095,
+      "eval_runtime": 23.7829,
+      "eval_samples_per_second": 682.171,
+      "eval_steps_per_second": 5.34,
       "step": 500
     },
     {
-      "epoch": 0.33,
-      "learning_rate": 4.585e-05,
-      "loss": 0.8186,
+      "epoch": 0.88,
+      "learning_rate": 4.0041991601679665e-05,
+      "loss": 0.6918,
       "step": 1000
     },
     {
-      "epoch": 0.33,
-      "eval_accuracy": 0.6870685404339251,
-      "eval_f1": 0.7088564060365691,
-      "eval_loss": 0.7797691226005554,
-      "eval_precision": 0.753903679506183,
-      "eval_recall": 0.6870685404339251,
-      "eval_runtime": 24.786,
-      "eval_samples_per_second": 654.562,
-      "eval_steps_per_second": 27.273,
+      "epoch": 0.88,
+      "eval_accuracy": 0.7271942800788954,
+      "eval_f1": 0.7414418295002464,
+      "eval_loss": 0.6605744957923889,
+      "eval_precision": 0.7743449767911329,
+      "eval_qwk": 0.5976543204298063,
+      "eval_recall": 0.7271942800788954,
+      "eval_runtime": 23.6177,
+      "eval_samples_per_second": 686.943,
+      "eval_steps_per_second": 5.377,
       "step": 1000
     },
     {
-      "epoch": 0.49,
-      "learning_rate": 4.376666666666667e-05,
-      "loss": 0.7658,
+      "epoch": 1.31,
+      "learning_rate": 3.504299140171966e-05,
+      "loss": 0.621,
       "step": 1500
     },
     {
-      "epoch": 0.49,
-      "eval_accuracy": 0.7040803747534516,
-      "eval_f1": 0.7250925730375316,
-      "eval_loss": 0.7284208536148071,
-      "eval_precision": 0.777831708888876,
-      "eval_recall": 0.7040803747534516,
-      "eval_runtime": 24.7754,
-      "eval_samples_per_second": 654.842,
-      "eval_steps_per_second": 27.285,
+      "epoch": 1.31,
+      "eval_accuracy": 0.7785996055226825,
+      "eval_f1": 0.7848033042674865,
+      "eval_loss": 0.6530709266662598,
+      "eval_precision": 0.7942290484031055,
+      "eval_qwk": 0.6828581526810108,
+      "eval_recall": 0.7785996055226825,
+      "eval_runtime": 23.6972,
+      "eval_samples_per_second": 684.637,
+      "eval_steps_per_second": 5.359,
       "step": 1500
     },
     {
-      "epoch": 0.66,
-      "learning_rate": 4.1683333333333335e-05,
-      "loss": 0.7318,
+      "epoch": 1.75,
+      "learning_rate": 3.0043991201759648e-05,
+      "loss": 0.5936,
       "step": 2000
     },
     {
-      "epoch": 0.66,
-      "eval_accuracy": 0.7891395463510849,
-      "eval_f1": 0.7822709378126483,
-      "eval_loss": 0.7148900032043457,
-      "eval_precision": 0.7780706309716242,
-      "eval_recall": 0.7891395463510849,
-      "eval_runtime": 24.6279,
-      "eval_samples_per_second": 658.766,
-      "eval_steps_per_second": 27.449,
+      "epoch": 1.75,
+      "eval_accuracy": 0.7501849112426036,
+      "eval_f1": 0.7649773762321463,
+      "eval_loss": 0.6148771643638611,
+      "eval_precision": 0.7952109430358816,
+      "eval_qwk": 0.6613885538385067,
+      "eval_recall": 0.7501849112426036,
+      "eval_runtime": 23.8294,
+      "eval_samples_per_second": 680.838,
+      "eval_steps_per_second": 5.33,
       "step": 2000
     },
     {
-      "epoch": 0.82,
-      "learning_rate": 3.960000000000001e-05,
-      "loss": 0.6977,
+      "epoch": 2.19,
+      "learning_rate": 2.504499100179964e-05,
+      "loss": 0.5434,
       "step": 2500
     },
     {
-      "epoch": 0.82,
-      "eval_accuracy": 0.7249753451676528,
-      "eval_f1": 0.7429677966966411,
-      "eval_loss": 0.6928562521934509,
-      "eval_precision": 0.7807492770659993,
-      "eval_recall": 0.7249753451676528,
-      "eval_runtime": 24.6337,
-      "eval_samples_per_second": 658.61,
-      "eval_steps_per_second": 27.442,
+      "epoch": 2.19,
+      "eval_accuracy": 0.7548076923076923,
+      "eval_f1": 0.7709100781310829,
+      "eval_loss": 0.6947299242019653,
+      "eval_precision": 0.8080766037849636,
+      "eval_qwk": 0.6597495777741245,
+      "eval_recall": 0.7548076923076923,
+      "eval_runtime": 23.6425,
+      "eval_samples_per_second": 686.22,
+      "eval_steps_per_second": 5.372,
       "step": 2500
     },
     {
-      "epoch": 0.99,
-      "learning_rate": 3.7516666666666666e-05,
-      "loss": 0.7032,
+      "epoch": 2.63,
+      "learning_rate": 2.0045990801839633e-05,
+      "loss": 0.4808,
       "step": 3000
     },
     {
-      "epoch": 0.99,
-      "eval_accuracy": 0.7484590729783037,
-      "eval_f1": 0.7594497306385251,
-      "eval_loss": 0.6671503782272339,
-      "eval_precision": 0.7788847542194782,
-      "eval_recall": 0.7484590729783037,
-      "eval_runtime": 24.5732,
-      "eval_samples_per_second": 660.231,
-      "eval_steps_per_second": 27.51,
+      "epoch": 2.63,
+      "eval_accuracy": 0.7347140039447732,
+      "eval_f1": 0.7545089976363922,
+      "eval_loss": 0.645616352558136,
+      "eval_precision": 0.8030403955873724,
+      "eval_qwk": 0.6513487818562542,
+      "eval_recall": 0.7347140039447732,
+      "eval_runtime": 23.8892,
+      "eval_samples_per_second": 679.136,
+      "eval_steps_per_second": 5.316,
       "step": 3000
     },
     {
-      "epoch": 1.15,
-      "learning_rate": 3.543333333333333e-05,
-      "loss": 0.6227,
+      "epoch": 3.07,
+      "learning_rate": 1.5046990601879624e-05,
+      "loss": 0.4643,
       "step": 3500
     },
     {
-      "epoch": 1.15,
-      "eval_accuracy": 0.7747781065088757,
-      "eval_f1": 0.7788314002376725,
-      "eval_loss": 0.6905977129936218,
-      "eval_precision": 0.7840673104698108,
-      "eval_recall": 0.7747781065088757,
-      "eval_runtime": 24.7721,
-      "eval_samples_per_second": 654.931,
-      "eval_steps_per_second": 27.289,
+      "epoch": 3.07,
+      "eval_accuracy": 0.7427268244575936,
+      "eval_f1": 0.7611673940281857,
+      "eval_loss": 0.7119439244270325,
+      "eval_precision": 0.8053303788937228,
+      "eval_qwk": 0.6549274566801152,
+      "eval_recall": 0.7427268244575936,
+      "eval_runtime": 23.6156,
+      "eval_samples_per_second": 687.004,
+      "eval_steps_per_second": 5.378,
       "step": 3500
     },
     {
-      "epoch": 1.31,
-      "learning_rate": 3.3350000000000004e-05,
-      "loss": 0.6309,
+      "epoch": 3.51,
+      "learning_rate": 1.0047990401919617e-05,
+      "loss": 0.3854,
       "step": 4000
     },
     {
-      "epoch": 1.31,
-      "eval_accuracy": 0.7607864891518737,
-      "eval_f1": 0.7709664128632473,
-      "eval_loss": 0.6686555743217468,
-      "eval_precision": 0.7878354043029604,
-      "eval_recall": 0.7607864891518737,
-      "eval_runtime": 24.6117,
-      "eval_samples_per_second": 659.198,
-      "eval_steps_per_second": 27.467,
+      "epoch": 3.51,
+      "eval_accuracy": 0.7685527613412229,
+      "eval_f1": 0.7819273980616418,
+      "eval_loss": 0.7430822253227234,
+      "eval_precision": 0.8091264019259906,
+      "eval_qwk": 0.6812961348271991,
+      "eval_recall": 0.7685527613412229,
+      "eval_runtime": 23.6906,
+      "eval_samples_per_second": 684.828,
+      "eval_steps_per_second": 5.361,
       "step": 4000
     },
     {
-      "epoch": 1.48,
-      "learning_rate": 3.126666666666666e-05,
-      "loss": 0.6256,
+      "epoch": 3.94,
+      "learning_rate": 5.048990201959608e-06,
+      "loss": 0.3642,
       "step": 4500
     },
     {
-      "epoch": 1.48,
-      "eval_accuracy": 0.6715359960552268,
-      "eval_f1": 0.7001277240194975,
-      "eval_loss": 0.6691098809242249,
-      "eval_precision": 0.7948795089365344,
-      "eval_recall": 0.6715359960552268,
-      "eval_runtime": 24.6266,
-      "eval_samples_per_second": 658.801,
-      "eval_steps_per_second": 27.45,
+      "epoch": 3.94,
+      "eval_accuracy": 0.7693540433925049,
+      "eval_f1": 0.7827174333765768,
+      "eval_loss": 0.7832001447677612,
+      "eval_precision": 0.8104759531984178,
+      "eval_qwk": 0.6799731235431681,
+      "eval_recall": 0.7693540433925049,
+      "eval_runtime": 23.5733,
+      "eval_samples_per_second": 688.237,
+      "eval_steps_per_second": 5.387,
       "step": 4500
     },
     {
-      "epoch": 1.64,
-      "learning_rate": 2.91875e-05,
-      "loss": 0.6152,
+      "epoch": 4.38,
+      "learning_rate": 4.999000199960008e-08,
+      "loss": 0.3199,
       "step": 5000
     },
     {
-      "epoch": 1.64,
-      "eval_accuracy": 0.7810034516765286,
-      "eval_f1": 0.7815583569804468,
-      "eval_loss": 0.6452860832214355,
-      "eval_precision": 0.7840264023525906,
-      "eval_recall": 0.7810034516765286,
-      "eval_runtime": 24.5688,
-      "eval_samples_per_second": 660.348,
-      "eval_steps_per_second": 27.515,
+      "epoch": 4.38,
+      "eval_accuracy": 0.7682445759368837,
+      "eval_f1": 0.7820013322186519,
+      "eval_loss": 0.8646416664123535,
+      "eval_precision": 0.8117448424467829,
+      "eval_qwk": 0.6776167553082479,
+      "eval_recall": 0.7682445759368837,
+      "eval_runtime": 23.7003,
+      "eval_samples_per_second": 684.547,
+      "eval_steps_per_second": 5.359,
       "step": 5000
     },
     {
-      "epoch": 1.81,
-      "learning_rate": 2.710416666666667e-05,
-      "loss": 0.6068,
-      "step": 5500
-    },
-    {
-      "epoch": 1.81,
-      "eval_accuracy": 0.7764423076923077,
-      "eval_f1": 0.7843276562987542,
-      "eval_loss": 0.6676753759384155,
-      "eval_precision": 0.8001617824983851,
-      "eval_recall": 0.7764423076923077,
-      "eval_runtime": 24.7224,
-      "eval_samples_per_second": 656.247,
-      "eval_steps_per_second": 27.344,
-      "step": 5500
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 2.5020833333333333e-05,
-      "loss": 0.6016,
-      "step": 6000
-    },
-    {
-      "epoch": 1.97,
-      "eval_accuracy": 0.7220784023668639,
-      "eval_f1": 0.7405665635983271,
-      "eval_loss": 0.6461244821548462,
-      "eval_precision": 0.782308061609766,
-      "eval_recall": 0.7220784023668639,
-      "eval_runtime": 24.6746,
-      "eval_samples_per_second": 657.518,
-      "eval_steps_per_second": 27.397,
-      "step": 6000
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 2.2937500000000002e-05,
-      "loss": 0.521,
-      "step": 6500
-    },
-    {
-      "epoch": 2.14,
-      "eval_accuracy": 0.7281188362919132,
-      "eval_f1": 0.7466913370432308,
-      "eval_loss": 0.7071252465248108,
-      "eval_precision": 0.7878399669333115,
-      "eval_recall": 0.7281188362919132,
-      "eval_runtime": 24.7635,
-      "eval_samples_per_second": 655.158,
-      "eval_steps_per_second": 27.298,
-      "step": 6500
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 2.0854166666666668e-05,
-      "loss": 0.5137,
-      "step": 7000
-    },
-    {
-      "epoch": 2.3,
-      "eval_accuracy": 0.7346523668639053,
-      "eval_f1": 0.752395675119321,
-      "eval_loss": 0.6949610114097595,
-      "eval_precision": 0.7902262301474485,
-      "eval_recall": 0.7346523668639053,
-      "eval_runtime": 24.761,
-      "eval_samples_per_second": 655.225,
-      "eval_steps_per_second": 27.301,
-      "step": 7000
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 1.8770833333333333e-05,
-      "loss": 0.504,
-      "step": 7500
-    },
-    {
-      "epoch": 2.47,
-      "eval_accuracy": 0.7453772189349113,
-      "eval_f1": 0.7618113431887504,
-      "eval_loss": 0.6824287176132202,
-      "eval_precision": 0.7979845371909847,
-      "eval_recall": 0.7453772189349113,
-      "eval_runtime": 24.7586,
-      "eval_samples_per_second": 655.287,
-      "eval_steps_per_second": 27.304,
-      "step": 7500
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 1.6691666666666668e-05,
-      "loss": 0.5049,
-      "step": 8000
-    },
-    {
-      "epoch": 2.63,
-      "eval_accuracy": 0.7816814595660749,
-      "eval_f1": 0.7878816281164417,
-      "eval_loss": 0.6992799043655396,
-      "eval_precision": 0.7967192296633613,
-      "eval_recall": 0.7816814595660749,
-      "eval_runtime": 24.7558,
-      "eval_samples_per_second": 655.361,
-      "eval_steps_per_second": 27.307,
-      "step": 8000
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.4608333333333335e-05,
-      "loss": 0.4987,
-      "step": 8500
-    },
-    {
-      "epoch": 2.79,
-      "eval_accuracy": 0.7680596646942801,
-      "eval_f1": 0.779595614035326,
-      "eval_loss": 0.7158842086791992,
-      "eval_precision": 0.8023493022129313,
-      "eval_recall": 0.7680596646942801,
-      "eval_runtime": 24.6159,
-      "eval_samples_per_second": 659.087,
-      "eval_steps_per_second": 27.462,
-      "step": 8500
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 1.2525000000000001e-05,
-      "loss": 0.4978,
-      "step": 9000
-    },
-    {
-      "epoch": 2.96,
-      "eval_accuracy": 0.7757026627218935,
-      "eval_f1": 0.7846497142505,
-      "eval_loss": 0.6755026578903198,
-      "eval_precision": 0.7992058416315558,
-      "eval_recall": 0.7757026627218935,
-      "eval_runtime": 24.6076,
-      "eval_samples_per_second": 659.309,
-      "eval_steps_per_second": 27.471,
-      "step": 9000
-    },
-    {
-      "epoch": 3.12,
-      "learning_rate": 1.0441666666666667e-05,
-      "loss": 0.4236,
-      "step": 9500
-    },
-    {
-      "epoch": 3.12,
-      "eval_accuracy": 0.7638067061143984,
-      "eval_f1": 0.7757104932531658,
-      "eval_loss": 0.740465521812439,
-      "eval_precision": 0.7974750781708845,
-      "eval_recall": 0.7638067061143984,
-      "eval_runtime": 24.6338,
-      "eval_samples_per_second": 658.607,
-      "eval_steps_per_second": 27.442,
-      "step": 9500
-    },
-    {
-      "epoch": 3.29,
-      "learning_rate": 8.358333333333333e-06,
-      "loss": 0.3987,
-      "step": 10000
-    },
-    {
-      "epoch": 3.29,
-      "eval_accuracy": 0.7555473372781065,
-      "eval_f1": 0.7701303707262471,
-      "eval_loss": 0.7569716572761536,
-      "eval_precision": 0.80053210602621,
-      "eval_recall": 0.7555473372781065,
-      "eval_runtime": 24.6348,
-      "eval_samples_per_second": 658.58,
-      "eval_steps_per_second": 27.441,
-      "step": 10000
-    },
-    {
-      "epoch": 3.45,
-      "learning_rate": 6.275e-06,
-      "loss": 0.3914,
-      "step": 10500
-    },
-    {
-      "epoch": 3.45,
-      "eval_accuracy": 0.7660256410256411,
-      "eval_f1": 0.7781937770368635,
-      "eval_loss": 0.7907547354698181,
-      "eval_precision": 0.8023937992057022,
-      "eval_recall": 0.7660256410256411,
-      "eval_runtime": 24.7673,
-      "eval_samples_per_second": 655.057,
-      "eval_steps_per_second": 27.294,
-      "step": 10500
-    },
-    {
-      "epoch": 3.62,
-      "learning_rate": 4.195833333333334e-06,
-      "loss": 0.4007,
-      "step": 11000
-    },
-    {
-      "epoch": 3.62,
-      "eval_accuracy": 0.7821745562130178,
-      "eval_f1": 0.7886478688599108,
-      "eval_loss": 0.7599502801895142,
-      "eval_precision": 0.7983440482619892,
-      "eval_recall": 0.7821745562130178,
-      "eval_runtime": 24.6844,
-      "eval_samples_per_second": 657.258,
-      "eval_steps_per_second": 27.386,
-      "step": 11000
-    },
-    {
-      "epoch": 3.78,
-      "learning_rate": 2.1125e-06,
-      "loss": 0.3878,
-      "step": 11500
-    },
-    {
-      "epoch": 3.78,
-      "eval_accuracy": 0.7740384615384616,
-      "eval_f1": 0.7836077149829459,
-      "eval_loss": 0.7647958993911743,
-      "eval_precision": 0.8007153395014647,
-      "eval_recall": 0.7740384615384616,
-      "eval_runtime": 24.6189,
-      "eval_samples_per_second": 659.006,
-      "eval_steps_per_second": 27.459,
-      "step": 11500
-    },
-    {
-      "epoch": 3.94,
-      "learning_rate": 2.916666666666667e-08,
-      "loss": 0.3829,
-      "step": 12000
-    },
-    {
-      "epoch": 3.94,
-      "eval_accuracy": 0.7662721893491125,
-      "eval_f1": 0.7779778325177736,
-      "eval_loss": 0.752331018447876,
-      "eval_precision": 0.7995290654452684,
-      "eval_recall": 0.7662721893491125,
-      "eval_runtime": 24.6488,
-      "eval_samples_per_second": 658.206,
-      "eval_steps_per_second": 27.425,
-      "step": 12000
-    },
-    {
-      "epoch": 3.94,
-      "step": 12000,
-      "total_flos": 1.502407617014661e+17,
-      "train_loss": 0.5753211771647135,
-      "train_runtime": 3636.5588,
-      "train_samples_per_second": 158.391,
-      "train_steps_per_second": 3.3
+      "epoch": 4.38,
+      "step": 5001,
+      "total_flos": 1.6838837781764506e+17,
+      "train_loss": 0.5313903366928219,
+      "train_runtime": 2667.6757,
+      "train_samples_per_second": 239.957,
+      "train_steps_per_second": 1.875
     }
   ],
-  "max_steps": 12000,
-  "num_train_epochs": 4,
-  "total_flos": 1.502407617014661e+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 sha256:2a1b1bde55632273fa2d79fa4e7449972067ad76fcbd823e01f19e5d013eba50
+oid sha256:124ce131b35effd9a1e5de6412af80ddd87768156ca8d94b92ddde0c2440032d
 size 3439