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: robbert-v2-dutch-base-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: 0.8262
-        name: Test accuracy
-      - type: f1
-        value: 0.8267
-        name: Test f1
-      - type: precision
-        value: 0.8274
-        name: Test precision
-      - type: recall
-        value: 0.8262
-        name: Test recall
+    - name: Test accuracy
+      type: accuracy
+      value: 0.8070512820512821
+    - name: Test f1
+      type: f1
+      value: 0.8144966061997005
+    - name: Test precision
+      type: precision
+      value: 0.8275999429062602
+    - name: Test qwk
+      type: qwk
+      value: 0.7336245557372719
+    - name: Test recall
+      type: recall
+      value: 0.8070512820512821
+    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
 ---
+
+# robbert-v2-dutch-base-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: 0.7412639349881154
+- best_model_checkpoint: trained/hebban-reviews/robbert-v2-dutch-base/checkpoint-3500
+
+# Test results of best checkpoint
+- accuracy: 0.8070512820512821
+- f1: 0.8144966061997005
+- precision: 0.8275999429062602
+- qwk: 0.7336245557372719
+- recall: 0.8070512820512821
+
+## 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": 0.8070512820512821,
+    "epoch": 4.38,
+    "f1": 0.8144966061997005,
+    "precision": 0.8275999429062602,
+    "qwk": 0.7336245557372719,
+    "recall": 0.8070512820512821,
+    "train_loss": 0.4127759954567505,
+    "train_runtime": 2598.1165,
+    "train_samples_per_second": 246.382,
+    "train_steps_per_second": 1.925
+}

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/robbert-v2-dutch-base/runs/Jul28_17-46-38_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": "pdelobelle/robbert-v2-dutch-base",
+    "model_revision": "main",
+    "mp_parameters": "",
+    "n_trials": 8,
+    "no_cuda": false,
+    "num_train_epochs": 3.0,
+    "optim": "adamw_hf",
+    "output_dir": "trained/hebban-reviews/robbert-v2-dutch-base",
+    "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/robbert-v2-dutch-base",
+    "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

@@ -32,7 +32,7 @@
   "pad_token_id": 1,
   "position_embedding_type": "absolute",
   "torch_dtype": "float32",
-  "transformers_version": "4.21.0.dev0",
+  "transformers_version": "4.21.0",
   "type_vocab_size": 1,
   "use_cache": true,
   "vocab_size": 40000

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:f46bcf05bc1695e23df1ded5d9c92981bf3ac3c0841aea3a0bc2137f9dbb4b21
+oid sha256:96807aa4e98e6651797f42bd066ef8dae2ab02a72f92de17869d26d0fd8c0f6e
 size 467129325

test_predictions.txt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0dcf9c1e19881d0b081d2625d8c1b915685e29f291684c3ccefde0a959415e7b
+size 65401041

test_results.json

@@ -0,0 +1,7 @@
+{
+    "accuracy": 0.8070512820512821,
+    "f1": 0.8144966061997005,
+    "precision": 0.8275999429062602,
+    "qwk": 0.7336245557372719,
+    "recall": 0.8070512820512821
+}

train_results.json

@@ -0,0 +1,7 @@
+{
+    "epoch": 4.38,
+    "train_loss": 0.4127759954567505,
+    "train_runtime": 2598.1165,
+    "train_samples_per_second": 246.382,
+    "train_steps_per_second": 1.925
+}

trainer_state.json

@@ -1,457 +1,215 @@
 {
-  "best_metric": 0.826051364529945,
-  "best_model_checkpoint": "/home/bram/shares/predict/trained/dutch/hebban-reviews/robbert-v2-dutch-base/checkpoint-2500",
-  "epoch": 3.9447731755424065,
-  "global_step": 12000,
+  "best_metric": 0.7412639349881154,
+  "best_model_checkpoint": "trained/hebban-reviews/robbert-v2-dutch-base/checkpoint-3500",
+  "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": 4.7925000000000006e-05,
-      "loss": 0.7262,
+      "epoch": 0.44,
+      "learning_rate": 4.501099780043991e-05,
+      "loss": 0.6723,
       "step": 500
     },
     {
-      "epoch": 0.16,
-      "eval_accuracy": 0.7467948717948718,
-      "eval_f1": 0.7588505397999891,
-      "eval_loss": 0.6503846049308777,
-      "eval_precision": 0.7826608700318252,
-      "eval_recall": 0.7467948717948718,
-      "eval_runtime": 24.2662,
-      "eval_samples_per_second": 668.584,
-      "eval_steps_per_second": 27.858,
+      "epoch": 0.44,
+      "eval_accuracy": 0.8079388560157791,
+      "eval_f1": 0.8117501994363496,
+      "eval_loss": 0.5760409235954285,
+      "eval_precision": 0.8174343819162313,
+      "eval_qwk": 0.7302784941209646,
+      "eval_recall": 0.8079388560157791,
+      "eval_runtime": 23.4439,
+      "eval_samples_per_second": 692.036,
+      "eval_steps_per_second": 5.417,
       "step": 500
     },
     {
-      "epoch": 0.33,
-      "learning_rate": 4.5841666666666665e-05,
-      "loss": 0.6388,
+      "epoch": 0.88,
+      "learning_rate": 4.0011997600479906e-05,
+      "loss": 0.5754,
       "step": 1000
     },
     {
-      "epoch": 0.33,
-      "eval_accuracy": 0.7707716962524654,
-      "eval_f1": 0.7828012503260621,
-      "eval_loss": 0.6046772599220276,
-      "eval_precision": 0.806817293633085,
-      "eval_recall": 0.7707716962524654,
-      "eval_runtime": 24.3206,
-      "eval_samples_per_second": 667.089,
-      "eval_steps_per_second": 27.795,
+      "epoch": 0.88,
+      "eval_accuracy": 0.7649778106508875,
+      "eval_f1": 0.7798244436633573,
+      "eval_loss": 0.5568957328796387,
+      "eval_precision": 0.8129894746393622,
+      "eval_qwk": 0.6942695698574534,
+      "eval_recall": 0.7649778106508875,
+      "eval_runtime": 23.1665,
+      "eval_samples_per_second": 700.32,
+      "eval_steps_per_second": 5.482,
       "step": 1000
     },
     {
-      "epoch": 0.49,
-      "learning_rate": 4.375833333333334e-05,
-      "loss": 0.6036,
+      "epoch": 1.31,
+      "learning_rate": 3.50129974005199e-05,
+      "loss": 0.5052,
       "step": 1500
     },
     {
-      "epoch": 0.49,
-      "eval_accuracy": 0.7991247534516766,
-      "eval_f1": 0.8068009793163453,
-      "eval_loss": 0.5817954540252686,
-      "eval_precision": 0.8200694446499367,
-      "eval_recall": 0.7991247534516766,
-      "eval_runtime": 24.3415,
-      "eval_samples_per_second": 666.516,
-      "eval_steps_per_second": 27.771,
+      "epoch": 1.31,
+      "eval_accuracy": 0.7829758382642998,
+      "eval_f1": 0.7967934169486903,
+      "eval_loss": 0.5705748200416565,
+      "eval_precision": 0.8279826615797163,
+      "eval_qwk": 0.7174911347556729,
+      "eval_recall": 0.7829758382642998,
+      "eval_runtime": 23.1285,
+      "eval_samples_per_second": 701.473,
+      "eval_steps_per_second": 5.491,
       "step": 1500
     },
     {
-      "epoch": 0.66,
-      "learning_rate": 4.1675e-05,
-      "loss": 0.5894,
+      "epoch": 1.75,
+      "learning_rate": 3.001399720055989e-05,
+      "loss": 0.4723,
       "step": 2000
     },
     {
-      "epoch": 0.66,
-      "eval_accuracy": 0.8214373767258383,
-      "eval_f1": 0.8223555383757819,
-      "eval_loss": 0.5838125944137573,
-      "eval_precision": 0.8240754296334577,
-      "eval_recall": 0.8214373767258383,
-      "eval_runtime": 24.3264,
-      "eval_samples_per_second": 666.929,
-      "eval_steps_per_second": 27.789,
+      "epoch": 1.75,
+      "eval_accuracy": 0.7924063116370809,
+      "eval_f1": 0.8033559165914831,
+      "eval_loss": 0.5601416230201721,
+      "eval_precision": 0.8252798007612112,
+      "eval_qwk": 0.7270656679689509,
+      "eval_recall": 0.7924063116370809,
+      "eval_runtime": 23.2157,
+      "eval_samples_per_second": 698.837,
+      "eval_steps_per_second": 5.47,
       "step": 2000
     },
     {
-      "epoch": 0.82,
-      "learning_rate": 3.960000000000001e-05,
-      "loss": 0.5678,
+      "epoch": 2.19,
+      "learning_rate": 2.5024995000999802e-05,
+      "loss": 0.4285,
       "step": 2500
     },
     {
-      "epoch": 0.82,
-      "eval_accuracy": 0.8257519723865878,
-      "eval_f1": 0.826051364529945,
-      "eval_loss": 0.565325915813446,
-      "eval_precision": 0.8264063729685338,
-      "eval_recall": 0.8257519723865878,
-      "eval_runtime": 24.3378,
-      "eval_samples_per_second": 666.617,
-      "eval_steps_per_second": 27.776,
+      "epoch": 2.19,
+      "eval_accuracy": 0.8101577909270217,
+      "eval_f1": 0.8186486636369545,
+      "eval_loss": 0.691852867603302,
+      "eval_precision": 0.8361687645268726,
+      "eval_qwk": 0.7349979059232339,
+      "eval_recall": 0.8101577909270217,
+      "eval_runtime": 23.1035,
+      "eval_samples_per_second": 702.232,
+      "eval_steps_per_second": 5.497,
       "step": 2500
     },
     {
-      "epoch": 0.99,
-      "learning_rate": 3.7516666666666666e-05,
-      "loss": 0.5821,
+      "epoch": 2.63,
+      "learning_rate": 2.0025994801039795e-05,
+      "loss": 0.3553,
       "step": 3000
     },
     {
-      "epoch": 0.99,
-      "eval_accuracy": 0.8164447731755424,
-      "eval_f1": 0.8171577310703354,
-      "eval_loss": 0.5594866275787354,
-      "eval_precision": 0.8191229796098939,
-      "eval_recall": 0.8164447731755424,
-      "eval_runtime": 24.3358,
-      "eval_samples_per_second": 666.673,
-      "eval_steps_per_second": 27.778,
+      "epoch": 2.63,
+      "eval_accuracy": 0.7843318540433925,
+      "eval_f1": 0.7982445447583074,
+      "eval_loss": 0.674355685710907,
+      "eval_precision": 0.8311821725851122,
+      "eval_qwk": 0.7133869717906458,
+      "eval_recall": 0.7843318540433925,
+      "eval_runtime": 23.1222,
+      "eval_samples_per_second": 701.663,
+      "eval_steps_per_second": 5.493,
       "step": 3000
     },
     {
-      "epoch": 1.15,
-      "learning_rate": 3.543333333333333e-05,
-      "loss": 0.4741,
+      "epoch": 3.07,
+      "learning_rate": 1.5026994601079786e-05,
+      "loss": 0.3433,
       "step": 3500
     },
     {
-      "epoch": 1.15,
-      "eval_accuracy": 0.7975838264299803,
-      "eval_f1": 0.8063837559416328,
-      "eval_loss": 0.5798861384391785,
-      "eval_precision": 0.8218943463788534,
-      "eval_recall": 0.7975838264299803,
-      "eval_runtime": 24.3305,
-      "eval_samples_per_second": 666.816,
-      "eval_steps_per_second": 27.784,
+      "epoch": 3.07,
+      "eval_accuracy": 0.813732741617357,
+      "eval_f1": 0.821008951036937,
+      "eval_loss": 0.8146640658378601,
+      "eval_precision": 0.8342737114916078,
+      "eval_qwk": 0.7412639349881154,
+      "eval_recall": 0.813732741617357,
+      "eval_runtime": 23.1151,
+      "eval_samples_per_second": 701.878,
+      "eval_steps_per_second": 5.494,
       "step": 3500
     },
     {
-      "epoch": 1.31,
-      "learning_rate": 3.3350000000000004e-05,
-      "loss": 0.4763,
+      "epoch": 3.51,
+      "learning_rate": 1.0037992401519696e-05,
+      "loss": 0.2751,
       "step": 4000
     },
     {
-      "epoch": 1.31,
-      "eval_accuracy": 0.8000493096646942,
-      "eval_f1": 0.8080426775793622,
-      "eval_loss": 0.5934886336326599,
-      "eval_precision": 0.8221336258804294,
-      "eval_recall": 0.8000493096646942,
-      "eval_runtime": 24.3209,
-      "eval_samples_per_second": 667.08,
-      "eval_steps_per_second": 27.795,
+      "epoch": 3.51,
+      "eval_accuracy": 0.8128698224852071,
+      "eval_f1": 0.8199992286734568,
+      "eval_loss": 0.8419223427772522,
+      "eval_precision": 0.8332550382998175,
+      "eval_qwk": 0.7383621154665407,
+      "eval_recall": 0.8128698224852071,
+      "eval_runtime": 23.1425,
+      "eval_samples_per_second": 701.047,
+      "eval_steps_per_second": 5.488,
       "step": 4000
     },
     {
-      "epoch": 1.48,
-      "learning_rate": 3.126666666666666e-05,
-      "loss": 0.4719,
+      "epoch": 3.94,
+      "learning_rate": 5.038992201559688e-06,
+      "loss": 0.2659,
       "step": 4500
     },
     {
-      "epoch": 1.48,
-      "eval_accuracy": 0.7595537475345168,
-      "eval_f1": 0.7769614913537937,
-      "eval_loss": 0.5927842855453491,
-      "eval_precision": 0.8215588992651691,
-      "eval_recall": 0.7595537475345168,
-      "eval_runtime": 24.3097,
-      "eval_samples_per_second": 667.388,
-      "eval_steps_per_second": 27.808,
+      "epoch": 3.94,
+      "eval_accuracy": 0.8022682445759369,
+      "eval_f1": 0.8125481330626305,
+      "eval_loss": 0.8309345841407776,
+      "eval_precision": 0.8344366995643693,
+      "eval_qwk": 0.7292405615679007,
+      "eval_recall": 0.8022682445759369,
+      "eval_runtime": 23.1428,
+      "eval_samples_per_second": 701.039,
+      "eval_steps_per_second": 5.488,
       "step": 4500
     },
     {
-      "epoch": 1.64,
-      "learning_rate": 2.91875e-05,
-      "loss": 0.477,
+      "epoch": 4.38,
+      "learning_rate": 3.9992001599680065e-08,
+      "loss": 0.2346,
       "step": 5000
     },
     {
-      "epoch": 1.64,
-      "eval_accuracy": 0.8143491124260355,
-      "eval_f1": 0.8178014358117583,
-      "eval_loss": 0.571167528629303,
-      "eval_precision": 0.8224283181214769,
-      "eval_recall": 0.8143491124260355,
-      "eval_runtime": 24.3019,
-      "eval_samples_per_second": 667.601,
-      "eval_steps_per_second": 27.817,
+      "epoch": 4.38,
+      "eval_accuracy": 0.8102810650887574,
+      "eval_f1": 0.8179537922797215,
+      "eval_loss": 0.8737895488739014,
+      "eval_precision": 0.8320255779887702,
+      "eval_qwk": 0.7367921505621005,
+      "eval_recall": 0.8102810650887574,
+      "eval_runtime": 23.1382,
+      "eval_samples_per_second": 701.178,
+      "eval_steps_per_second": 5.489,
       "step": 5000
     },
     {
-      "epoch": 1.81,
-      "learning_rate": 2.710416666666667e-05,
-      "loss": 0.4627,
-      "step": 5500
-    },
-    {
-      "epoch": 1.81,
-      "eval_accuracy": 0.8163831360946746,
-      "eval_f1": 0.8209409596272614,
-      "eval_loss": 0.5852541327476501,
-      "eval_precision": 0.8276030279019199,
-      "eval_recall": 0.8163831360946746,
-      "eval_runtime": 24.3288,
-      "eval_samples_per_second": 666.863,
-      "eval_steps_per_second": 27.786,
-      "step": 5500
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 2.5020833333333333e-05,
-      "loss": 0.4669,
-      "step": 6000
-    },
-    {
-      "epoch": 1.97,
-      "eval_accuracy": 0.823594674556213,
-      "eval_f1": 0.8244437619387351,
-      "eval_loss": 0.5960806608200073,
-      "eval_precision": 0.8258010274342865,
-      "eval_recall": 0.823594674556213,
-      "eval_runtime": 24.228,
-      "eval_samples_per_second": 669.639,
-      "eval_steps_per_second": 27.902,
-      "step": 6000
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 2.2937500000000002e-05,
-      "loss": 0.3604,
-      "step": 6500
-    },
-    {
-      "epoch": 2.14,
-      "eval_accuracy": 0.8014669625246549,
-      "eval_f1": 0.8102801213854947,
-      "eval_loss": 0.6862322092056274,
-      "eval_precision": 0.8275335401956307,
-      "eval_recall": 0.8014669625246549,
-      "eval_runtime": 24.2792,
-      "eval_samples_per_second": 668.227,
-      "eval_steps_per_second": 27.843,
-      "step": 6500
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 2.0854166666666668e-05,
-      "loss": 0.3489,
-      "step": 7000
-    },
-    {
-      "epoch": 2.3,
-      "eval_accuracy": 0.7792776134122288,
-      "eval_f1": 0.7927818598604747,
-      "eval_loss": 0.6760165095329285,
-      "eval_precision": 0.822269294544987,
-      "eval_recall": 0.7792776134122288,
-      "eval_runtime": 24.2892,
-      "eval_samples_per_second": 667.95,
-      "eval_steps_per_second": 27.831,
-      "step": 7000
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 1.8775000000000002e-05,
-      "loss": 0.3394,
-      "step": 7500
-    },
-    {
-      "epoch": 2.47,
-      "eval_accuracy": 0.8185404339250493,
-      "eval_f1": 0.8224920113739561,
-      "eval_loss": 0.6916875839233398,
-      "eval_precision": 0.8283683902286886,
-      "eval_recall": 0.8185404339250493,
-      "eval_runtime": 24.2568,
-      "eval_samples_per_second": 668.844,
-      "eval_steps_per_second": 27.868,
-      "step": 7500
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 1.6695833333333334e-05,
-      "loss": 0.3462,
-      "step": 8000
-    },
-    {
-      "epoch": 2.63,
-      "eval_accuracy": 0.8147805719921104,
-      "eval_f1": 0.8200537266341018,
-      "eval_loss": 0.6634742617607117,
-      "eval_precision": 0.8279502059904518,
-      "eval_recall": 0.8147805719921104,
-      "eval_runtime": 24.237,
-      "eval_samples_per_second": 669.39,
-      "eval_steps_per_second": 27.891,
-      "step": 8000
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.4612500000000001e-05,
-      "loss": 0.3498,
-      "step": 8500
-    },
-    {
-      "epoch": 2.79,
-      "eval_accuracy": 0.8173076923076923,
-      "eval_f1": 0.8219188790015944,
-      "eval_loss": 0.6741124987602234,
-      "eval_precision": 0.8291747460835226,
-      "eval_recall": 0.8173076923076923,
-      "eval_runtime": 24.3305,
-      "eval_samples_per_second": 666.817,
-      "eval_steps_per_second": 27.784,
-      "step": 8500
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 1.2529166666666667e-05,
-      "loss": 0.335,
-      "step": 9000
-    },
-    {
-      "epoch": 2.96,
-      "eval_accuracy": 0.8091715976331361,
-      "eval_f1": 0.8160105785088146,
-      "eval_loss": 0.6432847380638123,
-      "eval_precision": 0.8271351948602929,
-      "eval_recall": 0.8091715976331361,
-      "eval_runtime": 24.2377,
-      "eval_samples_per_second": 669.372,
-      "eval_steps_per_second": 27.89,
-      "step": 9000
-    },
-    {
-      "epoch": 3.12,
-      "learning_rate": 1.0445833333333334e-05,
-      "loss": 0.2724,
-      "step": 9500
-    },
-    {
-      "epoch": 3.12,
-      "eval_accuracy": 0.8155818540433925,
-      "eval_f1": 0.8200452160739635,
-      "eval_loss": 0.778809666633606,
-      "eval_precision": 0.8263415739712193,
-      "eval_recall": 0.8155818540433925,
-      "eval_runtime": 24.3292,
-      "eval_samples_per_second": 666.853,
-      "eval_steps_per_second": 27.786,
-      "step": 9500
-    },
-    {
-      "epoch": 3.29,
-      "learning_rate": 8.362500000000002e-06,
-      "loss": 0.2404,
-      "step": 10000
-    },
-    {
-      "epoch": 3.29,
-      "eval_accuracy": 0.8102194280078896,
-      "eval_f1": 0.8165639264989629,
-      "eval_loss": 0.7925969958305359,
-      "eval_precision": 0.8272230495650681,
-      "eval_recall": 0.8102194280078896,
-      "eval_runtime": 24.3267,
-      "eval_samples_per_second": 666.921,
-      "eval_steps_per_second": 27.788,
-      "step": 10000
-    },
-    {
-      "epoch": 3.45,
-      "learning_rate": 6.2791666666666665e-06,
-      "loss": 0.2417,
-      "step": 10500
-    },
-    {
-      "epoch": 3.45,
-      "eval_accuracy": 0.811698717948718,
-      "eval_f1": 0.8176421251758528,
-      "eval_loss": 0.8324716687202454,
-      "eval_precision": 0.8284127391855651,
-      "eval_recall": 0.811698717948718,
-      "eval_runtime": 24.2309,
-      "eval_samples_per_second": 669.559,
-      "eval_steps_per_second": 27.898,
-      "step": 10500
-    },
-    {
-      "epoch": 3.62,
-      "learning_rate": 4.2000000000000004e-06,
-      "loss": 0.2426,
-      "step": 11000
-    },
-    {
-      "epoch": 3.62,
-      "eval_accuracy": 0.8218071992110454,
-      "eval_f1": 0.8237488575881482,
-      "eval_loss": 0.835399866104126,
-      "eval_precision": 0.8261715829036445,
-      "eval_recall": 0.8218071992110454,
-      "eval_runtime": 24.2418,
-      "eval_samples_per_second": 669.257,
-      "eval_steps_per_second": 27.886,
-      "step": 11000
-    },
-    {
-      "epoch": 3.78,
-      "learning_rate": 2.1166666666666666e-06,
-      "loss": 0.2399,
-      "step": 11500
-    },
-    {
-      "epoch": 3.78,
-      "eval_accuracy": 0.8130547337278107,
-      "eval_f1": 0.8181997138713092,
-      "eval_loss": 0.8061773777008057,
-      "eval_precision": 0.8260171371000359,
-      "eval_recall": 0.8130547337278107,
-      "eval_runtime": 24.2269,
-      "eval_samples_per_second": 669.67,
-      "eval_steps_per_second": 27.903,
-      "step": 11500
-    },
-    {
-      "epoch": 3.94,
-      "learning_rate": 3.3333333333333334e-08,
-      "loss": 0.2292,
-      "step": 12000
-    },
-    {
-      "epoch": 3.94,
-      "eval_accuracy": 0.8134245562130178,
-      "eval_f1": 0.8183934985407206,
-      "eval_loss": 0.8132809996604919,
-      "eval_precision": 0.8258917495347268,
-      "eval_recall": 0.8134245562130178,
-      "eval_runtime": 24.2378,
-      "eval_samples_per_second": 669.369,
-      "eval_steps_per_second": 27.89,
-      "step": 12000
-    },
-    {
-      "epoch": 3.94,
-      "step": 12000,
-      "total_flos": 1.481337892587438e+17,
-      "train_loss": 0.42010608355204265,
-      "train_runtime": 3460.7395,
-      "train_samples_per_second": 166.438,
-      "train_steps_per_second": 3.467
+      "epoch": 4.38,
+      "step": 5001,
+      "total_flos": 1.6838837781764506e+17,
+      "train_loss": 0.4127759954567505,
+      "train_runtime": 2598.1165,
+      "train_samples_per_second": 246.382,
+      "train_steps_per_second": 1.925
     }
   ],
-  "max_steps": 12000,
-  "num_train_epochs": 4,
-  "total_flos": 1.481337892587438e+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:a595e075a32f1895e812ed866d982ad1bfebea5e607b63f47a60a916b06e1199
+oid sha256:5f88658b17ce89f74b414f36c684b79e029180c576f6545e98a028462e121694
 size 3375