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: xlm-roberta-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: 0.8137
-        name: Test accuracy
-      - type: f1
-        value: 0.8176
-        name: Test f1
-      - type: precision
-        value: 0.8232
-        name: Test precision
-      - type: recall
-        value: 0.8137
-        name: Test recall
----
+    - name: Test accuracy
+      type: accuracy
+      value: 0.8094674556213017
+    - name: Test f1
+      type: f1
+      value: 0.812677483587223
+    - name: Test precision
+      type: precision
+      value: 0.8173602585519025
+    - name: Test qwk
+      type: qwk
+      value: 0.7369243423166991
+    - name: Test recall
+      type: recall
+      value: 0.8094674556213017
+    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
+---
+
+# xlm-roberta-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: 0.741533273748008
+- best_model_checkpoint: trained/hebban-reviews/xlm-roberta-base/checkpoint-2000
+
+# Test results of best checkpoint
+- accuracy: 0.8094674556213017
+- f1: 0.812677483587223
+- precision: 0.8173602585519025
+- qwk: 0.7369243423166991
+- recall: 0.8094674556213017
+
+## 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": 0.8094674556213017,
+    "epoch": 4.38,
+    "f1": 0.812677483587223,
+    "precision": 0.8173602585519025,
+    "qwk": 0.7369243423166991,
+    "recall": 0.8094674556213017,
+    "train_loss": 0.4945967703420051,
+    "train_runtime": 2688.9042,
+    "train_samples_per_second": 238.063,
+    "train_steps_per_second": 1.86
+}

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/xlm-roberta-base/runs/Jul28_18-36-45_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": "xlm-roberta-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/xlm-roberta-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/xlm-roberta-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

@@ -31,7 +31,7 @@
   "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": 250002

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 sha256:71bc1033e544364f7f5b1a44e5c7a8c073c3f1b62d52a5272e43fb8b73b15d72
+oid sha256:2b34ca84ce38964385525b3df415ed8b650a3507b975458ba2140f3232e27f66
 size 1112255469

test_predictions.txt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57409f6279da21f877909a078eaae93752ee7b7a85febe2328ff47d19a452b4b
+size 65402901

test_results.json

@@ -0,0 +1,7 @@
+{
+    "accuracy": 0.8094674556213017,
+    "f1": 0.812677483587223,
+    "precision": 0.8173602585519025,
+    "qwk": 0.7369243423166991,
+    "recall": 0.8094674556213017
+}

train_results.json

@@ -0,0 +1,7 @@
+{
+    "epoch": 4.38,
+    "train_loss": 0.4945967703420051,
+    "train_runtime": 2688.9042,
+    "train_samples_per_second": 238.063,
+    "train_steps_per_second": 1.86
+}

trainer_state.json

@@ -1,457 +1,215 @@
 {
-  "best_metric": 0.8169879527109274,
-  "best_model_checkpoint": "/home/bram/shares/predict/trained/dutch/hebban-reviews/xlm-roberta-base/checkpoint-11000",
-  "epoch": 3.9447731755424065,
-  "global_step": 12000,
+  "best_metric": 0.741533273748008,
+  "best_model_checkpoint": "trained/hebban-reviews/xlm-roberta-base/checkpoint-2000",
+  "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": 4.794166666666667e-05,
-      "loss": 0.913,
+      "epoch": 0.44,
+      "learning_rate": 4.504099180163968e-05,
+      "loss": 0.7622,
       "step": 500
     },
     {
-      "epoch": 0.16,
-      "eval_accuracy": 0.7665803747534516,
-      "eval_f1": 0.7649545454029377,
-      "eval_loss": 0.7253644466400146,
-      "eval_precision": 0.7641565688780922,
-      "eval_recall": 0.7665803747534516,
-      "eval_runtime": 24.3431,
-      "eval_samples_per_second": 666.472,
-      "eval_steps_per_second": 27.77,
+      "epoch": 0.44,
+      "eval_accuracy": 0.7567800788954635,
+      "eval_f1": 0.7729405363796663,
+      "eval_loss": 0.626587986946106,
+      "eval_precision": 0.8122754990827801,
+      "eval_qwk": 0.6618145848522559,
+      "eval_recall": 0.7567800788954635,
+      "eval_runtime": 23.2604,
+      "eval_samples_per_second": 697.494,
+      "eval_steps_per_second": 5.46,
       "step": 500
     },
     {
-      "epoch": 0.33,
-      "learning_rate": 4.5858333333333334e-05,
-      "loss": 0.7582,
+      "epoch": 0.88,
+      "learning_rate": 4.0041991601679665e-05,
+      "loss": 0.6228,
       "step": 1000
     },
     {
-      "epoch": 0.33,
-      "eval_accuracy": 0.6934787968441815,
-      "eval_f1": 0.7165538695299023,
-      "eval_loss": 0.7026467323303223,
-      "eval_precision": 0.7706797791371907,
-      "eval_recall": 0.6934787968441815,
-      "eval_runtime": 24.1169,
-      "eval_samples_per_second": 672.723,
-      "eval_steps_per_second": 28.03,
+      "epoch": 0.88,
+      "eval_accuracy": 0.7375493096646942,
+      "eval_f1": 0.7560139522475998,
+      "eval_loss": 0.592847466468811,
+      "eval_precision": 0.8006409272816242,
+      "eval_qwk": 0.6464780542008888,
+      "eval_recall": 0.7375493096646942,
+      "eval_runtime": 23.1917,
+      "eval_samples_per_second": 699.56,
+      "eval_steps_per_second": 5.476,
       "step": 1000
     },
     {
-      "epoch": 0.49,
-      "learning_rate": 4.3775e-05,
-      "loss": 0.6847,
+      "epoch": 1.31,
+      "learning_rate": 3.504299140171966e-05,
+      "loss": 0.5623,
       "step": 1500
     },
     {
-      "epoch": 0.49,
-      "eval_accuracy": 0.7233727810650887,
-      "eval_f1": 0.7442026185547411,
-      "eval_loss": 0.6611877083778381,
-      "eval_precision": 0.8038626962859567,
-      "eval_recall": 0.7233727810650887,
-      "eval_runtime": 24.1771,
-      "eval_samples_per_second": 671.047,
-      "eval_steps_per_second": 27.96,
+      "epoch": 1.31,
+      "eval_accuracy": 0.7543145956607495,
+      "eval_f1": 0.7717090298131432,
+      "eval_loss": 0.5968530178070068,
+      "eval_precision": 0.8134758130284254,
+      "eval_qwk": 0.6828220433375901,
+      "eval_recall": 0.7543145956607495,
+      "eval_runtime": 23.2366,
+      "eval_samples_per_second": 698.21,
+      "eval_steps_per_second": 5.466,
       "step": 1500
     },
     {
-      "epoch": 0.66,
-      "learning_rate": 4.1691666666666666e-05,
-      "loss": 0.6532,
+      "epoch": 1.75,
+      "learning_rate": 3.0043991201759648e-05,
+      "loss": 0.5421,
       "step": 2000
     },
     {
-      "epoch": 0.66,
-      "eval_accuracy": 0.8156434911242604,
-      "eval_f1": 0.8093687626348003,
-      "eval_loss": 0.6557860374450684,
-      "eval_precision": 0.8059505966835585,
-      "eval_recall": 0.8156434911242604,
-      "eval_runtime": 24.3135,
-      "eval_samples_per_second": 667.283,
-      "eval_steps_per_second": 27.803,
+      "epoch": 1.75,
+      "eval_accuracy": 0.8150271203155819,
+      "eval_f1": 0.8184520029418662,
+      "eval_loss": 0.5808575749397278,
+      "eval_precision": 0.8233166962202109,
+      "eval_qwk": 0.741533273748008,
+      "eval_recall": 0.8150271203155819,
+      "eval_runtime": 23.1787,
+      "eval_samples_per_second": 699.952,
+      "eval_steps_per_second": 5.479,
       "step": 2000
     },
     {
-      "epoch": 0.82,
-      "learning_rate": 3.960833333333334e-05,
-      "loss": 0.6281,
+      "epoch": 2.19,
+      "learning_rate": 2.504499100179964e-05,
+      "loss": 0.5033,
       "step": 2500
     },
     {
-      "epoch": 0.82,
-      "eval_accuracy": 0.7695389546351085,
-      "eval_f1": 0.7795959943899675,
-      "eval_loss": 0.616236686706543,
-      "eval_precision": 0.7973850176423627,
-      "eval_recall": 0.7695389546351085,
-      "eval_runtime": 24.3023,
-      "eval_samples_per_second": 667.59,
-      "eval_steps_per_second": 27.816,
+      "epoch": 2.19,
+      "eval_accuracy": 0.789447731755424,
+      "eval_f1": 0.8012347601760753,
+      "eval_loss": 0.640012800693512,
+      "eval_precision": 0.8269193049518779,
+      "eval_qwk": 0.7122705559860744,
+      "eval_recall": 0.789447731755424,
+      "eval_runtime": 23.1944,
+      "eval_samples_per_second": 699.48,
+      "eval_steps_per_second": 5.475,
       "step": 2500
     },
     {
-      "epoch": 0.99,
-      "learning_rate": 3.7525e-05,
-      "loss": 0.6281,
+      "epoch": 2.63,
+      "learning_rate": 2.0055988802239554e-05,
+      "loss": 0.4493,
       "step": 3000
     },
     {
-      "epoch": 0.99,
-      "eval_accuracy": 0.8015902366863905,
-      "eval_f1": 0.803170220169187,
-      "eval_loss": 0.5991469621658325,
-      "eval_precision": 0.806066258666526,
-      "eval_recall": 0.8015902366863905,
-      "eval_runtime": 24.3429,
-      "eval_samples_per_second": 666.477,
-      "eval_steps_per_second": 27.77,
+      "epoch": 2.63,
+      "eval_accuracy": 0.7778599605522682,
+      "eval_f1": 0.7918891407245614,
+      "eval_loss": 0.6219201683998108,
+      "eval_precision": 0.8239935574770723,
+      "eval_qwk": 0.7029409846056511,
+      "eval_recall": 0.7778599605522682,
+      "eval_runtime": 23.0915,
+      "eval_samples_per_second": 702.598,
+      "eval_steps_per_second": 5.5,
       "step": 3000
     },
     {
-      "epoch": 1.15,
-      "learning_rate": 3.544583333333333e-05,
-      "loss": 0.5668,
+      "epoch": 3.07,
+      "learning_rate": 1.5066986602679465e-05,
+      "loss": 0.4391,
       "step": 3500
     },
     {
-      "epoch": 1.15,
-      "eval_accuracy": 0.7874753451676528,
-      "eval_f1": 0.7971578305557919,
-      "eval_loss": 0.5982191562652588,
-      "eval_precision": 0.8156521173286116,
-      "eval_recall": 0.7874753451676528,
-      "eval_runtime": 24.3367,
-      "eval_samples_per_second": 666.646,
-      "eval_steps_per_second": 27.777,
+      "epoch": 3.07,
+      "eval_accuracy": 0.784948224852071,
+      "eval_f1": 0.7976824915929677,
+      "eval_loss": 0.6451554298400879,
+      "eval_precision": 0.8255295840982949,
+      "eval_qwk": 0.7107576916204621,
+      "eval_recall": 0.784948224852071,
+      "eval_runtime": 23.2205,
+      "eval_samples_per_second": 698.694,
+      "eval_steps_per_second": 5.469,
       "step": 3500
     },
     {
-      "epoch": 1.31,
-      "learning_rate": 3.3362500000000005e-05,
-      "loss": 0.567,
+      "epoch": 3.51,
+      "learning_rate": 1.0067986402719456e-05,
+      "loss": 0.3756,
       "step": 4000
     },
     {
-      "epoch": 1.31,
-      "eval_accuracy": 0.8007889546351085,
-      "eval_f1": 0.8041408078539543,
-      "eval_loss": 0.6023094654083252,
-      "eval_precision": 0.8088979038333125,
-      "eval_recall": 0.8007889546351085,
-      "eval_runtime": 24.2641,
-      "eval_samples_per_second": 668.643,
-      "eval_steps_per_second": 27.86,
+      "epoch": 3.51,
+      "eval_accuracy": 0.7970290927021696,
+      "eval_f1": 0.8070579315874782,
+      "eval_loss": 0.6958814263343811,
+      "eval_precision": 0.8265215874136934,
+      "eval_qwk": 0.7249078215037359,
+      "eval_recall": 0.7970290927021696,
+      "eval_runtime": 23.1336,
+      "eval_samples_per_second": 701.319,
+      "eval_steps_per_second": 5.49,
       "step": 4000
     },
     {
-      "epoch": 1.48,
-      "learning_rate": 3.127916666666667e-05,
-      "loss": 0.5704,
+      "epoch": 3.94,
+      "learning_rate": 5.068986202759449e-06,
+      "loss": 0.3633,
       "step": 4500
     },
     {
-      "epoch": 1.48,
-      "eval_accuracy": 0.7429117357001972,
-      "eval_f1": 0.7619810076683907,
-      "eval_loss": 0.6065093278884888,
-      "eval_precision": 0.8107750610152921,
-      "eval_recall": 0.7429117357001972,
-      "eval_runtime": 24.3353,
-      "eval_samples_per_second": 666.686,
-      "eval_steps_per_second": 27.779,
+      "epoch": 3.94,
+      "eval_accuracy": 0.7962894477317555,
+      "eval_f1": 0.8063568557789076,
+      "eval_loss": 0.6968616247177124,
+      "eval_precision": 0.8256437219303169,
+      "eval_qwk": 0.7253147876480203,
+      "eval_recall": 0.7962894477317555,
+      "eval_runtime": 23.1633,
+      "eval_samples_per_second": 700.418,
+      "eval_steps_per_second": 5.483,
       "step": 4500
     },
     {
-      "epoch": 1.64,
-      "learning_rate": 2.9195833333333333e-05,
-      "loss": 0.5596,
+      "epoch": 4.38,
+      "learning_rate": 6.998600279944012e-08,
+      "loss": 0.3263,
       "step": 5000
     },
     {
-      "epoch": 1.64,
-      "eval_accuracy": 0.8072608481262328,
-      "eval_f1": 0.8104579115041036,
-      "eval_loss": 0.5900022983551025,
-      "eval_precision": 0.8144869336926288,
-      "eval_recall": 0.8072608481262328,
-      "eval_runtime": 24.3429,
-      "eval_samples_per_second": 666.477,
-      "eval_steps_per_second": 27.77,
+      "epoch": 4.38,
+      "eval_accuracy": 0.7972140039447732,
+      "eval_f1": 0.8072907591794813,
+      "eval_loss": 0.7634969353675842,
+      "eval_precision": 0.826968890246297,
+      "eval_qwk": 0.7254593503280649,
+      "eval_recall": 0.7972140039447732,
+      "eval_runtime": 23.3008,
+      "eval_samples_per_second": 696.286,
+      "eval_steps_per_second": 5.45,
       "step": 5000
     },
     {
-      "epoch": 1.81,
-      "learning_rate": 2.7116666666666667e-05,
-      "loss": 0.5495,
-      "step": 5500
-    },
-    {
-      "epoch": 1.81,
-      "eval_accuracy": 0.810034516765286,
-      "eval_f1": 0.8141305380075001,
-      "eval_loss": 0.613106906414032,
-      "eval_precision": 0.8219044900382881,
-      "eval_recall": 0.810034516765286,
-      "eval_runtime": 24.3418,
-      "eval_samples_per_second": 666.509,
-      "eval_steps_per_second": 27.771,
-      "step": 5500
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 2.5033333333333336e-05,
-      "loss": 0.5449,
-      "step": 6000
-    },
-    {
-      "epoch": 1.97,
-      "eval_accuracy": 0.8124383629191322,
-      "eval_f1": 0.8140798132169556,
-      "eval_loss": 0.6060279011726379,
-      "eval_precision": 0.816286158402022,
-      "eval_recall": 0.8124383629191322,
-      "eval_runtime": 24.332,
-      "eval_samples_per_second": 666.777,
-      "eval_steps_per_second": 27.782,
-      "step": 6000
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 2.2950000000000002e-05,
-      "loss": 0.4898,
-      "step": 6500
-    },
-    {
-      "epoch": 2.14,
-      "eval_accuracy": 0.7848865877712031,
-      "eval_f1": 0.7964804879952159,
-      "eval_loss": 0.6215521693229675,
-      "eval_precision": 0.820125366434727,
-      "eval_recall": 0.7848865877712031,
-      "eval_runtime": 24.3196,
-      "eval_samples_per_second": 667.117,
-      "eval_steps_per_second": 27.797,
-      "step": 6500
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 2.0866666666666668e-05,
-      "loss": 0.4837,
-      "step": 7000
-    },
-    {
-      "epoch": 2.3,
-      "eval_accuracy": 0.7318786982248521,
-      "eval_f1": 0.7528823670992008,
-      "eval_loss": 0.6411539912223816,
-      "eval_precision": 0.8100101192694165,
-      "eval_recall": 0.7318786982248521,
-      "eval_runtime": 24.3327,
-      "eval_samples_per_second": 666.758,
-      "eval_steps_per_second": 27.782,
-      "step": 7000
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 1.87875e-05,
-      "loss": 0.4671,
-      "step": 7500
-    },
-    {
-      "epoch": 2.47,
-      "eval_accuracy": 0.803870808678501,
-      "eval_f1": 0.810953536758238,
-      "eval_loss": 0.6316830515861511,
-      "eval_precision": 0.8241951591967538,
-      "eval_recall": 0.803870808678501,
-      "eval_runtime": 24.3331,
-      "eval_samples_per_second": 666.747,
-      "eval_steps_per_second": 27.781,
-      "step": 7500
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 1.670416666666667e-05,
-      "loss": 0.4791,
-      "step": 8000
-    },
-    {
-      "epoch": 2.63,
-      "eval_accuracy": 0.8032544378698225,
-      "eval_f1": 0.8091355876498971,
-      "eval_loss": 0.5908682942390442,
-      "eval_precision": 0.8179762946015251,
-      "eval_recall": 0.8032544378698225,
-      "eval_runtime": 24.3361,
-      "eval_samples_per_second": 666.664,
-      "eval_steps_per_second": 27.778,
-      "step": 8000
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.4620833333333334e-05,
-      "loss": 0.4739,
-      "step": 8500
-    },
-    {
-      "epoch": 2.79,
-      "eval_accuracy": 0.8067061143984221,
-      "eval_f1": 0.8120892015198095,
-      "eval_loss": 0.6165759563446045,
-      "eval_precision": 0.8199753861011193,
-      "eval_recall": 0.8067061143984221,
-      "eval_runtime": 24.1747,
-      "eval_samples_per_second": 671.116,
-      "eval_steps_per_second": 27.963,
-      "step": 8500
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 1.25375e-05,
-      "loss": 0.4587,
-      "step": 9000
-    },
-    {
-      "epoch": 2.96,
-      "eval_accuracy": 0.8041173570019724,
-      "eval_f1": 0.8104680056938384,
-      "eval_loss": 0.5887444019317627,
-      "eval_precision": 0.820271897699825,
-      "eval_recall": 0.8041173570019724,
-      "eval_runtime": 24.1945,
-      "eval_samples_per_second": 670.564,
-      "eval_steps_per_second": 27.94,
-      "step": 9000
-    },
-    {
-      "epoch": 3.12,
-      "learning_rate": 1.0454166666666667e-05,
-      "loss": 0.4147,
-      "step": 9500
-    },
-    {
-      "epoch": 3.12,
-      "eval_accuracy": 0.780448717948718,
-      "eval_f1": 0.7927207824065717,
-      "eval_loss": 0.6190515160560608,
-      "eval_precision": 0.8178157106781372,
-      "eval_recall": 0.780448717948718,
-      "eval_runtime": 24.1866,
-      "eval_samples_per_second": 670.784,
-      "eval_steps_per_second": 27.949,
-      "step": 9500
-    },
-    {
-      "epoch": 3.29,
-      "learning_rate": 8.370833333333333e-06,
-      "loss": 0.3861,
-      "step": 10000
-    },
-    {
-      "epoch": 3.29,
-      "eval_accuracy": 0.7917899408284024,
-      "eval_f1": 0.8013705698417323,
-      "eval_loss": 0.6606641411781311,
-      "eval_precision": 0.8190127006775076,
-      "eval_recall": 0.7917899408284024,
-      "eval_runtime": 24.3416,
-      "eval_samples_per_second": 666.514,
-      "eval_steps_per_second": 27.771,
-      "step": 10000
-    },
-    {
-      "epoch": 3.45,
-      "learning_rate": 6.2875e-06,
-      "loss": 0.3897,
-      "step": 10500
-    },
-    {
-      "epoch": 3.45,
-      "eval_accuracy": 0.788646449704142,
-      "eval_f1": 0.7987527519476123,
-      "eval_loss": 0.6613931059837341,
-      "eval_precision": 0.8177063830689373,
-      "eval_recall": 0.788646449704142,
-      "eval_runtime": 24.342,
-      "eval_samples_per_second": 666.503,
-      "eval_steps_per_second": 27.771,
-      "step": 10500
-    },
-    {
-      "epoch": 3.62,
-      "learning_rate": 4.204166666666667e-06,
-      "loss": 0.3877,
-      "step": 11000
-    },
-    {
-      "epoch": 3.62,
-      "eval_accuracy": 0.8135478303747534,
-      "eval_f1": 0.8169879527109274,
-      "eval_loss": 0.6640126705169678,
-      "eval_precision": 0.8215474627622835,
-      "eval_recall": 0.8135478303747534,
-      "eval_runtime": 24.3344,
-      "eval_samples_per_second": 666.71,
-      "eval_steps_per_second": 27.78,
-      "step": 11000
-    },
-    {
-      "epoch": 3.78,
-      "learning_rate": 2.1208333333333335e-06,
-      "loss": 0.3795,
-      "step": 11500
-    },
-    {
-      "epoch": 3.78,
-      "eval_accuracy": 0.8053500986193294,
-      "eval_f1": 0.8113411583628731,
-      "eval_loss": 0.6599082350730896,
-      "eval_precision": 0.8205901740056264,
-      "eval_recall": 0.8053500986193294,
-      "eval_runtime": 24.3271,
-      "eval_samples_per_second": 666.91,
-      "eval_steps_per_second": 27.788,
-      "step": 11500
-    },
-    {
-      "epoch": 3.94,
-      "learning_rate": 3.7500000000000005e-08,
-      "loss": 0.3863,
-      "step": 12000
-    },
-    {
-      "epoch": 3.94,
-      "eval_accuracy": 0.8009122287968442,
-      "eval_f1": 0.8081790831347971,
-      "eval_loss": 0.6572125554084778,
-      "eval_precision": 0.8200806840462704,
-      "eval_recall": 0.8009122287968442,
-      "eval_runtime": 24.3366,
-      "eval_samples_per_second": 666.651,
-      "eval_steps_per_second": 27.777,
-      "step": 12000
-    },
-    {
-      "epoch": 3.94,
-      "step": 12000,
-      "total_flos": 1.4962346001065574e+17,
-      "train_loss": 0.5341533139546712,
-      "train_runtime": 3936.6668,
-      "train_samples_per_second": 146.317,
-      "train_steps_per_second": 3.048
+      "epoch": 4.38,
+      "step": 5001,
+      "total_flos": 1.6838837781764506e+17,
+      "train_loss": 0.4945967703420051,
+      "train_runtime": 2688.9042,
+      "train_samples_per_second": 238.063,
+      "train_steps_per_second": 1.86
     }
   ],
-  "max_steps": 12000,
-  "num_train_epochs": 4,
-  "total_flos": 1.4962346001065574e+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 sha256:5972df5698b926a1dd5ed254cf898167128666e13f9d4375f9c4a79adc0d7e20
+oid sha256:ddb8c2875c5f66f935f06a86d2ad870e85baa6cb41cfcc194feb29fc278345d1
 size 3375