Upload folder using huggingface_hub

archive.tar.gz/Model3.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:451e8dd90599dc55001a410b3f5ca46eceec25a3ce29338839d3059a3e59a19c
+size 2986632

archive.tar.gz/model3_attention_20260518_121438/training.log

@@ -0,0 +1,28 @@
+[12:14:38] ======================================================================
+[12:14:38] MODEL 3: DS-CNN + Temporal Self-Attention — model3_attention_20260518_121438
+[12:14:38] GPUs: 1  Epochs: 5
+[12:14:38] Batch: 128  LR: 0.001  Heads: 4
+[12:14:38] Focal gamma=2.0  Mixup=0.4  CutMix=1.0
+[12:14:38] Cosine T0=10  T_mult=2
+[12:14:38] Hard Neg Mining: buffer=10000, thresh=0.3
+[12:14:38] Output: /data/Platform_Build/V6_Experiment/Step5_Training_Data/Model3/runs/model3_attention_20260518_121438
+[12:14:38] ======================================================================
+[12:14:39] Dataset: train=620,448  val=77,554  test=77,554
+[12:14:39]   hey_ava             :     27,792
+[12:14:39]   hey_t_mobile        :     33,562
+[12:14:39]   hey_turkcell        :     49,736
+[12:14:39]   hey_cassandra       :     28,886
+[12:14:39]   hey_dtag            :     28,392
+[12:14:39]   negative            :    452,080
+[12:14:39] Model: 474,758 params, 1.81 MB
+[12:14:39] Class weights:
+[12:14:39]   hey_ava             : 1.3669
+[12:14:39]   hey_t_mobile        : 1.1319
+[12:14:39]   hey_turkcell        : 0.7638
+[12:14:39]   hey_cassandra       : 1.3152
+[12:14:39]   hey_dtag            : 1.3381
+[12:14:39]   negative            : 0.0840
+[12:14:40] 
+Starting training: 5 epochs, patience=15
+[12:14:40] Epoch   TrLoss   TrAcc   VaLoss   VaAcc    VaF1      FAR      FRR  HardN         LR   Time
+[12:14:40] ────────────────────────────────────────────────────────────────────────────────────────────────────

archive.tar.gz/model3_attention_20260518_124432/training.log

@@ -0,0 +1,28 @@
+[12:44:32] ======================================================================
+[12:44:32] MODEL 3: DS-CNN + Temporal Self-Attention — model3_attention_20260518_124432
+[12:44:32] GPUs: 1  Epochs: 5
+[12:44:32] Batch: 64  LR: 0.001  Heads: 4
+[12:44:32] Focal gamma=2.0  Mixup=0.4  CutMix=1.0
+[12:44:32] Cosine T0=10  T_mult=2
+[12:44:32] Hard Neg Mining: buffer=10000, thresh=0.3
+[12:44:32] Output: /data/Platform_Build/V6_Experiment/Step5_Training_Data/Model3/runs/model3_attention_20260518_124432
+[12:44:32] ======================================================================
+[12:44:33] Dataset: train=620,448  val=77,554  test=77,554
+[12:44:33]   hey_ava             :     27,792
+[12:44:33]   hey_t_mobile        :     33,562
+[12:44:33]   hey_turkcell        :     49,736
+[12:44:33]   hey_cassandra       :     28,886
+[12:44:33]   hey_dtag            :     28,392
+[12:44:33]   negative            :    452,080
+[12:44:33] Model: 474,758 params, 1.81 MB
+[12:44:33] Class weights:
+[12:44:33]   hey_ava             : 1.3669
+[12:44:33]   hey_t_mobile        : 1.1319
+[12:44:33]   hey_turkcell        : 0.7638
+[12:44:33]   hey_cassandra       : 1.3152
+[12:44:33]   hey_dtag            : 1.3381
+[12:44:33]   negative            : 0.0840
+[12:44:34] 
+Starting training: 5 epochs, patience=15
+[12:44:34] Epoch   TrLoss   TrAcc   VaLoss   VaAcc    VaF1      FAR      FRR  HardN         LR   Time
+[12:44:34] ────────────────────────────────────────────────────────────────────────────────────────────────────

archive.tar.gz/model3_attention_20260518_132913/training.log

@@ -0,0 +1,29 @@
+[13:29:13] ======================================================================
+[13:29:13] MODEL 3: DS-CNN + Temporal Self-Attention — model3_attention_20260518_132913
+[13:29:13] GPUs: 1  Epochs: 5
+[13:29:13] Batch: 64  LR: 0.001  Heads: 4
+[13:29:13] Focal gamma=2.0  Mixup=0.4  CutMix=1.0
+[13:29:13] Cosine T0=10  T_mult=2
+[13:29:13] Hard Neg Mining: buffer=10000, thresh=0.3
+[13:29:13] Output: /data/Platform_Build/V6_Experiment/Step5_Training_Data/Model3/runs/model3_attention_20260518_132913
+[13:29:13] ======================================================================
+[13:29:13] Dataset: train=620,448  val=77,554  test=77,554
+[13:29:13]   hey_ava             :     27,792
+[13:29:13]   hey_t_mobile        :     33,562
+[13:29:13]   hey_turkcell        :     49,736
+[13:29:13]   hey_cassandra       :     28,886
+[13:29:13]   hey_dtag            :     28,392
+[13:29:13]   negative            :    452,080
+[13:29:14] Model: 474,758 params, 1.81 MB
+[13:29:14] Class weights:
+[13:29:14]   hey_ava             : 1.3669
+[13:29:14]   hey_t_mobile        : 1.1319
+[13:29:14]   hey_turkcell        : 0.7638
+[13:29:14]   hey_cassandra       : 1.3152
+[13:29:14]   hey_dtag            : 1.3381
+[13:29:14]   negative            : 0.0840
+[13:29:15] 
+Starting training: 5 epochs, patience=15
+[13:29:15] Epoch   TrLoss   TrAcc   VaLoss   VaAcc    VaF1      FAR      FRR  HardN         LR   Time
+[13:29:15] ────────────────────────────────────────────────────────────────────────────────────────────────────
+[15:03:34]     1   0.3031   12.2%   0.1678   20.8%  28.09% 100.000%   0.000% 10000   9.76e-04 5659.2s

archive.tar.gz/model3_attention_20260519_062645/.ipynb_checkpoints/training-checkpoint.log

@@ -0,0 +1,28 @@
+[06:26:45] ======================================================================
+[06:26:45] MODEL 3: DS-CNN + Temporal Self-Attention — model3_attention_20260519_062645
+[06:26:45] GPUs: 1  Epochs: 200
+[06:26:45] Batch: 64  LR: 0.001  Heads: 4
+[06:26:45] Focal gamma=2.0  Mixup=0.4  CutMix=1.0
+[06:26:45] Cosine T0=10  T_mult=2
+[06:26:45] Hard Neg Mining: buffer=10000, thresh=0.3
+[06:26:45] Output: /opt/app-root/src/data/aditya_wakeword/Step5_Training_Data/Model3/runs/model3_attention_20260519_062645
+[06:26:45] ======================================================================
+[06:26:48] Dataset: train=620,448  val=77,554  test=77,554
+[06:26:48]   hey_ava             :     27,792
+[06:26:48]   hey_t_mobile        :     33,562
+[06:26:48]   hey_turkcell        :     49,736
+[06:26:48]   hey_cassandra       :     28,886
+[06:26:48]   hey_dtag            :     28,392
+[06:26:48]   negative            :    452,080
+[06:26:48] Model: 474,758 params, 1.81 MB
+[06:26:48] Class weights:
+[06:26:48]   hey_ava             : 1.3669
+[06:26:48]   hey_t_mobile        : 1.1319
+[06:26:48]   hey_turkcell        : 0.7638
+[06:26:48]   hey_cassandra       : 1.3152
+[06:26:48]   hey_dtag            : 1.3381
+[06:26:48]   negative            : 0.0840
+[06:26:50] 
+Starting training: 200 epochs, patience=15
+[06:26:50] Epoch   TrLoss   TrAcc   VaLoss   VaAcc    VaF1      FAR      FRR  HardN         LR   Time
+[06:26:50] ────────────────────────────────────────────────────────────────────────────────────────────────────

archive.tar.gz/model3_attention_20260519_062645/best_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ec95fc72c6a6e7ebbdd9143e68389b46fa6d3a50e183f6ce83c4fe1cba59844
+size 5738498

archive.tar.gz/model3_attention_20260519_062645/training.log

@@ -0,0 +1,33 @@
+[06:26:45] ======================================================================
+[06:26:45] MODEL 3: DS-CNN + Temporal Self-Attention — model3_attention_20260519_062645
+[06:26:45] GPUs: 1  Epochs: 200
+[06:26:45] Batch: 64  LR: 0.001  Heads: 4
+[06:26:45] Focal gamma=2.0  Mixup=0.4  CutMix=1.0
+[06:26:45] Cosine T0=10  T_mult=2
+[06:26:45] Hard Neg Mining: buffer=10000, thresh=0.3
+[06:26:45] Output: /opt/app-root/src/data/aditya_wakeword/Step5_Training_Data/Model3/runs/model3_attention_20260519_062645
+[06:26:45] ======================================================================
+[06:26:48] Dataset: train=620,448  val=77,554  test=77,554
+[06:26:48]   hey_ava             :     27,792
+[06:26:48]   hey_t_mobile        :     33,562
+[06:26:48]   hey_turkcell        :     49,736
+[06:26:48]   hey_cassandra       :     28,886
+[06:26:48]   hey_dtag            :     28,392
+[06:26:48]   negative            :    452,080
+[06:26:48] Model: 474,758 params, 1.81 MB
+[06:26:48] Class weights:
+[06:26:48]   hey_ava             : 1.3669
+[06:26:48]   hey_t_mobile        : 1.1319
+[06:26:48]   hey_turkcell        : 0.7638
+[06:26:48]   hey_cassandra       : 1.3152
+[06:26:48]   hey_dtag            : 1.3381
+[06:26:48]   negative            : 0.0840
+[06:26:50] 
+Starting training: 200 epochs, patience=15
+[06:26:50] Epoch   TrLoss   TrAcc   VaLoss   VaAcc    VaF1      FAR      FRR  HardN         LR   Time
+[06:26:50] ────────────────────────────────────────────────────────────────────────────────────────────────────
+[07:29:42]     1   0.3917    4.6%   0.3906    4.6%   1.46% 100.000%   0.000% 10000   9.76e-04 3771.4s
+[07:29:42]   ✓ Best model saved (F1=1.46%, FAR=100.000%, FRR=0.000%)
+[08:34:15]     2   0.3908    4.6%   0.3906    4.5%   1.43% 100.000%   0.000% 10000   9.05e-04 3873.5s
+[09:37:48]     3   0.3907    4.6%   0.3905    4.7%   1.48% 100.000%   0.000% 10000   7.94e-04 3812.9s
+[09:37:48]   ✓ Best model saved (F1=1.48%, FAR=100.000%, FRR=0.000%)

archive.tar.gz/model3_attention_20260520_120607/.ipynb_checkpoints/training-checkpoint.log

@@ -0,0 +1,93 @@
+[12:06:07] ======================================================================
+[12:06:07] MODEL 3: DS-CNN + Temporal Self-Attention — model3_attention_20260520_120607
+[12:06:07] GPUs: 1  Epochs: 200
+[12:06:07] Batch: 64  LR: 0.001  Heads: 4
+[12:06:07] Focal gamma=2.0  Mixup=0.4  CutMix=1.0
+[12:06:07] Cosine T0=10  T_mult=2
+[12:06:07] Hard Neg Mining: buffer=10000, thresh=0.3
+[12:06:07] Output: /opt/app-root/src/data/aditya_wakeword/Step5_Training_Data/Model3/runs/model3_attention_20260520_120607
+[12:06:07] ======================================================================
+[12:06:08] Dataset: train=620,448  val=77,554  test=77,554
+[12:06:08]   hey_ava             :     27,792
+[12:06:08]   hey_t_mobile        :     33,562
+[12:06:08]   hey_turkcell        :     49,736
+[12:06:08]   hey_cassandra       :     28,886
+[12:06:08]   hey_dtag            :     28,392
+[12:06:08]   negative            :    452,080
+[12:06:08] Model: 474,758 params, 1.81 MB
+[12:06:08] Class weights:
+[12:06:08]   hey_ava             : 1.2306
+[12:06:08]   hey_t_mobile        : 1.1198
+[12:06:08]   hey_turkcell        : 0.9199
+[12:06:08]   hey_cassandra       : 1.2071
+[12:06:08]   hey_dtag            : 1.2175
+[12:06:08]   negative            : 0.3051
+[12:06:09] 
+Starting training: 200 epochs, patience=15
+[12:06:09] Epoch   TrLoss   TrAcc   VaLoss   VaAcc    VaF1      FAR      FRR  HardN         LR   Time
+[12:06:09] ────────────────────────────────────────────────────────────────────────────────────────────────────
+[12:14:56]     1   0.5089   71.5%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   9.76e-04 526.8s
+[12:14:56]   ✓ Best model saved (F1=14.05%, FAR=0.000%, FRR=100.000%)
+[12:23:48]     2   0.5080   72.8%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   9.05e-04 532.6s
+[12:32:48]     3   0.5079   72.8%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   7.94e-04 539.0s
+[12:41:35]     4   0.5079   72.8%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   6.55e-04 527.2s
+[12:50:20]     5   0.5079   72.8%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   5.00e-04 524.9s
+[12:59:05]     6   0.5079   72.9%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   3.45e-04 525.3s
+[13:07:56]     7   0.5078   72.9%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   2.06e-04 531.0s
+[13:16:45]     8   0.5078   72.9%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   9.55e-05 528.9s
+[13:25:30]     9   0.5078   72.9%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   2.45e-05 525.3s
+[13:34:15]    10   0.5078   72.9%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   1.00e-03 524.5s
+[13:43:01]    11   0.5079   72.8%   0.5079   72.9%  14.05%   0.000% 100.000% 10000   9.94e-04 525.9s
+[13:51:47]    12   0.5080   72.9%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   9.76e-04 526.8s
+[14:00:33]    13   0.5079   72.8%   0.5079   72.9%  14.05%   0.000% 100.000% 10000   9.46e-04 525.5s
+[14:09:06]    14   0.5079   72.8%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   9.05e-04 513.0s
+[14:17:39]    15   0.5079   72.8%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   8.54e-04 512.6s
+[14:26:11]    16   0.5079   72.9%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   7.94e-04 512.0s
+[14:26:11] 
+  Early stopping at epoch 16 (patience=15)
+[14:26:11] 
+======================================================================
+[14:26:11] FINAL TEST EVALUATION
+[14:26:11] ======================================================================
+[14:26:11] Loaded best model from epoch 1
+[14:27:14] 
+  Test Accuracy:  72.87%
+[14:27:14]   Test Macro F1:  14.05%
+[14:27:14]   Overall FAR:    0.0000%
+[14:27:14]   Overall FRR:    100.0000%
+[14:27:14] 
+  Per-class FAR:
+[14:27:14]     hey_ava             : 0.0000%
+[14:27:14]     hey_t_mobile        : 0.0000%
+[14:27:14]     hey_turkcell        : 0.0000%
+[14:27:14]     hey_cassandra       : 0.0000%
+[14:27:14]     hey_dtag            : 0.0000%
+[14:27:14] 
+  Per-class FRR:
+[14:27:14]     hey_ava             : 100.0000%
+[14:27:14]     hey_t_mobile        : 100.0000%
+[14:27:14]     hey_turkcell        : 100.0000%
+[14:27:14]     hey_cassandra       : 100.0000%
+[14:27:14]     hey_dtag            : 100.0000%
+[14:27:14] 
+  Classification Report:
+               precision    recall  f1-score   support
+
+      hey_ava       0.00      0.00      0.00      3473
+ hey_t_mobile       0.00      0.00      0.00      4195
+ hey_turkcell       0.00      0.00      0.00      6217
+hey_cassandra       0.00      0.00      0.00      3610
+     hey_dtag       0.00      0.00      0.00      3549
+     negative       0.73      1.00      0.84     56510
+
+     accuracy                           0.73     77554
+    macro avg       0.12      0.17      0.14     77554
+ weighted avg       0.53      0.73      0.61     77554
+
+[14:27:15] 
+  PDF Report: /opt/app-root/src/data/aditya_wakeword/Step5_Training_Data/Model3/runs/model3_attention_20260520_120607/training_report.pdf
+[14:27:15] 
+  All results saved to: /opt/app-root/src/data/aditya_wakeword/Step5_Training_Data/Model3/runs/model3_attention_20260520_120607
+[14:27:15] ======================================================================
+[14:27:15] TRAINING COMPLETE
+[14:27:15] ======================================================================

archive.tar.gz/model3_attention_20260520_120607/best_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38a2df666ed8b43495a67d0187e2ceed51b9b3f7bdb229f6c830a0fdb4742049
+size 5738114

archive.tar.gz/model3_attention_20260520_120607/test_results.json

@@ -0,0 +1,104 @@
+{
+  "run_name": "model3_attention_20260520_120607",
+  "model_name": "Model3_DSCNN_Attention",
+  "n_params": "474,758",
+  "best_epoch": 1,
+  "best_val_f1": 14.050503242232566,
+  "test_accuracy": 72.86535833096939,
+  "test_macro_f1": 14.050503242232566,
+  "test_far": 0.0,
+  "test_frr": 100.0,
+  "per_class_far": {
+    "hey_ava": 0.0,
+    "hey_t_mobile": 0.0,
+    "hey_turkcell": 0.0,
+    "hey_cassandra": 0.0,
+    "hey_dtag": 0.0
+  },
+  "per_class_frr": {
+    "hey_ava": 1.0,
+    "hey_t_mobile": 1.0,
+    "hey_turkcell": 1.0,
+    "hey_cassandra": 1.0,
+    "hey_dtag": 1.0
+  },
+  "confusion_matrix": [
+    [
+      0,
+      0,
+      0,
+      0,
+      0,
+      3473
+    ],
+    [
+      0,
+      0,
+      0,
+      0,
+      0,
+      4195
+    ],
+    [
+      0,
+      0,
+      0,
+      0,
+      0,
+      6217
+    ],
+    [
+      0,
+      0,
+      0,
+      0,
+      0,
+      3610
+    ],
+    [
+      0,
+      0,
+      0,
+      0,
+      0,
+      3549
+    ],
+    [
+      0,
+      0,
+      0,
+      0,
+      0,
+      56510
+    ]
+  ],
+  "classification_report": "               precision    recall  f1-score   support\n\n      hey_ava       0.00      0.00      0.00      3473\n hey_t_mobile       0.00      0.00      0.00      4195\n hey_turkcell       0.00      0.00      0.00      6217\nhey_cassandra       0.00      0.00      0.00      3610\n     hey_dtag       0.00      0.00      0.00      3549\n     negative       0.73      1.00      0.84     56510\n\n     accuracy                           0.73     77554\n    macro avg       0.12      0.17      0.14     77554\n weighted avg       0.53      0.73      0.61     77554\n",
+  "args": {
+    "epochs": 200,
+    "batch_size": 64,
+    "lr": 0.001,
+    "embedding_dim": 256,
+    "dropout": 0.1,
+    "num_heads": 4,
+    "focal_gamma": 2.0,
+    "mixup_alpha": 0.4,
+    "cutmix_alpha": 1.0,
+    "cosine_t0": 10,
+    "cosine_t_mult": 2,
+    "hard_neg_buffer": 10000,
+    "hard_neg_thresh": 0.3,
+    "patience": 15,
+    "num_workers": 4,
+    "seed": 42,
+    "single_gpu": 1,
+    "port": "29502"
+  },
+  "class_names": [
+    "hey_ava",
+    "hey_t_mobile",
+    "hey_turkcell",
+    "hey_cassandra",
+    "hey_dtag",
+    "negative"
+  ]
+}

archive.tar.gz/model3_attention_20260520_120607/training.log

@@ -0,0 +1,93 @@
+[12:06:07] ======================================================================
+[12:06:07] MODEL 3: DS-CNN + Temporal Self-Attention — model3_attention_20260520_120607
+[12:06:07] GPUs: 1  Epochs: 200
+[12:06:07] Batch: 64  LR: 0.001  Heads: 4
+[12:06:07] Focal gamma=2.0  Mixup=0.4  CutMix=1.0
+[12:06:07] Cosine T0=10  T_mult=2
+[12:06:07] Hard Neg Mining: buffer=10000, thresh=0.3
+[12:06:07] Output: /opt/app-root/src/data/aditya_wakeword/Step5_Training_Data/Model3/runs/model3_attention_20260520_120607
+[12:06:07] ======================================================================
+[12:06:08] Dataset: train=620,448  val=77,554  test=77,554
+[12:06:08]   hey_ava             :     27,792
+[12:06:08]   hey_t_mobile        :     33,562
+[12:06:08]   hey_turkcell        :     49,736
+[12:06:08]   hey_cassandra       :     28,886
+[12:06:08]   hey_dtag            :     28,392
+[12:06:08]   negative            :    452,080
+[12:06:08] Model: 474,758 params, 1.81 MB
+[12:06:08] Class weights:
+[12:06:08]   hey_ava             : 1.2306
+[12:06:08]   hey_t_mobile        : 1.1198
+[12:06:08]   hey_turkcell        : 0.9199
+[12:06:08]   hey_cassandra       : 1.2071
+[12:06:08]   hey_dtag            : 1.2175
+[12:06:08]   negative            : 0.3051
+[12:06:09] 
+Starting training: 200 epochs, patience=15
+[12:06:09] Epoch   TrLoss   TrAcc   VaLoss   VaAcc    VaF1      FAR      FRR  HardN         LR   Time
+[12:06:09] ────────────────────────────────────────────────────────────────────────────────────────────────────
+[12:14:56]     1   0.5089   71.5%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   9.76e-04 526.8s
+[12:14:56]   ✓ Best model saved (F1=14.05%, FAR=0.000%, FRR=100.000%)
+[12:23:48]     2   0.5080   72.8%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   9.05e-04 532.6s
+[12:32:48]     3   0.5079   72.8%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   7.94e-04 539.0s
+[12:41:35]     4   0.5079   72.8%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   6.55e-04 527.2s
+[12:50:20]     5   0.5079   72.8%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   5.00e-04 524.9s
+[12:59:05]     6   0.5079   72.9%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   3.45e-04 525.3s
+[13:07:56]     7   0.5078   72.9%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   2.06e-04 531.0s
+[13:16:45]     8   0.5078   72.9%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   9.55e-05 528.9s
+[13:25:30]     9   0.5078   72.9%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   2.45e-05 525.3s
+[13:34:15]    10   0.5078   72.9%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   1.00e-03 524.5s
+[13:43:01]    11   0.5079   72.8%   0.5079   72.9%  14.05%   0.000% 100.000% 10000   9.94e-04 525.9s
+[13:51:47]    12   0.5080   72.9%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   9.76e-04 526.8s
+[14:00:33]    13   0.5079   72.8%   0.5079   72.9%  14.05%   0.000% 100.000% 10000   9.46e-04 525.5s
+[14:09:06]    14   0.5079   72.8%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   9.05e-04 513.0s
+[14:17:39]    15   0.5079   72.8%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   8.54e-04 512.6s
+[14:26:11]    16   0.5079   72.9%   0.5078   72.9%  14.05%   0.000% 100.000% 10000   7.94e-04 512.0s
+[14:26:11] 
+  Early stopping at epoch 16 (patience=15)
+[14:26:11] 
+======================================================================
+[14:26:11] FINAL TEST EVALUATION
+[14:26:11] ======================================================================
+[14:26:11] Loaded best model from epoch 1
+[14:27:14] 
+  Test Accuracy:  72.87%
+[14:27:14]   Test Macro F1:  14.05%
+[14:27:14]   Overall FAR:    0.0000%
+[14:27:14]   Overall FRR:    100.0000%
+[14:27:14] 
+  Per-class FAR:
+[14:27:14]     hey_ava             : 0.0000%
+[14:27:14]     hey_t_mobile        : 0.0000%
+[14:27:14]     hey_turkcell        : 0.0000%
+[14:27:14]     hey_cassandra       : 0.0000%
+[14:27:14]     hey_dtag            : 0.0000%
+[14:27:14] 
+  Per-class FRR:
+[14:27:14]     hey_ava             : 100.0000%
+[14:27:14]     hey_t_mobile        : 100.0000%
+[14:27:14]     hey_turkcell        : 100.0000%
+[14:27:14]     hey_cassandra       : 100.0000%
+[14:27:14]     hey_dtag            : 100.0000%
+[14:27:14] 
+  Classification Report:
+               precision    recall  f1-score   support
+
+      hey_ava       0.00      0.00      0.00      3473
+ hey_t_mobile       0.00      0.00      0.00      4195
+ hey_turkcell       0.00      0.00      0.00      6217
+hey_cassandra       0.00      0.00      0.00      3610
+     hey_dtag       0.00      0.00      0.00      3549
+     negative       0.73      1.00      0.84     56510
+
+     accuracy                           0.73     77554
+    macro avg       0.12      0.17      0.14     77554
+ weighted avg       0.53      0.73      0.61     77554
+
+[14:27:15] 
+  PDF Report: /opt/app-root/src/data/aditya_wakeword/Step5_Training_Data/Model3/runs/model3_attention_20260520_120607/training_report.pdf
+[14:27:15] 
+  All results saved to: /opt/app-root/src/data/aditya_wakeword/Step5_Training_Data/Model3/runs/model3_attention_20260520_120607
+[14:27:15] ======================================================================
+[14:27:15] TRAINING COMPLETE
+[14:27:15] ======================================================================

archive.tar.gz/model3_attention_20260520_120607/training_history.json

@@ -0,0 +1,164 @@
+{
+  "epoch": [
+    1,
+    2,
+    3,
+    4,
+    5,
+    6,
+    7,
+    8,
+    9,
+    10,
+    11,
+    12,
+    13,
+    14,
+    15,
+    16
+  ],
+  "train_loss": [
+    0.5089186223927941,
+    0.5079592704096672,
+    0.5079361379899703,
+    0.5079197311290704,
+    0.507888638421455,
+    0.5078568547994836,
+    0.507841204999033,
+    0.5078368117524644,
+    0.5078120109652539,
+    0.5078068754917694,
+    0.5079417400872849,
+    0.5079604534973122,
+    0.5079326448611465,
+    0.5079439199988285,
+    0.5079238594100917,
+    0.5079193356696203
+  ],
+  "train_acc": [
+    71.48769921089277,
+    72.82173526226211,
+    72.82769869513642,
+    72.84188199494558,
+    72.81013074423642,
+    72.86347929238228,
+    72.86347929238228,
+    72.86347929238228,
+    72.86347929238228,
+    72.86347929238228,
+    72.81432126463459,
+    72.86347929238228,
+    72.78772757749239,
+    72.84994068801898,
+    72.84816777554283,
+    72.86347929238228
+  ],
+  "val_loss": [
+    0.5078213921818796,
+    0.5078061159835396,
+    0.5078027595830436,
+    0.5077764186086041,
+    0.507795118164308,
+    0.5077809013500072,
+    0.5077789777232082,
+    0.5077743728463799,
+    0.5077709429236529,
+    0.5077690220754533,
+    0.5078863813587935,
+    0.5077990599640525,
+    0.5078714346826667,
+    0.5077777711963496,
+    0.5077813545321868,
+    0.5078466098153158
+  ],
+  "val_acc": [
+    72.86535833096939,
+    72.86535833096939,
+    72.86535833096939,
+    72.86535833096939,
+    72.86535833096939,
+    72.86535833096939,
+    72.86535833096939,
+    72.86535833096939,
+    72.86535833096939,
+    72.86535833096939,
+    72.86535833096939,
+    72.86535833096939,
+    72.86535833096939,
+    72.86535833096939,
+    72.86535833096939,
+    72.86535833096939
+  ],
+  "val_f1": [
+    14.050503242232566,
+    14.050503242232566,
+    14.050503242232566,
+    14.050503242232566,
+    14.050503242232566,
+    14.050503242232566,
+    14.050503242232566,
+    14.050503242232566,
+    14.050503242232566,
+    14.050503242232566,
+    14.050503242232566,
+    14.050503242232566,
+    14.050503242232566,
+    14.050503242232566,
+    14.050503242232566,
+    14.050503242232566
+  ],
+  "val_far": [
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0
+  ],
+  "val_frr": [
+    100.0,
+    100.0,
+    100.0,
+    100.0,
+    100.0,
+    100.0,
+    100.0,
+    100.0,
+    100.0,
+    100.0,
+    100.0,
+    100.0,
+    100.0,
+    100.0,
+    100.0,
+    100.0
+  ],
+  "lr": [
+    0.0009755282581475768,
+    0.0009045084971874737,
+    0.0007938926261462366,
+    0.0006545084971874737,
+    0.0005,
+    0.00034549150281252633,
+    0.00020610737385376348,
+    9.549150281252633e-05,
+    2.4471741852423235e-05,
+    0.001,
+    0.0009938441702975688,
+    0.0009755282581475768,
+    0.0009455032620941839,
+    0.0009045084971874737,
+    0.0008535533905932737,
+    0.0007938926261462366
+  ]
+}