Upload 44 files

.gitattributes

@@ -33,3 +33,9 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+src/nn_binary.keras filter=lfs diff=lfs merge=lfs -text
+src/nn_category.keras filter=lfs diff=lfs merge=lfs -text
+src/nn_categorys.keras filter=lfs diff=lfs merge=lfs -text
+src/nn_vectorizer_binary.keras filter=lfs diff=lfs merge=lfs -text
+src/nn_vectorizer_category.keras filter=lfs diff=lfs merge=lfs -text
+src/nn_vectorizer_categorys.keras filter=lfs diff=lfs merge=lfs -text

src/binary_model/checkpoint-400/config.json

@@ -0,0 +1,35 @@
+{
+  "architectures": [
+    "BertForSequenceClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "dtype": "float32",
+  "emb_size": 312,
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 312,
+  "id2label": {
+    "0": "negative",
+    "1": "positive"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 600,
+  "label2id": {
+    "negative": 0,
+    "positive": 1
+  },
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 3,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "problem_type": "single_label_classification",
+  "transformers_version": "4.57.0",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 29564
+}

src/binary_model/checkpoint-400/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:452079fe712b0a898545d8ec8d36ae4dbca4e4f34a67ee61bce99be89efb0276
+size 47145624

src/binary_model/checkpoint-400/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98b2f3d51ee4fb8db44f65ab5aa0dfcb6a3a42faac9a58970d809a7fac691ff5
+size 94323147

src/binary_model/checkpoint-400/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:987e2565b4d0e8df1d1d7fabe9aae58ea75b005c720faa6553599a28da5eb789
+size 14455

src/binary_model/checkpoint-400/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d2a13aedfb10730f658c90eb03b45908f32db49971da0e67d91a64b15a963525
+size 1465

src/binary_model/checkpoint-400/trainer_state.json

@@ -0,0 +1,734 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 100.0,
+  "eval_steps": 500,
+  "global_step": 400,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.7171610593795776,
+      "learning_rate": 1.985e-05,
+      "loss": 0.6936,
+      "step": 4
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 0.931102991104126,
+      "learning_rate": 1.9650000000000003e-05,
+      "loss": 0.6903,
+      "step": 8
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 0.9753431677818298,
+      "learning_rate": 1.9450000000000002e-05,
+      "loss": 0.691,
+      "step": 12
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 0.7282153964042664,
+      "learning_rate": 1.925e-05,
+      "loss": 0.6909,
+      "step": 16
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 0.7919716238975525,
+      "learning_rate": 1.9050000000000002e-05,
+      "loss": 0.6853,
+      "step": 20
+    },
+    {
+      "epoch": 6.0,
+      "grad_norm": 2.4427859783172607,
+      "learning_rate": 1.885e-05,
+      "loss": 0.6789,
+      "step": 24
+    },
+    {
+      "epoch": 7.0,
+      "grad_norm": 0.7891402840614319,
+      "learning_rate": 1.8650000000000003e-05,
+      "loss": 0.6753,
+      "step": 28
+    },
+    {
+      "epoch": 8.0,
+      "grad_norm": 0.6697407960891724,
+      "learning_rate": 1.845e-05,
+      "loss": 0.6762,
+      "step": 32
+    },
+    {
+      "epoch": 9.0,
+      "grad_norm": 0.7295678853988647,
+      "learning_rate": 1.825e-05,
+      "loss": 0.6714,
+      "step": 36
+    },
+    {
+      "epoch": 10.0,
+      "grad_norm": 0.7619945406913757,
+      "learning_rate": 1.805e-05,
+      "loss": 0.6697,
+      "step": 40
+    },
+    {
+      "epoch": 11.0,
+      "grad_norm": 1.8652944564819336,
+      "learning_rate": 1.785e-05,
+      "loss": 0.6609,
+      "step": 44
+    },
+    {
+      "epoch": 12.0,
+      "grad_norm": 0.9776553511619568,
+      "learning_rate": 1.7650000000000002e-05,
+      "loss": 0.6554,
+      "step": 48
+    },
+    {
+      "epoch": 13.0,
+      "grad_norm": 0.8226175308227539,
+      "learning_rate": 1.7450000000000004e-05,
+      "loss": 0.6505,
+      "step": 52
+    },
+    {
+      "epoch": 14.0,
+      "grad_norm": 1.9432940483093262,
+      "learning_rate": 1.7250000000000003e-05,
+      "loss": 0.6451,
+      "step": 56
+    },
+    {
+      "epoch": 15.0,
+      "grad_norm": 1.1705070734024048,
+      "learning_rate": 1.705e-05,
+      "loss": 0.6293,
+      "step": 60
+    },
+    {
+      "epoch": 16.0,
+      "grad_norm": 1.1913769245147705,
+      "learning_rate": 1.6850000000000003e-05,
+      "loss": 0.6219,
+      "step": 64
+    },
+    {
+      "epoch": 17.0,
+      "grad_norm": 1.1586858034133911,
+      "learning_rate": 1.665e-05,
+      "loss": 0.6151,
+      "step": 68
+    },
+    {
+      "epoch": 18.0,
+      "grad_norm": 1.3686275482177734,
+      "learning_rate": 1.645e-05,
+      "loss": 0.6057,
+      "step": 72
+    },
+    {
+      "epoch": 19.0,
+      "grad_norm": 1.2270820140838623,
+      "learning_rate": 1.6250000000000002e-05,
+      "loss": 0.5921,
+      "step": 76
+    },
+    {
+      "epoch": 20.0,
+      "grad_norm": 2.155693531036377,
+      "learning_rate": 1.605e-05,
+      "loss": 0.5771,
+      "step": 80
+    },
+    {
+      "epoch": 21.0,
+      "grad_norm": 1.8586078882217407,
+      "learning_rate": 1.5850000000000002e-05,
+      "loss": 0.562,
+      "step": 84
+    },
+    {
+      "epoch": 22.0,
+      "grad_norm": 2.844381809234619,
+      "learning_rate": 1.565e-05,
+      "loss": 0.5483,
+      "step": 88
+    },
+    {
+      "epoch": 23.0,
+      "grad_norm": 1.532677412033081,
+      "learning_rate": 1.545e-05,
+      "loss": 0.5323,
+      "step": 92
+    },
+    {
+      "epoch": 24.0,
+      "grad_norm": 2.501610040664673,
+      "learning_rate": 1.525e-05,
+      "loss": 0.5106,
+      "step": 96
+    },
+    {
+      "epoch": 25.0,
+      "grad_norm": 3.366448402404785,
+      "learning_rate": 1.505e-05,
+      "loss": 0.5028,
+      "step": 100
+    },
+    {
+      "epoch": 26.0,
+      "grad_norm": 2.540175199508667,
+      "learning_rate": 1.4850000000000002e-05,
+      "loss": 0.4743,
+      "step": 104
+    },
+    {
+      "epoch": 27.0,
+      "grad_norm": 2.1043853759765625,
+      "learning_rate": 1.4650000000000002e-05,
+      "loss": 0.4676,
+      "step": 108
+    },
+    {
+      "epoch": 28.0,
+      "grad_norm": 2.4121694564819336,
+      "learning_rate": 1.4450000000000002e-05,
+      "loss": 0.4418,
+      "step": 112
+    },
+    {
+      "epoch": 29.0,
+      "grad_norm": 1.871059775352478,
+      "learning_rate": 1.425e-05,
+      "loss": 0.4188,
+      "step": 116
+    },
+    {
+      "epoch": 30.0,
+      "grad_norm": 3.22082257270813,
+      "learning_rate": 1.4050000000000001e-05,
+      "loss": 0.3973,
+      "step": 120
+    },
+    {
+      "epoch": 31.0,
+      "grad_norm": 2.0184738636016846,
+      "learning_rate": 1.3850000000000001e-05,
+      "loss": 0.3767,
+      "step": 124
+    },
+    {
+      "epoch": 32.0,
+      "grad_norm": 1.8004070520401,
+      "learning_rate": 1.3650000000000001e-05,
+      "loss": 0.3687,
+      "step": 128
+    },
+    {
+      "epoch": 33.0,
+      "grad_norm": 2.161533832550049,
+      "learning_rate": 1.3450000000000002e-05,
+      "loss": 0.3419,
+      "step": 132
+    },
+    {
+      "epoch": 34.0,
+      "grad_norm": 2.215999126434326,
+      "learning_rate": 1.325e-05,
+      "loss": 0.3259,
+      "step": 136
+    },
+    {
+      "epoch": 35.0,
+      "grad_norm": 1.8289316892623901,
+      "learning_rate": 1.305e-05,
+      "loss": 0.2965,
+      "step": 140
+    },
+    {
+      "epoch": 36.0,
+      "grad_norm": 1.7603213787078857,
+      "learning_rate": 1.285e-05,
+      "loss": 0.2784,
+      "step": 144
+    },
+    {
+      "epoch": 37.0,
+      "grad_norm": 1.9211527109146118,
+      "learning_rate": 1.2650000000000001e-05,
+      "loss": 0.2624,
+      "step": 148
+    },
+    {
+      "epoch": 38.0,
+      "grad_norm": 1.7408591508865356,
+      "learning_rate": 1.2450000000000003e-05,
+      "loss": 0.2301,
+      "step": 152
+    },
+    {
+      "epoch": 39.0,
+      "grad_norm": 1.8422377109527588,
+      "learning_rate": 1.2250000000000001e-05,
+      "loss": 0.2316,
+      "step": 156
+    },
+    {
+      "epoch": 40.0,
+      "grad_norm": 2.905261754989624,
+      "learning_rate": 1.2050000000000002e-05,
+      "loss": 0.2066,
+      "step": 160
+    },
+    {
+      "epoch": 41.0,
+      "grad_norm": 1.5432759523391724,
+      "learning_rate": 1.1850000000000002e-05,
+      "loss": 0.2084,
+      "step": 164
+    },
+    {
+      "epoch": 42.0,
+      "grad_norm": 1.6602318286895752,
+      "learning_rate": 1.1650000000000002e-05,
+      "loss": 0.1901,
+      "step": 168
+    },
+    {
+      "epoch": 43.0,
+      "grad_norm": 1.7276387214660645,
+      "learning_rate": 1.145e-05,
+      "loss": 0.1635,
+      "step": 172
+    },
+    {
+      "epoch": 44.0,
+      "grad_norm": 3.0626723766326904,
+      "learning_rate": 1.125e-05,
+      "loss": 0.1493,
+      "step": 176
+    },
+    {
+      "epoch": 45.0,
+      "grad_norm": 1.6950130462646484,
+      "learning_rate": 1.1050000000000001e-05,
+      "loss": 0.128,
+      "step": 180
+    },
+    {
+      "epoch": 46.0,
+      "grad_norm": 1.41054105758667,
+      "learning_rate": 1.0850000000000001e-05,
+      "loss": 0.1241,
+      "step": 184
+    },
+    {
+      "epoch": 47.0,
+      "grad_norm": 1.694176435470581,
+      "learning_rate": 1.065e-05,
+      "loss": 0.126,
+      "step": 188
+    },
+    {
+      "epoch": 48.0,
+      "grad_norm": 1.3726774454116821,
+      "learning_rate": 1.045e-05,
+      "loss": 0.1127,
+      "step": 192
+    },
+    {
+      "epoch": 49.0,
+      "grad_norm": 2.0337917804718018,
+      "learning_rate": 1.025e-05,
+      "loss": 0.1056,
+      "step": 196
+    },
+    {
+      "epoch": 50.0,
+      "grad_norm": 1.3560911417007446,
+      "learning_rate": 1.005e-05,
+      "loss": 0.0995,
+      "step": 200
+    },
+    {
+      "epoch": 51.0,
+      "grad_norm": 1.0479848384857178,
+      "learning_rate": 9.85e-06,
+      "loss": 0.0848,
+      "step": 204
+    },
+    {
+      "epoch": 52.0,
+      "grad_norm": 0.9078042507171631,
+      "learning_rate": 9.65e-06,
+      "loss": 0.0789,
+      "step": 208
+    },
+    {
+      "epoch": 53.0,
+      "grad_norm": 1.6278938055038452,
+      "learning_rate": 9.450000000000001e-06,
+      "loss": 0.077,
+      "step": 212
+    },
+    {
+      "epoch": 54.0,
+      "grad_norm": 1.9590917825698853,
+      "learning_rate": 9.250000000000001e-06,
+      "loss": 0.0807,
+      "step": 216
+    },
+    {
+      "epoch": 55.0,
+      "grad_norm": 1.2972891330718994,
+      "learning_rate": 9.050000000000001e-06,
+      "loss": 0.0614,
+      "step": 220
+    },
+    {
+      "epoch": 56.0,
+      "grad_norm": 0.8540873527526855,
+      "learning_rate": 8.85e-06,
+      "loss": 0.0606,
+      "step": 224
+    },
+    {
+      "epoch": 57.0,
+      "grad_norm": 0.6654326319694519,
+      "learning_rate": 8.65e-06,
+      "loss": 0.0551,
+      "step": 228
+    },
+    {
+      "epoch": 58.0,
+      "grad_norm": 0.9245683550834656,
+      "learning_rate": 8.45e-06,
+      "loss": 0.054,
+      "step": 232
+    },
+    {
+      "epoch": 59.0,
+      "grad_norm": 0.5625425577163696,
+      "learning_rate": 8.25e-06,
+      "loss": 0.0496,
+      "step": 236
+    },
+    {
+      "epoch": 60.0,
+      "grad_norm": 0.664634644985199,
+      "learning_rate": 8.050000000000001e-06,
+      "loss": 0.0493,
+      "step": 240
+    },
+    {
+      "epoch": 61.0,
+      "grad_norm": 0.5101817846298218,
+      "learning_rate": 7.850000000000001e-06,
+      "loss": 0.0442,
+      "step": 244
+    },
+    {
+      "epoch": 62.0,
+      "grad_norm": 0.5927309393882751,
+      "learning_rate": 7.650000000000001e-06,
+      "loss": 0.0423,
+      "step": 248
+    },
+    {
+      "epoch": 63.0,
+      "grad_norm": 0.7394993305206299,
+      "learning_rate": 7.450000000000001e-06,
+      "loss": 0.0434,
+      "step": 252
+    },
+    {
+      "epoch": 64.0,
+      "grad_norm": 0.653026819229126,
+      "learning_rate": 7.25e-06,
+      "loss": 0.0373,
+      "step": 256
+    },
+    {
+      "epoch": 65.0,
+      "grad_norm": 0.4957493543624878,
+      "learning_rate": 7.05e-06,
+      "loss": 0.0345,
+      "step": 260
+    },
+    {
+      "epoch": 66.0,
+      "grad_norm": 0.6404949426651001,
+      "learning_rate": 6.850000000000001e-06,
+      "loss": 0.0347,
+      "step": 264
+    },
+    {
+      "epoch": 67.0,
+      "grad_norm": 0.4832979440689087,
+      "learning_rate": 6.650000000000001e-06,
+      "loss": 0.0318,
+      "step": 268
+    },
+    {
+      "epoch": 68.0,
+      "grad_norm": 0.5346927046775818,
+      "learning_rate": 6.450000000000001e-06,
+      "loss": 0.0334,
+      "step": 272
+    },
+    {
+      "epoch": 69.0,
+      "grad_norm": 0.46299833059310913,
+      "learning_rate": 6.25e-06,
+      "loss": 0.0329,
+      "step": 276
+    },
+    {
+      "epoch": 70.0,
+      "grad_norm": 0.39228323101997375,
+      "learning_rate": 6.0500000000000005e-06,
+      "loss": 0.0299,
+      "step": 280
+    },
+    {
+      "epoch": 71.0,
+      "grad_norm": 0.4643970727920532,
+      "learning_rate": 5.85e-06,
+      "loss": 0.0297,
+      "step": 284
+    },
+    {
+      "epoch": 72.0,
+      "grad_norm": 0.4702988862991333,
+      "learning_rate": 5.65e-06,
+      "loss": 0.0292,
+      "step": 288
+    },
+    {
+      "epoch": 73.0,
+      "grad_norm": 0.4042636752128601,
+      "learning_rate": 5.450000000000001e-06,
+      "loss": 0.0276,
+      "step": 292
+    },
+    {
+      "epoch": 74.0,
+      "grad_norm": 0.49854159355163574,
+      "learning_rate": 5.2500000000000006e-06,
+      "loss": 0.0285,
+      "step": 296
+    },
+    {
+      "epoch": 75.0,
+      "grad_norm": 0.33747512102127075,
+      "learning_rate": 5.050000000000001e-06,
+      "loss": 0.0259,
+      "step": 300
+    },
+    {
+      "epoch": 76.0,
+      "grad_norm": 0.5222832560539246,
+      "learning_rate": 4.85e-06,
+      "loss": 0.027,
+      "step": 304
+    },
+    {
+      "epoch": 77.0,
+      "grad_norm": 0.3840760588645935,
+      "learning_rate": 4.65e-06,
+      "loss": 0.0257,
+      "step": 308
+    },
+    {
+      "epoch": 78.0,
+      "grad_norm": 0.3676559627056122,
+      "learning_rate": 4.450000000000001e-06,
+      "loss": 0.0253,
+      "step": 312
+    },
+    {
+      "epoch": 79.0,
+      "grad_norm": 0.3206919729709625,
+      "learning_rate": 4.25e-06,
+      "loss": 0.0245,
+      "step": 316
+    },
+    {
+      "epoch": 80.0,
+      "grad_norm": 0.38936108350753784,
+      "learning_rate": 4.05e-06,
+      "loss": 0.0246,
+      "step": 320
+    },
+    {
+      "epoch": 81.0,
+      "grad_norm": 1.3330600261688232,
+      "learning_rate": 3.85e-06,
+      "loss": 0.0245,
+      "step": 324
+    },
+    {
+      "epoch": 82.0,
+      "grad_norm": 0.3317999839782715,
+      "learning_rate": 3.65e-06,
+      "loss": 0.0225,
+      "step": 328
+    },
+    {
+      "epoch": 83.0,
+      "grad_norm": 0.35797789692878723,
+      "learning_rate": 3.45e-06,
+      "loss": 0.0237,
+      "step": 332
+    },
+    {
+      "epoch": 84.0,
+      "grad_norm": 0.3166642189025879,
+      "learning_rate": 3.2500000000000002e-06,
+      "loss": 0.0233,
+      "step": 336
+    },
+    {
+      "epoch": 85.0,
+      "grad_norm": 0.3116203248500824,
+      "learning_rate": 3.05e-06,
+      "loss": 0.0235,
+      "step": 340
+    },
+    {
+      "epoch": 86.0,
+      "grad_norm": 0.3509286940097809,
+      "learning_rate": 2.85e-06,
+      "loss": 0.0221,
+      "step": 344
+    },
+    {
+      "epoch": 87.0,
+      "grad_norm": 0.33957698941230774,
+      "learning_rate": 2.6500000000000005e-06,
+      "loss": 0.0219,
+      "step": 348
+    },
+    {
+      "epoch": 88.0,
+      "grad_norm": 0.36599016189575195,
+      "learning_rate": 2.4500000000000003e-06,
+      "loss": 0.0219,
+      "step": 352
+    },
+    {
+      "epoch": 89.0,
+      "grad_norm": 0.30192670226097107,
+      "learning_rate": 2.25e-06,
+      "loss": 0.0215,
+      "step": 356
+    },
+    {
+      "epoch": 90.0,
+      "grad_norm": 0.4861908257007599,
+      "learning_rate": 2.05e-06,
+      "loss": 0.0216,
+      "step": 360
+    },
+    {
+      "epoch": 91.0,
+      "grad_norm": 0.43383175134658813,
+      "learning_rate": 1.85e-06,
+      "loss": 0.0211,
+      "step": 364
+    },
+    {
+      "epoch": 92.0,
+      "grad_norm": 0.32720497250556946,
+      "learning_rate": 1.6500000000000003e-06,
+      "loss": 0.0218,
+      "step": 368
+    },
+    {
+      "epoch": 93.0,
+      "grad_norm": 0.36105918884277344,
+      "learning_rate": 1.45e-06,
+      "loss": 0.0212,
+      "step": 372
+    },
+    {
+      "epoch": 94.0,
+      "grad_norm": 0.3829093277454376,
+      "learning_rate": 1.25e-06,
+      "loss": 0.0202,
+      "step": 376
+    },
+    {
+      "epoch": 95.0,
+      "grad_norm": 0.3548564016819,
+      "learning_rate": 1.0500000000000001e-06,
+      "loss": 0.0216,
+      "step": 380
+    },
+    {
+      "epoch": 96.0,
+      "grad_norm": 0.52253657579422,
+      "learning_rate": 8.500000000000001e-07,
+      "loss": 0.0211,
+      "step": 384
+    },
+    {
+      "epoch": 97.0,
+      "grad_norm": 0.29113584756851196,
+      "learning_rate": 6.5e-07,
+      "loss": 0.0216,
+      "step": 388
+    },
+    {
+      "epoch": 98.0,
+      "grad_norm": 0.35965240001678467,
+      "learning_rate": 4.5000000000000003e-07,
+      "loss": 0.0209,
+      "step": 392
+    },
+    {
+      "epoch": 99.0,
+      "grad_norm": 0.2798146605491638,
+      "learning_rate": 2.5000000000000004e-07,
+      "loss": 0.0208,
+      "step": 396
+    },
+    {
+      "epoch": 100.0,
+      "grad_norm": 0.30020079016685486,
+      "learning_rate": 5.0000000000000004e-08,
+      "loss": 0.0208,
+      "step": 400
+    }
+  ],
+  "logging_steps": 500,
+  "max_steps": 400,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 100,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 23228751974400.0,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

src/binary_model/checkpoint-400/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f896b886afa5b5445f6670dc5555187301baf60f22ffc418da5b22535c2d9b7
+size 5841

src/binary_model/runs/Nov19_19-55-03_vadim-HP-Laptop-15s-eq1xxx/events.out.tfevents.1763571303.vadim-HP-Laptop-15s-eq1xxx.128002.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53658ae0abd1e72ffd984ba81c8e70ba0b924e0893117cf404b99e33198dea5b
+size 26486

src/category_model/checkpoint-400/config.json

@@ -0,0 +1,39 @@
+{
+  "architectures": [
+    "BertForSequenceClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "dtype": "float32",
+  "emb_size": 312,
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 312,
+  "id2label": {
+    "0": "\u043a\u0443\u043b\u044c\u0442\u0443\u0440\u0430",
+    "1": "\u043f\u043e\u043b\u0438\u0442\u0438\u043a\u0430",
+    "2": "\u0441\u043f\u043e\u0440\u0442",
+    "3": "\u044d\u043a\u043e\u043d\u043e\u043c\u0438\u043a\u0430"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 600,
+  "label2id": {
+    "\u043a\u0443\u043b\u044c\u0442\u0443\u0440\u0430": 0,
+    "\u043f\u043e\u043b\u0438\u0442\u0438\u043a\u0430": 1,
+    "\u0441\u043f\u043e\u0440\u0442": 2,
+    "\u044d\u043a\u043e\u043d\u043e\u043c\u0438\u043a\u0430": 3
+  },
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 3,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "problem_type": "single_label_classification",
+  "transformers_version": "4.57.0",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 29564
+}

src/category_model/checkpoint-400/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d171659daf0b9242a1355e1f748db5f1fcb61a51537acafa86dd40ee74ed82f
+size 47148128

src/category_model/checkpoint-400/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87a1c7b428390f59facee41bc1170aabd0e0fcd483460b5361b4b834c02b1f88
+size 94328139

src/category_model/checkpoint-400/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80f4b2228e39d06e9189b284c4d5fbf902b5c3de450a889d8a8ff7c84c225c15
+size 14455

src/category_model/checkpoint-400/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d2a13aedfb10730f658c90eb03b45908f32db49971da0e67d91a64b15a963525
+size 1465

src/category_model/checkpoint-400/trainer_state.json

@@ -0,0 +1,734 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 100.0,
+  "eval_steps": 500,
+  "global_step": 400,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 1.0,
+      "grad_norm": 3.455427408218384,
+      "learning_rate": 1.985e-05,
+      "loss": 1.372,
+      "step": 4
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 2.609254837036133,
+      "learning_rate": 1.9650000000000003e-05,
+      "loss": 1.3635,
+      "step": 8
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 3.7442009449005127,
+      "learning_rate": 1.9450000000000002e-05,
+      "loss": 1.3225,
+      "step": 12
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 3.725454568862915,
+      "learning_rate": 1.925e-05,
+      "loss": 1.3125,
+      "step": 16
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 3.681874990463257,
+      "learning_rate": 1.9050000000000002e-05,
+      "loss": 1.2952,
+      "step": 20
+    },
+    {
+      "epoch": 6.0,
+      "grad_norm": 4.666588306427002,
+      "learning_rate": 1.885e-05,
+      "loss": 1.3249,
+      "step": 24
+    },
+    {
+      "epoch": 7.0,
+      "grad_norm": 2.378511428833008,
+      "learning_rate": 1.8650000000000003e-05,
+      "loss": 1.28,
+      "step": 28
+    },
+    {
+      "epoch": 8.0,
+      "grad_norm": 4.4941725730896,
+      "learning_rate": 1.845e-05,
+      "loss": 1.2995,
+      "step": 32
+    },
+    {
+      "epoch": 9.0,
+      "grad_norm": 3.8662474155426025,
+      "learning_rate": 1.825e-05,
+      "loss": 1.2559,
+      "step": 36
+    },
+    {
+      "epoch": 10.0,
+      "grad_norm": 3.9468913078308105,
+      "learning_rate": 1.805e-05,
+      "loss": 1.2227,
+      "step": 40
+    },
+    {
+      "epoch": 11.0,
+      "grad_norm": 3.712695360183716,
+      "learning_rate": 1.785e-05,
+      "loss": 1.2797,
+      "step": 44
+    },
+    {
+      "epoch": 12.0,
+      "grad_norm": 2.5063178539276123,
+      "learning_rate": 1.7650000000000002e-05,
+      "loss": 1.232,
+      "step": 48
+    },
+    {
+      "epoch": 13.0,
+      "grad_norm": 5.289675712585449,
+      "learning_rate": 1.7450000000000004e-05,
+      "loss": 1.2657,
+      "step": 52
+    },
+    {
+      "epoch": 14.0,
+      "grad_norm": 3.8385415077209473,
+      "learning_rate": 1.7250000000000003e-05,
+      "loss": 1.179,
+      "step": 56
+    },
+    {
+      "epoch": 15.0,
+      "grad_norm": 4.603259086608887,
+      "learning_rate": 1.705e-05,
+      "loss": 1.2584,
+      "step": 60
+    },
+    {
+      "epoch": 16.0,
+      "grad_norm": 3.869927406311035,
+      "learning_rate": 1.6850000000000003e-05,
+      "loss": 1.1632,
+      "step": 64
+    },
+    {
+      "epoch": 17.0,
+      "grad_norm": 3.1474037170410156,
+      "learning_rate": 1.665e-05,
+      "loss": 1.1961,
+      "step": 68
+    },
+    {
+      "epoch": 18.0,
+      "grad_norm": 5.3139424324035645,
+      "learning_rate": 1.645e-05,
+      "loss": 1.1511,
+      "step": 72
+    },
+    {
+      "epoch": 19.0,
+      "grad_norm": 7.552919387817383,
+      "learning_rate": 1.6250000000000002e-05,
+      "loss": 1.1045,
+      "step": 76
+    },
+    {
+      "epoch": 20.0,
+      "grad_norm": 3.9046547412872314,
+      "learning_rate": 1.605e-05,
+      "loss": 1.1123,
+      "step": 80
+    },
+    {
+      "epoch": 21.0,
+      "grad_norm": 4.048532962799072,
+      "learning_rate": 1.5850000000000002e-05,
+      "loss": 1.0915,
+      "step": 84
+    },
+    {
+      "epoch": 22.0,
+      "grad_norm": 4.49423885345459,
+      "learning_rate": 1.565e-05,
+      "loss": 1.0988,
+      "step": 88
+    },
+    {
+      "epoch": 23.0,
+      "grad_norm": 7.544517517089844,
+      "learning_rate": 1.545e-05,
+      "loss": 1.1344,
+      "step": 92
+    },
+    {
+      "epoch": 24.0,
+      "grad_norm": 5.467407703399658,
+      "learning_rate": 1.525e-05,
+      "loss": 1.1133,
+      "step": 96
+    },
+    {
+      "epoch": 25.0,
+      "grad_norm": 3.8020665645599365,
+      "learning_rate": 1.505e-05,
+      "loss": 1.0526,
+      "step": 100
+    },
+    {
+      "epoch": 26.0,
+      "grad_norm": 4.590837478637695,
+      "learning_rate": 1.4850000000000002e-05,
+      "loss": 1.0252,
+      "step": 104
+    },
+    {
+      "epoch": 27.0,
+      "grad_norm": 5.339500427246094,
+      "learning_rate": 1.4650000000000002e-05,
+      "loss": 0.9814,
+      "step": 108
+    },
+    {
+      "epoch": 28.0,
+      "grad_norm": 3.674020528793335,
+      "learning_rate": 1.4450000000000002e-05,
+      "loss": 0.9983,
+      "step": 112
+    },
+    {
+      "epoch": 29.0,
+      "grad_norm": 7.431573867797852,
+      "learning_rate": 1.425e-05,
+      "loss": 0.9252,
+      "step": 116
+    },
+    {
+      "epoch": 30.0,
+      "grad_norm": 6.031501770019531,
+      "learning_rate": 1.4050000000000001e-05,
+      "loss": 1.0251,
+      "step": 120
+    },
+    {
+      "epoch": 31.0,
+      "grad_norm": 6.9349799156188965,
+      "learning_rate": 1.3850000000000001e-05,
+      "loss": 0.8766,
+      "step": 124
+    },
+    {
+      "epoch": 32.0,
+      "grad_norm": 3.9678494930267334,
+      "learning_rate": 1.3650000000000001e-05,
+      "loss": 0.9679,
+      "step": 128
+    },
+    {
+      "epoch": 33.0,
+      "grad_norm": 7.587587356567383,
+      "learning_rate": 1.3450000000000002e-05,
+      "loss": 0.9791,
+      "step": 132
+    },
+    {
+      "epoch": 34.0,
+      "grad_norm": 6.44096040725708,
+      "learning_rate": 1.325e-05,
+      "loss": 0.8987,
+      "step": 136
+    },
+    {
+      "epoch": 35.0,
+      "grad_norm": 5.321375846862793,
+      "learning_rate": 1.305e-05,
+      "loss": 0.8359,
+      "step": 140
+    },
+    {
+      "epoch": 36.0,
+      "grad_norm": 4.376260757446289,
+      "learning_rate": 1.285e-05,
+      "loss": 0.8474,
+      "step": 144
+    },
+    {
+      "epoch": 37.0,
+      "grad_norm": 5.06814432144165,
+      "learning_rate": 1.2650000000000001e-05,
+      "loss": 0.8176,
+      "step": 148
+    },
+    {
+      "epoch": 38.0,
+      "grad_norm": 4.7853899002075195,
+      "learning_rate": 1.2450000000000003e-05,
+      "loss": 0.8357,
+      "step": 152
+    },
+    {
+      "epoch": 39.0,
+      "grad_norm": 3.8893511295318604,
+      "learning_rate": 1.2250000000000001e-05,
+      "loss": 0.7759,
+      "step": 156
+    },
+    {
+      "epoch": 40.0,
+      "grad_norm": 4.117180824279785,
+      "learning_rate": 1.2050000000000002e-05,
+      "loss": 0.776,
+      "step": 160
+    },
+    {
+      "epoch": 41.0,
+      "grad_norm": 3.4978015422821045,
+      "learning_rate": 1.1850000000000002e-05,
+      "loss": 0.8084,
+      "step": 164
+    },
+    {
+      "epoch": 42.0,
+      "grad_norm": 3.2947819232940674,
+      "learning_rate": 1.1650000000000002e-05,
+      "loss": 0.7224,
+      "step": 168
+    },
+    {
+      "epoch": 43.0,
+      "grad_norm": 7.838773250579834,
+      "learning_rate": 1.145e-05,
+      "loss": 0.7542,
+      "step": 172
+    },
+    {
+      "epoch": 44.0,
+      "grad_norm": 5.788562297821045,
+      "learning_rate": 1.125e-05,
+      "loss": 0.733,
+      "step": 176
+    },
+    {
+      "epoch": 45.0,
+      "grad_norm": 6.079436779022217,
+      "learning_rate": 1.1050000000000001e-05,
+      "loss": 0.7207,
+      "step": 180
+    },
+    {
+      "epoch": 46.0,
+      "grad_norm": 5.334987640380859,
+      "learning_rate": 1.0850000000000001e-05,
+      "loss": 0.6308,
+      "step": 184
+    },
+    {
+      "epoch": 47.0,
+      "grad_norm": 3.619874954223633,
+      "learning_rate": 1.065e-05,
+      "loss": 0.6654,
+      "step": 188
+    },
+    {
+      "epoch": 48.0,
+      "grad_norm": 4.317775726318359,
+      "learning_rate": 1.045e-05,
+      "loss": 0.5905,
+      "step": 192
+    },
+    {
+      "epoch": 49.0,
+      "grad_norm": 4.5337748527526855,
+      "learning_rate": 1.025e-05,
+      "loss": 0.6396,
+      "step": 196
+    },
+    {
+      "epoch": 50.0,
+      "grad_norm": 3.27056884765625,
+      "learning_rate": 1.005e-05,
+      "loss": 0.5575,
+      "step": 200
+    },
+    {
+      "epoch": 51.0,
+      "grad_norm": 5.867713928222656,
+      "learning_rate": 9.85e-06,
+      "loss": 0.5961,
+      "step": 204
+    },
+    {
+      "epoch": 52.0,
+      "grad_norm": 4.240111351013184,
+      "learning_rate": 9.65e-06,
+      "loss": 0.5803,
+      "step": 208
+    },
+    {
+      "epoch": 53.0,
+      "grad_norm": 8.360318183898926,
+      "learning_rate": 9.450000000000001e-06,
+      "loss": 0.6097,
+      "step": 212
+    },
+    {
+      "epoch": 54.0,
+      "grad_norm": 5.395203590393066,
+      "learning_rate": 9.250000000000001e-06,
+      "loss": 0.5235,
+      "step": 216
+    },
+    {
+      "epoch": 55.0,
+      "grad_norm": 8.306116104125977,
+      "learning_rate": 9.050000000000001e-06,
+      "loss": 0.6212,
+      "step": 220
+    },
+    {
+      "epoch": 56.0,
+      "grad_norm": 4.548465728759766,
+      "learning_rate": 8.85e-06,
+      "loss": 0.4939,
+      "step": 224
+    },
+    {
+      "epoch": 57.0,
+      "grad_norm": 5.0567755699157715,
+      "learning_rate": 8.65e-06,
+      "loss": 0.492,
+      "step": 228
+    },
+    {
+      "epoch": 58.0,
+      "grad_norm": 3.3125669956207275,
+      "learning_rate": 8.45e-06,
+      "loss": 0.4867,
+      "step": 232
+    },
+    {
+      "epoch": 59.0,
+      "grad_norm": 9.607614517211914,
+      "learning_rate": 8.25e-06,
+      "loss": 0.4979,
+      "step": 236
+    },
+    {
+      "epoch": 60.0,
+      "grad_norm": 4.669170379638672,
+      "learning_rate": 8.050000000000001e-06,
+      "loss": 0.5232,
+      "step": 240
+    },
+    {
+      "epoch": 61.0,
+      "grad_norm": 3.661278247833252,
+      "learning_rate": 7.850000000000001e-06,
+      "loss": 0.4184,
+      "step": 244
+    },
+    {
+      "epoch": 62.0,
+      "grad_norm": 6.294672012329102,
+      "learning_rate": 7.650000000000001e-06,
+      "loss": 0.4472,
+      "step": 248
+    },
+    {
+      "epoch": 63.0,
+      "grad_norm": 3.544436454772949,
+      "learning_rate": 7.450000000000001e-06,
+      "loss": 0.449,
+      "step": 252
+    },
+    {
+      "epoch": 64.0,
+      "grad_norm": 5.267669200897217,
+      "learning_rate": 7.25e-06,
+      "loss": 0.4789,
+      "step": 256
+    },
+    {
+      "epoch": 65.0,
+      "grad_norm": 7.3072333335876465,
+      "learning_rate": 7.05e-06,
+      "loss": 0.4661,
+      "step": 260
+    },
+    {
+      "epoch": 66.0,
+      "grad_norm": 2.6512272357940674,
+      "learning_rate": 6.850000000000001e-06,
+      "loss": 0.3777,
+      "step": 264
+    },
+    {
+      "epoch": 67.0,
+      "grad_norm": 4.13808536529541,
+      "learning_rate": 6.650000000000001e-06,
+      "loss": 0.4238,
+      "step": 268
+    },
+    {
+      "epoch": 68.0,
+      "grad_norm": 3.1775310039520264,
+      "learning_rate": 6.450000000000001e-06,
+      "loss": 0.3932,
+      "step": 272
+    },
+    {
+      "epoch": 69.0,
+      "grad_norm": 3.4776253700256348,
+      "learning_rate": 6.25e-06,
+      "loss": 0.3601,
+      "step": 276
+    },
+    {
+      "epoch": 70.0,
+      "grad_norm": 4.582927227020264,
+      "learning_rate": 6.0500000000000005e-06,
+      "loss": 0.4413,
+      "step": 280
+    },
+    {
+      "epoch": 71.0,
+      "grad_norm": 2.587031364440918,
+      "learning_rate": 5.85e-06,
+      "loss": 0.3916,
+      "step": 284
+    },
+    {
+      "epoch": 72.0,
+      "grad_norm": 3.7085821628570557,
+      "learning_rate": 5.65e-06,
+      "loss": 0.4055,
+      "step": 288
+    },
+    {
+      "epoch": 73.0,
+      "grad_norm": 5.436678886413574,
+      "learning_rate": 5.450000000000001e-06,
+      "loss": 0.3487,
+      "step": 292
+    },
+    {
+      "epoch": 74.0,
+      "grad_norm": 5.039726734161377,
+      "learning_rate": 5.2500000000000006e-06,
+      "loss": 0.3582,
+      "step": 296
+    },
+    {
+      "epoch": 75.0,
+      "grad_norm": 4.922318935394287,
+      "learning_rate": 5.050000000000001e-06,
+      "loss": 0.3563,
+      "step": 300
+    },
+    {
+      "epoch": 76.0,
+      "grad_norm": 4.511425971984863,
+      "learning_rate": 4.85e-06,
+      "loss": 0.3747,
+      "step": 304
+    },
+    {
+      "epoch": 77.0,
+      "grad_norm": 2.0960898399353027,
+      "learning_rate": 4.65e-06,
+      "loss": 0.3277,
+      "step": 308
+    },
+    {
+      "epoch": 78.0,
+      "grad_norm": 1.7806938886642456,
+      "learning_rate": 4.450000000000001e-06,
+      "loss": 0.317,
+      "step": 312
+    },
+    {
+      "epoch": 79.0,
+      "grad_norm": 3.6240742206573486,
+      "learning_rate": 4.25e-06,
+      "loss": 0.3506,
+      "step": 316
+    },
+    {
+      "epoch": 80.0,
+      "grad_norm": 3.0891218185424805,
+      "learning_rate": 4.05e-06,
+      "loss": 0.3342,
+      "step": 320
+    },
+    {
+      "epoch": 81.0,
+      "grad_norm": 3.1899912357330322,
+      "learning_rate": 3.85e-06,
+      "loss": 0.3692,
+      "step": 324
+    },
+    {
+      "epoch": 82.0,
+      "grad_norm": 1.9796233177185059,
+      "learning_rate": 3.65e-06,
+      "loss": 0.3087,
+      "step": 328
+    },
+    {
+      "epoch": 83.0,
+      "grad_norm": 4.603359222412109,
+      "learning_rate": 3.45e-06,
+      "loss": 0.3533,
+      "step": 332
+    },
+    {
+      "epoch": 84.0,
+      "grad_norm": 5.730408668518066,
+      "learning_rate": 3.2500000000000002e-06,
+      "loss": 0.3498,
+      "step": 336
+    },
+    {
+      "epoch": 85.0,
+      "grad_norm": 6.595205783843994,
+      "learning_rate": 3.05e-06,
+      "loss": 0.3618,
+      "step": 340
+    },
+    {
+      "epoch": 86.0,
+      "grad_norm": 11.516875267028809,
+      "learning_rate": 2.85e-06,
+      "loss": 0.3932,
+      "step": 344
+    },
+    {
+      "epoch": 87.0,
+      "grad_norm": 3.7310776710510254,
+      "learning_rate": 2.6500000000000005e-06,
+      "loss": 0.329,
+      "step": 348
+    },
+    {
+      "epoch": 88.0,
+      "grad_norm": 2.2054193019866943,
+      "learning_rate": 2.4500000000000003e-06,
+      "loss": 0.3097,
+      "step": 352
+    },
+    {
+      "epoch": 89.0,
+      "grad_norm": 2.450695037841797,
+      "learning_rate": 2.25e-06,
+      "loss": 0.3052,
+      "step": 356
+    },
+    {
+      "epoch": 90.0,
+      "grad_norm": 2.2963459491729736,
+      "learning_rate": 2.05e-06,
+      "loss": 0.3126,
+      "step": 360
+    },
+    {
+      "epoch": 91.0,
+      "grad_norm": 3.7548775672912598,
+      "learning_rate": 1.85e-06,
+      "loss": 0.3343,
+      "step": 364
+    },
+    {
+      "epoch": 92.0,
+      "grad_norm": 1.9919285774230957,
+      "learning_rate": 1.6500000000000003e-06,
+      "loss": 0.2815,
+      "step": 368
+    },
+    {
+      "epoch": 93.0,
+      "grad_norm": 3.4772584438323975,
+      "learning_rate": 1.45e-06,
+      "loss": 0.3316,
+      "step": 372
+    },
+    {
+      "epoch": 94.0,
+      "grad_norm": 2.701188564300537,
+      "learning_rate": 1.25e-06,
+      "loss": 0.3175,
+      "step": 376
+    },
+    {
+      "epoch": 95.0,
+      "grad_norm": 2.582921266555786,
+      "learning_rate": 1.0500000000000001e-06,
+      "loss": 0.2869,
+      "step": 380
+    },
+    {
+      "epoch": 96.0,
+      "grad_norm": 3.1191177368164062,
+      "learning_rate": 8.500000000000001e-07,
+      "loss": 0.3155,
+      "step": 384
+    },
+    {
+      "epoch": 97.0,
+      "grad_norm": 4.482059478759766,
+      "learning_rate": 6.5e-07,
+      "loss": 0.3525,
+      "step": 388
+    },
+    {
+      "epoch": 98.0,
+      "grad_norm": 3.904967784881592,
+      "learning_rate": 4.5000000000000003e-07,
+      "loss": 0.3339,
+      "step": 392
+    },
+    {
+      "epoch": 99.0,
+      "grad_norm": 5.207050323486328,
+      "learning_rate": 2.5000000000000004e-07,
+      "loss": 0.3153,
+      "step": 396
+    },
+    {
+      "epoch": 100.0,
+      "grad_norm": 4.7919745445251465,
+      "learning_rate": 5.0000000000000004e-08,
+      "loss": 0.3194,
+      "step": 400
+    }
+  ],
+  "logging_steps": 500,
+  "max_steps": 400,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 100,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 18809131622400.0,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

src/category_model/checkpoint-400/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1c578e2b795ff2107b64fd5f8c207b5966e7b63550fcf63db2897f2e5d55fcc
+size 5841

src/category_model/runs/Nov19_16-34-43_vadim-HP-Laptop-15s-eq1xxx/events.out.tfevents.1763559283.vadim-HP-Laptop-15s-eq1xxx.119293.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5883fd5d4a434aa097bd4f10219658cde4fb327f41063da35f097c27473aede7
+size 12093

src/category_model/runs/Nov19_16-46-52_vadim-HP-Laptop-15s-eq1xxx/events.out.tfevents.1763560013.vadim-HP-Laptop-15s-eq1xxx.120060.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b68fcd71c4edc7c64c33b783e6abaf286da860db632393842f9229d750f5079
+size 26866

src/ml_binary.joblib

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cbda2dff7c7d0f5b0e98a62c7ba63df2000e571d84347a232bbc98d2233c6fa4
+size 2651188

src/ml_category.joblib

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5dda5f9be7a82fcce0467fb6cd222ddf0dc69edf4bfaa25910ec7285463fb3a
+size 3561992

src/ml_categorys.joblib

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cfee13464c870a68c591c2b5005b24530d4fb5ee538671e91d05ab73d5f61d1
+size 3954215

src/multilabel_model/checkpoint-700/config.json

@@ -0,0 +1,39 @@
+{
+  "architectures": [
+    "BertForSequenceClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "dtype": "float32",
+  "emb_size": 312,
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 312,
+  "id2label": {
+    "0": "LABEL_0",
+    "1": "LABEL_1",
+    "2": "LABEL_2",
+    "3": "LABEL_3"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 600,
+  "label2id": {
+    "LABEL_0": 0,
+    "LABEL_1": 1,
+    "LABEL_2": 2,
+    "LABEL_3": 3
+  },
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 3,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "problem_type": "multi_label_classification",
+  "transformers_version": "4.57.0",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 29564
+}

src/multilabel_model/checkpoint-700/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cc2d851cd93e4c007e0b237673fb86cd555754dca253a379ec239b7acaea1df
+size 47148128

src/multilabel_model/checkpoint-700/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1e70436bc9452a2ffdf997eb531c55061f9aab4883a53385ee85478431dfeacb
+size 94328139

src/multilabel_model/checkpoint-700/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22564cefda826550225bd0fd67a748fd7cb1eeafc2de68c4b5a5600037865e5c
+size 14455

src/multilabel_model/checkpoint-700/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a7430e3783a2d51c132011d61c22c31c8fec694606fd4e0d6f02426d5c36b37
+size 1465

src/multilabel_model/checkpoint-700/trainer_state.json

@@ -0,0 +1,41 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 100.0,
+  "eval_steps": 500,
+  "global_step": 700,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 71.42857142857143,
+      "grad_norm": 0.5301845669746399,
+      "learning_rate": 0.0008614285714285715,
+      "loss": 0.6088,
+      "step": 500
+    }
+  ],
+  "logging_steps": 500,
+  "max_steps": 700,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 100,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 18440325120000.0,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

src/multilabel_model/checkpoint-700/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c3a9d8bd2f33a889f206bdf109d0a17adc92a638dcd4560c2c2ee7c25e5b5dd
+size 5841

src/multilabel_model/runs/Nov19_18-32-50_vadim-HP-Laptop-15s-eq1xxx/events.out.tfevents.1763566371.vadim-HP-Laptop-15s-eq1xxx.124852.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d2783d509460017f1b7d648662ea982a1f5396f2e7c00f4c11dbb544a735cc2
+size 5282

src/multilabel_model/runs/Nov19_18-43-43_vadim-HP-Laptop-15s-eq1xxx/events.out.tfevents.1763567023.vadim-HP-Laptop-15s-eq1xxx.125134.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dafabd32f21f5da3b60a5c0761ef6534d2056e706f11a059954470db19eafe02
+size 5282

src/multilabel_model/runs/Nov19_18-50-20_vadim-HP-Laptop-15s-eq1xxx/events.out.tfevents.1763567421.vadim-HP-Laptop-15s-eq1xxx.125341.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:957d4a549ffd58187f71c12287765ee092001e9ab79184c982bed40b6f04684d
+size 5282

src/multilabel_model/runs/Nov19_18-56-03_vadim-HP-Laptop-15s-eq1xxx/events.out.tfevents.1763567764.vadim-HP-Laptop-15s-eq1xxx.125471.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d994adcc370d91dba8a168cd68b7187a71c43389d7a23a0374fe853cfb2fcae7
+size 5282

src/multilabel_model/runs/Nov19_19-12-54_vadim-HP-Laptop-15s-eq1xxx/events.out.tfevents.1763568775.vadim-HP-Laptop-15s-eq1xxx.125830.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ef3f43e0de37cd4c14f85dbec9e4c9e366fb8843d1c3dafb4f10eb0b40bd512
+size 5282

src/multilabel_model/runs/Nov19_19-19-23_vadim-HP-Laptop-15s-eq1xxx/events.out.tfevents.1763569164.vadim-HP-Laptop-15s-eq1xxx.126011.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a2c9137dc4ed0b364ae44e5d8b5fd57e8e6b253a6b90b8ce64bddc29d7c4ee5
+size 4184

src/multilabel_model/runs/Nov19_19-20-09_vadim-HP-Laptop-15s-eq1xxx/events.out.tfevents.1763569210.vadim-HP-Laptop-15s-eq1xxx.126088.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88ccd53331d477ce2088f7a0647754563f0b733b0ae7546ad04cfc3971b870f8
+size 5847

src/nn_binary.keras

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56eaba127bef8116f6d3d52e7243dd644af3fb91e9fab9c492d0474b980d1452
+size 78129542

src/nn_category.keras

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:994e7f52d66de06e363fde235a31688b8eccbd106485d80b98683479fa2e6124
+size 38731315

src/nn_categorys.keras

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1c11cc21b869caba3002bcdeaad3763b51491aa320fbffb24cfb9505562c13e
+size 78239148

src/nn_vectorizer_binary.keras

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6632ffa21bd1c3acd6835ff4b1c09e8d87db21046e7fe548424f06ab972b37d3
+size 367823

src/nn_vectorizer_category.keras

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b82e82c137dd417b3d36b936218147bf1ac1b100ad74cb32810a86ec86ab960b
+size 391149

src/nn_vectorizer_categorys.keras

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d6a999f2756bed0a9b7febd2e41d564578a2417ee07fbf30f9c24521dcf8941
+size 402881

src/streamlit_app.py

@@ -1,40 +1,555 @@
-import altair as alt
-import numpy as np
-import pandas as pd
 import streamlit as st
+import pandas as pd
+import numpy as np
+import matplotlib.pyplot as plt
+import seaborn as sns
+from sklearn.metrics import classification_report, confusion_matrix, roc_curve, auc, precision_recall_curve, precision_score, recall_score, f1_score
+import json
+
+# Импортируем ваши модули
+try:
+    from use_ml import predict_sentiment, predict_category, predict_categorys
+except ImportError:
+    st.error("Модуль use_ml не найден")
+try:
+    from use_nn import predict_sentiment as nn_predict_sentiment
+    from use_nn import predict_category as nn_predict_category
+    from use_nn import predict_categorys as nn_predict_categorys
+except ImportError:
+    st.error("Модуль use_nn не найден")
+try:
+    from use_transformer import predict_sentiment as tf_predict_sentiment
+    from use_transformer import predict_category as tf_predict_category
+    from use_transformer import predict_categorys as tf_predict_categorys
+except ImportError:
+    st.error("Модуль use_transformer не найден")
+
+# Настройка страницы
+st.set_page_config(
+    page_title="Анализ классификаторов текста",
+    page_icon="📊",
+    layout="wide"
+)
+
+
+def load_models(task_type):
+    """Загрузка моделей в зависимости от типа задачи"""
+    models = {}
+
+    if task_type == "Бинарная":
+        try:
+            models["Классическая ML"] = predict_sentiment()
+        except:
+            pass
+        try:
+            models["Нейросеть"] = nn_predict_sentiment()
+        except:
+            pass
+        try:
+            models["Трансформер"] = tf_predict_sentiment()
+        except:
+            pass
+
+    elif task_type == "Многоклассовая":
+        try:
+            models["Классическая ML"] = predict_category()
+        except:
+            pass
+        try:
+            models["Нейросеть"] = nn_predict_category()
+        except:
+            pass
+        try:
+            models["Трансформер"] = tf_predict_category()
+        except:
+            pass
+
+    elif task_type == "Многометочная":
+        try:
+            models["Классическая ML"] = predict_categorys()
+        except:
+            pass
+        try:
+            models["Нейросеть"] = nn_predict_categorys()
+        except:
+            pass
+        try:
+            models["Трансформер"] = tf_predict_categorys()
+        except:
+            pass
+
+    return models
+
+
+def plot_probabilities(probs, labels, model_name):
+    """Визуализация вероятностей"""
+    fig, ax = plt.subplots(figsize=(10, 6))
+    y_pos = np.arange(len(labels))
+
+    if isinstance(probs, (np.ndarray, list)) and len(probs) > 1:
+        # Многоклассовая или многометочная
+        ax.barh(y_pos, probs, align='center')
+        ax.set_yticks(y_pos)
+        ax.set_yticklabels(labels)
+        ax.set_xlabel('Вероятность')
+        ax.set_title(f'Вероятности классов - {model_name}')
+    else:
+        # Бинарная
+        binary_probs = [1 - probs, probs] if isinstance(probs, (int, float)) else [1 - probs[0], probs[0]]
+        binary_labels = ['Negative', 'Positive']
+        ax.barh([0, 1], binary_probs, align='center')
+        ax.set_yticks([0, 1])
+        ax.set_yticklabels(binary_labels)
+        ax.set_xlabel('Вероятность')
+        ax.set_title(f'Вероятности классов - {model_name}')
+
+    plt.tight_layout()
+    return fig
+
+
+def calculate_and_display_binary_metrics(true_labels, predictions):
+    """Расчет и отображение метрик для бинарной классификации"""
+    # Преобразуем true_labels в числовой формат
+    y_true = [1 if label == 'positive' else 0 for label in true_labels]
+    y_pred = [1 if pred['probs'] >= 0.5 else 0 for pred in predictions]
+    y_scores = [pred['probs'] for pred in predictions]
+
+    # ROC curve
+    fpr, tpr, _ = roc_curve(y_true, y_scores)
+    roc_auc = auc(fpr, tpr)
+
+    # Precision-Recall curve
+    precision, recall, _ = precision_recall_curve(y_true, y_scores)
+
+    # Визуализация
+    fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(15, 6))
+
+    # ROC curve
+    ax1.plot(fpr, tpr, color='darkorange', lw=2, label=f'ROC curve (AUC = {roc_auc:.2f})')
+    ax1.plot([0, 1], [0, 1], color='navy', lw=2, linestyle='--')
+    ax1.set_xlim([0.0, 1.0])
+    ax1.set_ylim([0.0, 1.05])
+    ax1.set_xlabel('False Positive Rate')
+    ax1.set_ylabel('True Positive Rate')
+    ax1.set_title('ROC Curve')
+    ax1.legend(loc="lower right")
+
+    # Precision-Recall curve
+    ax2.plot(recall, precision, color='blue', lw=2)
+    ax2.set_xlim([0.0, 1.0])
+    ax2.set_ylim([0.0, 1.05])
+    ax2.set_xlabel('Recall')
+    ax2.set_ylabel('Precision')
+    ax2.set_title('Precision-Recall Curve')
+
+    st.pyplot(fig)
+
+    # Матрица ошибок
+    cm = confusion_matrix(y_true, y_pred)
+    fig, ax = plt.subplots(figsize=(8, 6))
+    sns.heatmap(cm, annot=True, fmt='d', cmap='Blues', ax=ax,
+                xticklabels=['Negative', 'Positive'],
+                yticklabels=['Negative', 'Positive'])
+    ax.set_title('Confusion Matrix')
+    ax.set_xlabel('Predicted')
+    ax.set_ylabel('Actual')
+    st.pyplot(fig)
+
+    # Отчет классификации
+    st.subheader("Отчет классификации")
+    report = classification_report(y_true, y_pred, output_dict=True)
+    report_df = pd.DataFrame(report).transpose()
+    st.dataframe(report_df, use_container_width=True)
+
+
+def calculate_and_display_multiclass_metrics(true_labels, predictions):
+    """Расчет и отображение метрик для многоклассовой классификации"""
+    # Получаем все уникальные классы
+    all_classes = list(set(true_labels))
+
+    # Предсказанные классы (класс с максимальной вероятностью)
+    y_pred = [pred['labels'][np.argmax(pred['probs'])] for pred in predictions]
+    y_true = true_labels
+
+    # Матрица ошибок
+    cm = confusion_matrix(y_true, y_pred, labels=all_classes)
+    fig, ax = plt.subplots(figsize=(10, 8))
+    sns.heatmap(cm, annot=True, fmt='d', cmap='Blues', ax=ax,
+                xticklabels=all_classes, yticklabels=all_classes)
+    ax.set_title('Confusion Matrix')
+    ax.set_xlabel('Predicted')
+    ax.set_ylabel('Actual')
+    plt.xticks(rotation=45)
+    plt.yticks(rotation=0)
+    st.pyplot(fig)
+
+    # Отчет классификации
+    st.subheader("Отчет классификации")
+    report = classification_report(y_true, y_pred, output_dict=True)
+    report_df = pd.DataFrame(report).transpose()
+    st.dataframe(report_df, use_container_width=True)
+
+    # Визуализация точности по классам
+    class_report = classification_report(y_true, y_pred, output_dict=True)
+    classes_metrics = {}
+    for class_name in all_classes:
+        if class_name in class_report:
+            classes_metrics[class_name] = {
+                'Precision': class_report[class_name]['precision'],
+                'Recall': class_report[class_name]['recall'],
+                'F1-Score': class_report[class_name]['f1-score']
+            }
+
+    metrics_df = pd.DataFrame(classes_metrics).T
+    fig, ax = plt.subplots(figsize=(12, 6))
+    metrics_df.plot(kind='bar', ax=ax)
+    ax.set_title('Метрики по классам')
+    ax.set_ylabel('Score')
+    ax.legend(bbox_to_anchor=(1.05, 1), loc='upper left')
+    plt.xticks(rotation=45)
+    st.pyplot(fig)
+
+
+def calculate_and_display_multilabel_metrics(true_labels, predictions):
+    """Расчет и отображение метрик для многометочной классификации"""
+    # Получаем все возможные метки из предсказаний
+    all_labels = predictions[0]['labels']
+
+    # Создаем бинарные матрицы для истинных и предсказанных меток
+    y_true_binary = np.zeros((len(true_labels), len(all_labels)))
+    y_pred_binary = np.zeros((len(predictions), len(all_labels)))
+
+    for i, (true_label_list, pred) in enumerate(zip(true_labels, predictions)):
+        for j, label in enumerate(all_labels):
+            # Истинные метки
+            if label in true_label_list:
+                y_true_binary[i, j] = 1
+
+            # Предсказанные метки (порог 0.5)
+            if pred['probs'][j] >= 0.5:
+                y_pred_binary[i, j] = 1
+
+    # Вычисляем метрики для каждой метки
+    metrics_per_label = {}
+    for j, label in enumerate(all_labels):
+        metrics_per_label[label] = {
+            'Precision': precision_score(y_true_binary[:, j], y_pred_binary[:, j]),
+            'Recall': recall_score(y_true_binary[:, j], y_pred_binary[:, j]),
+            'F1-Score': f1_score(y_true_binary[:, j], y_pred_binary[:, j]),
+            'Support': np.sum(y_true_binary[:, j])
+        }
+
+    # Сводная таблица метрик
+    st.subheader("Метрики по меткам")
+    metrics_df = pd.DataFrame(metrics_per_label).T
+    st.dataframe(metrics_df, use_container_width=True)
+
+    # Визуализация метрик
+    fig, axes = plt.subplots(2, 2, figsize=(15, 10))
+
+    # Precision по меткам
+    axes[0, 0].barh(range(len(all_labels)), [metrics_per_label[label]['Precision'] for label in all_labels])
+    axes[0, 0].set_yticks(range(len(all_labels)))
+    axes[0, 0].set_yticklabels(all_labels)
+    axes[0, 0].set_title('Precision по меткам')
+    axes[0, 0].set_xlim(0, 1)
+
+    # Recall по меткам
+    axes[0, 1].barh(range(len(all_labels)), [metrics_per_label[label]['Recall'] for label in all_labels])
+    axes[0, 1].set_yticks(range(len(all_labels)))
+    axes[0, 1].set_yticklabels(all_labels)
+    axes[0, 1].set_title('Recall по меткам')
+    axes[0, 1].set_xlim(0, 1)
+
+    # F1-Score по меткам
+    axes[1, 0].barh(range(len(all_labels)), [metrics_per_label[label]['F1-Score'] for label in all_labels])
+    axes[1, 0].set_yticks(range(len(all_labels)))
+    axes[1, 0].set_yticklabels(all_labels)
+    axes[1, 0].set_title('F1-Score по меткам')
+    axes[1, 0].set_xlim(0, 1)
+
+    # Support по меткам
+    axes[1, 1].barh(range(len(all_labels)), [metrics_per_label[label]['Support'] for label in all_labels])
+    axes[1, 1].set_yticks(range(len(all_labels)))
+    axes[1, 1].set_yticklabels(all_labels)
+    axes[1, 1].set_title('Support (количество примеров) по меткам')
+
+    plt.tight_layout()
+    st.pyplot(fig)
+
+    # Примеры предсказаний
+    st.subheader("Примеры предсказаний")
+    sample_indices = np.random.choice(len(predictions), min(5, len(predictions)), replace=False)
+
+    for idx in sample_indices:
+        with st.expander(f"Пример {idx + 1}"):
+            col1, col2 = st.columns(2)
+
+            with col1:
+                st.write("**Истинные метки:**")
+                st.write(true_labels[idx])
+
+            with col2:
+                st.write("**Предсказанные метки:**")
+                predicted_labels = [all_labels[i] for i, prob in enumerate(predictions[idx]['probs']) if prob >= 0.5]
+                st.write(predicted_labels)
+
+                st.write("**Вероятности:**")
+                prob_df = pd.DataFrame({
+                    'Метка': all_labels,
+                    'Вероятность': predictions[idx]['probs']
+                }).sort_values('Вероятность', ascending=False)
+                st.dataframe(prob_df, use_container_width=True)
+
+
+def process_test_file(uploaded_file, task_type):
+    """Обработка загруженного JSONL файла"""
+    data = []
+    for line in uploaded_file:
+        data.append(json.loads(line.decode('utf-8')))
+
+    df = pd.DataFrame(data)
+    return df
+
+
+def calculate_metrics(df, predictions, task_type):
+    """Расчет метрик качества"""
+    if task_type == "Бинарная":
+        y_true = df['label'].apply(lambda x: 1 if x == 'positive' else 0)
+        y_pred = [1 if pred['probs'] >= 0.5 else 0 for pred in predictions]
+        y_scores = [pred['probs'] for pred in predictions]
+
+        # ROC curve
+        fpr, tpr, _ = roc_curve(y_true, y_scores)
+        roc_auc = auc(fpr, tpr)
+
+        # Precision-Recall curve
+        precision, recall, _ = precision_recall_curve(y_true, y_scores)
+
+        return {
+            'fpr': fpr,
+            'tpr': tpr,
+            'roc_auc': roc_auc,
+            'precision': precision,
+            'recall': recall,
+            'y_true': y_true,
+            'y_pred': y_pred,
+            'y_scores': y_scores
+        }
+
+    elif task_type == "Многоклассовая":
+        # Для многоклассовой нужна более сложная обработка
+        return {"message": "Многоклассовые метрики требуют дополнительной реализации"}
+
+    else:
+        return {"message": "Многометочные метрики требуют дополнительной реализации"}
+
+
+# Основной интерфейс
+st.title("📊 Анализ классификаторов текста")
+
+# Сайдбар для навигации
+st.sidebar.title("Навигация")
+app_mode = st.sidebar.selectbox(
+    "Выберите режим",
+    ["Интерактивная классификация", "Анализ тестовой выборки"],
+    key="main_navigation"  # Уникальный ключ
+)
+
+# Интерактивная классификация
+if app_mode == "Интерактивная классификация":
+    st.header("🔍 Интерактивная классификация")
+
+    col1, col2 = st.columns([1, 1])
+
+    with col1:
+        task_type = st.selectbox(
+            "Тип задачи",
+            ["Бинарная", "Многоклассовая", "Многометочная"],
+            key="interactive_task_type"  # Уникальный ключ
+        )
+
+        available_models = list(load_models(task_type).keys())
+        if not available_models:
+            st.error("Нет доступных моделей для выбранного типа задачи")
+            st.stop()
+
+        selected_models = st.multiselect(
+            "Выберите модели для сравнения",
+            available_models,
+            default=available_models[0] if available_models else None,
+            key="interactive_models"  # Уникальный ключ
+        )
+
+        text_input = st.text_area(
+            "Введите текст для классификации",
+            height=150,
+            placeholder="Введите текст здесь...",
+            key="interactive_text_input"  # Уникальный ключ
+        )
+
+    with col2:
+        if text_input and selected_models:
+            models = load_models(task_type)
+
+            for model_name in selected_models:
+                st.subheader(f"Модель: {model_name}")
+
+                try:
+                    result = models[model_name](text_input)
+
+                    # Отображение результатов
+                    if task_type == "Бинарная":
+                        sentiment = "Positive" if result['probs'] >= 0.5 else "Negative"
+                        confidence = result['probs'] if result['probs'] >= 0.5 else 1 - result['probs']
+
+                        st.write(f"**Результат**: {sentiment}")
+                        st.write(f"**Уверенность**: {confidence:.3f}")
+
+                        # Визуализация вероятностей
+                        fig = plot_probabilities(result['probs'], result.get('labels', ['Negative', 'Positive']),
+                                                 model_name)
+                        st.pyplot(fig)
+
+                    else:
+                        if task_type == "Многоклассовая":
+                            predicted_idx = np.argmax(result['probs'])
+                            predicted_label = result['labels'][predicted_idx]
+                            confidence = result['probs'][predicted_idx]
+
+                            st.write(f"**Предсказанный класс**: {predicted_label}")
+                            st.write(f"**Уверенность**: {confidence:.3f}")
+
+                        else:  # Многометочная
+                            predicted_labels = [result['labels'][i] for i, prob in enumerate(result['probs']) if
+                                                prob >= 0.5]
+                            st.write(f"**Предсказанные классы**: {', '.join(predicted_labels)}")
+
+                        # Визуализация вероятностей
+                        fig = plot_probabilities(result['probs'], result['labels'], model_name)
+                        st.pyplot(fig)
+
+                        # Таблица вероятностей
+                        prob_df = pd.DataFrame({
+                            'Класс': result['labels'],
+                            'Вероятность': result['probs']
+                        }).sort_values('Вероятность', ascending=False)
+
+                        st.dataframe(prob_df, use_container_width=True)
+
+                except Exception as e:
+                    st.write("График не поддерживается у данной модели")
+
+# Анализ тестовой выборки
+elif app_mode == "Анализ тестовой выборки":
+    st.header("📈 Анализ тестовой выборки")
+
+    uploaded_file = st.file_uploader(
+        "Загрузите JSONL файл с тестовой выборкой",
+        type=['jsonl'],
+        help="Файл должен содержать поля 'text' и 'label' (для бинарной/многоклассовой) или 'labels' (для многометочной)",
+        key="file_uploader"
+    )
+
+    if uploaded_file:
+        task_type = st.selectbox(
+            "Тип задачи для анализа",
+            ["Бинарная", "Многоклассовая", "Многометочная"],
+            key="analysis_task_type"
+        )
+
+        available_models = list(load_models(task_type).keys())
+        if not available_models:
+            st.error("Нет доступных моделей для выбранного типа задачи")
+            st.stop()
+
+        selected_model = st.selectbox(
+            "Выберите модель для анализа",
+            available_models,
+            key="analysis_model"
+        )
+
+        if st.button("Запустить анализ", key="analyze_button"):
+            with st.spinner("Обработка данных..."):
+                # Загрузка и обработка данных
+                df = process_test_file(uploaded_file, task_type)
+                st.write(f"Загружено {len(df)} примеров")
+
+                # Проверка структуры данных
+                st.subheader("Структура данных")
+                st.dataframe(df.head(), use_container_width=True)
+
+                if task_type == "Многометочная" and 'labels' not in df.columns:
+                    st.error("Для многометочной классификации в файле должно быть поле 'labels'")
+                    st.stop()
+                elif task_type != "Многометочная" and 'label' not in df.columns:
+                    st.error("Для бинарной и многоклассовой классификации в файле должно быть поле 'label'")
+                    st.stop()
+
+                # Предсказания
+                model = load_models(task_type)[selected_model]
+                predictions = []
+                true_labels = []
+
+                progress_bar = st.progress(0)
+                for i, row in df.iterrows():
+                    try:
+                        result = model(row['text'])
+                        predictions.append(result)
+
+                        # Сохраняем истинные метки в нужном формате
+                        if task_type == "Многометочная":
+                            true_labels.append(row['labels'])
+                        else:
+                            true_labels.append(row['label'])
+
+                    except Exception as e:
+                        st.error(f"Ошибка при обработке примера {i}: {str(e)}")
+                        predictions.append(None)
+                        true_labels.append(None)
+
+                    progress_bar.progress((i + 1) / len(df))
+
+                # Удаляем примеры с ошибками
+                valid_indices = [i for i, pred in enumerate(predictions) if pred is not None]
+                predictions = [predictions[i] for i in valid_indices]
+                true_labels = [true_labels[i] for i in valid_indices]
+
+                st.write(f"Успешно обработано {len(predictions)} из {len(df)} примеров")
+
+                # Расчет и отображение метрик для разных типов задач
+                if task_type == "Бинарная":
+                    calculate_and_display_binary_metrics(true_labels, predictions)
+
+                elif task_type == "Многоклассовая":
+                    calculate_and_display_multiclass_metrics(true_labels, predictions)
+
+                elif task_type == "Многометочная":
+                    calculate_and_display_multilabel_metrics(true_labels, predictions)
+
+
+# Информация в сайдбаре
+st.sidebar.markdown("---")
+st.sidebar.info("""
+**Инструкция:**
+1. **Интерактивная классификация**: Тестируйте модели на произвольном тексте
+2. **Анализ тестовой выборки**: Загрузите JSONL файл для оценки качества
+""")
 
-"""
-# Welcome to Streamlit!
-
-Edit `/streamlit_app.py` to customize this app to your heart's desire :heart:.
-If you have any questions, checkout our [documentation](https://docs.streamlit.io) and [community
-forums](https://discuss.streamlit.io).
-
-In the meantime, below is an example of what you can do with just a few lines of code:
-"""
-
-num_points = st.slider("Number of points in spiral", 1, 10000, 1100)
-num_turns = st.slider("Number of turns in spiral", 1, 300, 31)
-
-indices = np.linspace(0, 1, num_points)
-theta = 2 * np.pi * num_turns * indices
-radius = indices
-
-x = radius * np.cos(theta)
-y = radius * np.sin(theta)
-
-df = pd.DataFrame({
-    "x": x,
-    "y": y,
-    "idx": indices,
-    "rand": np.random.randn(num_points),
-})
-
-st.altair_chart(alt.Chart(df, height=700, width=700)
-    .mark_point(filled=True)
-    .encode(
-        x=alt.X("x", axis=None),
-        y=alt.Y("y", axis=None),
-        color=alt.Color("idx", legend=None, scale=alt.Scale()),
-        size=alt.Size("rand", legend=None, scale=alt.Scale(range=[1, 150])),
-    ))
+# CSS для улучшения внешнего вида
+st.markdown("""
+<style>
+    .main-header {
+        font-size: 2.5rem;
+        color: #1f77b4;
+        text-align: center;
+        margin-bottom: 2rem;
+    }
+    .metric-card {
+        background-color: #f0f2f6;
+        padding: 1rem;
+        border-radius: 0.5rem;
+        margin: 0.5rem 0;
+    }
+</style>
+""", unsafe_allow_html=True)

src/use_ml.py

@@ -0,0 +1,76 @@
+#1 привести полученный текст к приемлемому виду
+#2 подать текст на вход к модели и получить результат
+import spacy
+from joblib import load
+
+def predict_sentiment():
+    model_binary = load("ml_binary.joblib")
+    def _inner(text: str):
+        pred = model_binary.predict([preprocess_text(text)])[0]
+        res = {
+            "labels": "positive" if pred == 1 else "negative",
+            "probs": pred
+        }
+        return res
+    return _inner
+
+
+
+def predict_category():
+    model_category = load("ml_category.joblib")
+
+    def _inner(text: str):
+        pred = model_category.predict([preprocess_text(text)])[0]
+        labels = [
+            "политика",
+            "экономика",
+            "спорт",
+            "культура"
+        ]
+        probs = [0, 0, 0, 0]
+        probs[pred] = 1
+        res = {
+            "labels": labels,
+            "probs": probs
+        }
+        return res
+    return _inner
+
+def predict_categorys():
+    model_categorys = load("ml_categorys.joblib")
+
+    def _inner(text: str):
+        pred = model_categorys.predict([preprocess_text(text)])[0]
+        labels = [
+            "политика",
+            "экономика",
+            "спорт",
+            "культура"
+        ]
+        res = {
+            "labels": labels,
+            "probs": pred
+        }
+        return res
+    return _inner
+
+def preprocess_text(text: str) -> str:
+    if text is None:
+        return ""
+
+    nlp = spacy.load("ru_core_news_md", disable=["ner"])
+
+    text = " ".join(text.split()).lower()
+
+    doc = nlp(text)
+    tokens = []
+
+    for t in doc:
+        if t.is_stop or t.is_punct or t.is_space:
+            continue
+        lemma = t.lemma_.strip()
+        if len(lemma) <= 1:
+            continue
+        tokens.append(lemma)
+
+    return " ".join(tokens)

src/use_nn.py

@@ -0,0 +1,56 @@
+import tensorflow as tf
+from use_ml import preprocess_text
+
+def predict_sentiment():
+    model = tf.keras.models.load_model("nn_binary.keras")
+    vectorizer = tf.keras.models.load_model("nn_vectorizer_binary.keras")
+    def _inner(text: str) -> str:
+        p_text = preprocess_text(text)
+        vec = vectorizer([p_text])
+        pred = model.predict(vec)[0][0]
+        res = {
+            "labels": "positive" if pred >= 0.5 else "negative",
+            "probs": pred
+        }
+        return res
+    return _inner
+
+def predict_category():
+    model = tf.keras.models.load_model("nn_category.keras")
+    vectorizer = tf.keras.models.load_model("nn_vectorizer_category.keras")
+    def _inner(text: str) -> str:
+        p_text = preprocess_text(text)
+        vec = vectorizer([p_text])
+        pred = model.predict(vec)[0]
+        labels = [
+            "политика",
+            "экономика",
+            "спорт",
+            "культура"
+        ]
+        res = {
+            "labels": labels,
+            "probs": pred
+        }
+        return res
+    return _inner
+
+def predict_categorys():
+    model = tf.keras.models.load_model("nn_categorys.keras")
+    vectorizer = tf.keras.models.load_model("nn_vectorizer_categorys.keras")
+    def _inner(text: str):
+        p_text = preprocess_text(text)
+        vec = vectorizer([p_text])
+        labels = [
+            "политика",
+            "экономика",
+            "спорт",
+            "культура"
+        ]
+        pred = model.predict(vec)[0]
+        res = {
+            "labels": labels,
+            "probs": pred
+        }
+        return res
+    return _inner

src/use_transformer.py

@@ -0,0 +1,86 @@
+from transformers import AutoTokenizer, AutoModelForSequenceClassification, pipeline
+import torch
+
+
+def predict_sentiment():
+    model_path = "./binary_model/checkpoint-400"
+    model_name = "cointegrated/rubert-tiny"
+    tokenizer = AutoTokenizer.from_pretrained(model_name)
+    model = AutoModelForSequenceClassification.from_pretrained(model_path)
+    clf = pipeline(
+        "text-classification",
+        model=model,
+        tokenizer=tokenizer,
+        return_all_scores=False
+    )
+    def _inner(text: str):
+        pred = clf(text)
+        res = {
+            "labels": pred[0]["label"],
+            "probs": pred[0]["score"]
+        }
+        return res
+    return _inner
+
+def predict_category():
+    model_path = "./category_model/checkpoint-400"
+    model_name = "cointegrated/rubert-tiny"
+    tokenizer = AutoTokenizer.from_pretrained(model_name)
+    model = AutoModelForSequenceClassification.from_pretrained(model_path)
+    clf = pipeline(
+        "text-classification",
+        model=model,
+        tokenizer=tokenizer,
+        return_all_scores=False
+    )
+    def _inner(text: str):
+        pred = clf(text)
+        labels = {"политика": 0, "экономика": 0, "спорт": 0, "культура": 0, pred[0]["label"]: pred[0]["score"]}
+        classes = [
+            "политика",
+            "экономика",
+            "спорт",
+            "культура"
+        ]
+        new_labels = []
+        for cl in classes:
+            new_labels.append(labels[cl])
+        res = {
+            "labels": classes,
+            "probs": new_labels
+        }
+        return res
+
+    return _inner
+
+def predict_categorys():
+    model_path = "./multilabel_model/checkpoint-700"
+    model_name = "cointegrated/rubert-tiny"
+    tokenizer = AutoTokenizer.from_pretrained(model_name)
+    model = AutoModelForSequenceClassification.from_pretrained(model_path)
+    model.eval()
+    classes = [
+        "политика",
+        "экономика",
+        "спорт",
+        "культура"
+    ]
+    def _inner(text: str):
+        input = tokenizer(
+            text,
+            return_tensors="pt",
+            truncation=True,
+            padding="max_length",
+            max_length=256
+        )
+        with torch.no_grad():
+            logits = model(**input).logits
+        probs = torch.sigmoid(logits).squeeze().tolist()
+        res = {
+            "labels": classes,
+            "probs": probs
+        }
+        return res
+
+
+    return _inner