mustafoyev202/uzbek_ner

README.md

@@ -25,11 +25,11 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [FacebookAI/xlm-roberta-large](https://huggingface.co/FacebookAI/xlm-roberta-large) on the Uzbek Ner dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.1458
-- Precision: 0.5927
-- Recall: 0.6370
-- F1: 0.6140
-- Accuracy: 0.9479
+- Loss: 0.1761
+- Precision: 0.5870
+- Recall: 0.6354
+- F1: 0.6102
+- Accuracy: 0.9386
 
 ## Model description
 
@@ -55,8 +55,8 @@ The following hyperparameters were used during training:
 - gradient_accumulation_steps: 8
 - total_train_batch_size: 64
 - optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
-- lr_scheduler_type: cosine
-- lr_scheduler_warmup_ratio: 0.1
+- lr_scheduler_type: cosine_with_restarts
+- lr_scheduler_warmup_ratio: 0.08
 - num_epochs: 3
 - mixed_precision_training: Native AMP
 
@@ -64,9 +64,12 @@ The following hyperparameters were used during training:
 
 | Training Loss | Epoch  | Step | Validation Loss | Precision | Recall | F1     | Accuracy |
 |:-------------:|:------:|:----:|:---------------:|:---------:|:------:|:------:|:--------:|
-| 0.1673        | 0.9324 | 200  | 0.1592          | 0.5325    | 0.6170 | 0.5717 | 0.9438   |
-| 0.1503        | 1.8625 | 400  | 0.1476          | 0.5929    | 0.6240 | 0.6081 | 0.9477   |
-| 0.1394        | 2.7925 | 600  | 0.1458          | 0.5927    | 0.6370 | 0.6140 | 0.9479   |
+| 0.2571        | 0.4662 | 100  | 0.2272          | 0.4924    | 0.5096 | 0.5008 | 0.9291   |
+| 0.2035        | 0.9324 | 200  | 0.1931          | 0.5411    | 0.5962 | 0.5673 | 0.9339   |
+| 0.1787        | 1.3963 | 300  | 0.1846          | 0.5693    | 0.6327 | 0.5993 | 0.9358   |
+| 0.1788        | 1.8625 | 400  | 0.1776          | 0.5741    | 0.6259 | 0.5989 | 0.9383   |
+| 0.176         | 2.3263 | 500  | 0.1759          | 0.5902    | 0.6231 | 0.6062 | 0.9390   |
+| 0.1676        | 2.7925 | 600  | 0.1761          | 0.5868    | 0.6351 | 0.6100 | 0.9386   |
 
 
 ### Framework versions

config.json

@@ -1,10 +1,33 @@
 {
-  "_name_or_path": "FacebookAI/xlm-roberta-large",
+  "_name_or_path": "xlm-roberta-large",
   "architectures": [
     "XLMRobertaForTokenClassification"
   ],
   "attention_probs_dropout_prob": 0.1,
   "bos_token_id": 0,
+  "class_weights": [
+    0.053862977772951126,
+    251.441162109375,
+    878.10986328125,
+    8.121392250061035,
+    6.7787580490112305,
+    35.58425521850586,
+    21.885408401489258,
+    4.541465759277344,
+    31.51692008972168,
+    2.9324467182159424,
+    6.667890548706055,
+    16.192096710205078,
+    8.184276580810547,
+    4.891348361968994,
+    2.876582145690918,
+    113.24829864501953,
+    125.99810791015625,
+    3.378345251083374,
+    3.566652774810791,
+    120.9261474609375,
+    223.83193969726562
+  ],
   "classifier_dropout": null,
   "eos_token_id": 2,
   "hidden_act": "gelu",
@@ -21,7 +44,18 @@
     "7": "LABEL_7",
     "8": "LABEL_8",
     "9": "LABEL_9",
-    "10": "LABEL_10"
+    "10": "LABEL_10",
+    "11": "LABEL_11",
+    "12": "LABEL_12",
+    "13": "LABEL_13",
+    "14": "LABEL_14",
+    "15": "LABEL_15",
+    "16": "LABEL_16",
+    "17": "LABEL_17",
+    "18": "LABEL_18",
+    "19": "LABEL_19",
+    "20": "LABEL_20",
+    "21": "LABEL_21"
   },
   "initializer_range": 0.02,
   "intermediate_size": 4096,
@@ -29,7 +63,18 @@
     "LABEL_0": 0,
     "LABEL_1": 1,
     "LABEL_10": 10,
+    "LABEL_11": 11,
+    "LABEL_12": 12,
+    "LABEL_13": 13,
+    "LABEL_14": 14,
+    "LABEL_15": 15,
+    "LABEL_16": 16,
+    "LABEL_17": 17,
+    "LABEL_18": 18,
+    "LABEL_19": 19,
     "LABEL_2": 2,
+    "LABEL_20": 20,
+    "LABEL_21": 21,
     "LABEL_3": 3,
     "LABEL_4": 4,
     "LABEL_5": 5,
@@ -39,19 +84,6 @@
     "LABEL_9": 9
   },
   "layer_norm_eps": 1e-05,
-  "loss_weight": [
-    1.0,
-    1.0,
-    1.0,
-    1.0,
-    1.0,
-    1.0,
-    1.0,
-    1.0,
-    1.0,
-    1.0,
-    1.0
-  ],
   "max_position_embeddings": 514,
   "model_type": "xlm-roberta",
   "num_attention_heads": 16,
@@ -59,7 +91,7 @@
   "output_past": true,
   "pad_token_id": 1,
   "position_embedding_type": "absolute",
-  "torch_dtype": "float32",
+  "torch_dtype": "float16",
   "transformers_version": "4.49.0",
   "type_vocab_size": 1,
   "use_cache": true,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c83f02de25ff8e8d8d2309647cc79245adeae65840b328b0cc93f9fead24d7aa
-size 2235456956
+oid sha256:131c49bb9375a2fb85424cda466c04b32b120961b80b1f90c8bde5e29395a7ea
+size 1117774660

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4978fa21f4e079dc360ba47132ee01a1228a34765b3e2d6231d3c955e59953ce
-size 5240
+oid sha256:9f2137e1f9d20a81d6b09a5a9121c3731c957f75f5a4028372eecaa3676d98ea
+size 5304