RuModernBERT-small ner finetuned

README.md

@@ -0,0 +1,83 @@
+---
+library_name: transformers
+license: apache-2.0
+base_model: deepvk/RuModernBERT-small
+tags:
+- generated_from_trainer
+metrics:
+- f1
+- precision
+- recall
+model-index:
+- name: rumodernbert_ner_ft_small
+  results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+# rumodernbert_ner_ft_small
+
+This model is a fine-tuned version of [deepvk/RuModernBERT-small](https://huggingface.co/deepvk/RuModernBERT-small) on an unknown dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.2842
+- F1: 0.8257
+- Precision: 0.8080
+- Recall: 0.8443
+
+## Model description
+
+More information needed
+
+## Intended uses & limitations
+
+More information needed
+
+## Training and evaluation data
+
+More information needed
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 2e-05
+- train_batch_size: 16
+- eval_batch_size: 32
+- seed: 42
+- gradient_accumulation_steps: 2
+- total_train_batch_size: 32
+- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
+- lr_scheduler_type: linear
+- lr_scheduler_warmup_steps: 0.1
+- num_epochs: 10
+- mixed_precision_training: Native AMP
+
+### Training results
+
+| Training Loss | Epoch | Step | Validation Loss | F1     | Precision | Recall |
+|:-------------:|:-----:|:----:|:---------------:|:------:|:---------:|:------:|
+| 1.3906        | 0.32  | 200  | 0.6057          | 0.4393 | 0.3840    | 0.5132 |
+| 0.8110        | 0.64  | 400  | 0.3561          | 0.6269 | 0.5835    | 0.6774 |
+| 0.6350        | 0.96  | 600  | 0.3008          | 0.7213 | 0.6812    | 0.7665 |
+| 0.4992        | 1.28  | 800  | 0.2633          | 0.7803 | 0.7699    | 0.7909 |
+| 0.3823        | 1.6   | 1000 | 0.2300          | 0.8084 | 0.7936    | 0.8238 |
+| 0.3995        | 1.92  | 1200 | 0.2162          | 0.8009 | 0.7918    | 0.8102 |
+| 0.2860        | 2.24  | 1400 | 0.2195          | 0.8109 | 0.7988    | 0.8234 |
+| 0.2833        | 2.56  | 1600 | 0.2065          | 0.8140 | 0.7955    | 0.8335 |
+| 0.2660        | 2.88  | 1800 | 0.2374          | 0.8145 | 0.8020    | 0.8274 |
+| 0.1466        | 3.2   | 2000 | 0.2693          | 0.8079 | 0.7838    | 0.8335 |
+| 0.1809        | 3.52  | 2200 | 0.2504          | 0.8299 | 0.8210    | 0.8391 |
+| 0.1685        | 3.84  | 2400 | 0.2267          | 0.8315 | 0.8134    | 0.8503 |
+| 0.0860        | 4.16  | 2600 | 0.2938          | 0.8273 | 0.8106    | 0.8447 |
+| 0.0884        | 4.48  | 2800 | 0.2980          | 0.8183 | 0.7959    | 0.8419 |
+| 0.1116        | 4.8   | 3000 | 0.2842          | 0.8257 | 0.8080    | 0.8443 |
+
+
+### Framework versions
+
+- Transformers 5.1.0
+- Pytorch 2.10.0+cu128
+- Datasets 4.7.0
+- Tokenizers 0.22.2

config.json

@@ -0,0 +1,122 @@
+{
+  "activation_function": "gelu",
+  "allow_embedding_resizing": true,
+  "architectures": [
+    "ModernBertForTokenClassification"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "attention_layer": "rope",
+  "attention_probs_dropout_prob": 0.0,
+  "attn_out_bias": false,
+  "attn_out_dropout_prob": 0.1,
+  "attn_qkv_bias": false,
+  "bert_layer": "prenorm",
+  "bos_token_id": null,
+  "classifier_activation": "gelu",
+  "classifier_bias": false,
+  "classifier_dropout": 0.0,
+  "classifier_pooling": "cls",
+  "cls_token_id": 50281,
+  "compile_model": true,
+  "decoder_bias": true,
+  "deterministic_flash_attn": false,
+  "dtype": "float32",
+  "embed_dropout_prob": 0.0,
+  "embed_norm": true,
+  "embedding_dropout": 0.0,
+  "embedding_layer": "sans_pos",
+  "eos_token_id": null,
+  "final_norm": true,
+  "global_attn_every_n_layers": 3,
+  "head_pred_act": "gelu",
+  "hidden_act": "gelu",
+  "hidden_activation": "gelu",
+  "hidden_size": 384,
+  "id2label": {
+    "0": "O",
+    "1": "B-PER",
+    "2": "I-PER",
+    "3": "B-ORG",
+    "4": "I-ORG",
+    "5": "B-LOC",
+    "6": "I-LOC"
+  },
+  "init_method": "full_megatron",
+  "initializer_cutoff_factor": 2.0,
+  "initializer_range": 0.02,
+  "intermediate_size": 576,
+  "label2id": {
+    "B-LOC": 5,
+    "B-ORG": 3,
+    "B-PER": 1,
+    "I-LOC": 6,
+    "I-ORG": 4,
+    "I-PER": 2,
+    "O": 0
+  },
+  "layer_types": [
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention"
+  ],
+  "local_attention": 256,
+  "local_attn_rotary_emb_base": 10000.0,
+  "loss_function": "fa_cross_entropy",
+  "loss_kwargs": {
+    "reduction": "mean"
+  },
+  "masked_prediction": true,
+  "max_position_embeddings": 8192,
+  "mlp_bias": false,
+  "mlp_dropout": 0.0,
+  "mlp_dropout_prob": 0.0,
+  "mlp_in_bias": false,
+  "mlp_layer": "glu",
+  "mlp_out_bias": false,
+  "model_type": "modernbert",
+  "norm_bias": false,
+  "norm_eps": 1e-05,
+  "norm_kwargs": {
+    "bias": false,
+    "eps": 1e-05
+  },
+  "normalization": "layernorm",
+  "num_attention_heads": 6,
+  "num_hidden_layers": 12,
+  "pad_token_id": 50283,
+  "padding": "unpadded",
+  "repad_logits_with_grad": false,
+  "rope_parameters": {
+    "full_attention": {
+      "rope_theta": 160000.0,
+      "rope_type": "default"
+    },
+    "sliding_attention": {
+      "rope_theta": 10000.0,
+      "rope_type": "default"
+    }
+  },
+  "rotary_emb_base": 160000.0,
+  "rotary_emb_dim": null,
+  "rotary_emb_interleaved": false,
+  "rotary_emb_scale_base": null,
+  "sep_token_id": 50282,
+  "skip_first_prenorm": true,
+  "sparse_pred_ignore_index": -100,
+  "sparse_prediction": false,
+  "tie_word_embeddings": true,
+  "transformers_version": "5.1.0",
+  "unpad_embeddings": true,
+  "use_cache": false,
+  "vocab_size": 50368
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4ced7cc9dfab7239745451679d0811a6a2cbf2a70874b519a50c31c664e9e1a
+size 138175724

tokenizer_config.json

@@ -0,0 +1,16 @@
+{
+  "backend": "tokenizers",
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "[CLS]",
+  "is_local": false,
+  "mask_token": "[MASK]",
+  "model_input_names": [
+    "input_ids",
+    "attention_mask"
+  ],
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": "[UNK]"
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fdd24b1aaf09472a19daaeb68d733040305aab072a68a23a778e9aa05546f6c5
+size 5201