gliner-nightmare

Private GLiNER NER model fine-tuned for PII detection.

Model Details

Field	Value
Base model	mdeberta-v3-base
Architecture	GLiNER (markerV0)
Hidden size	512
Fine-tuned by	arthrod
Framework	GLiNER

Evaluation Status

Not evaluated. This model was not included in any evaluation run. None of the 15 checkpoints have been benchmarked.

Checkpoint	Evaluated
checkpoint-250	No
checkpoint-500	No
checkpoint-750	No
checkpoint-1000	No
checkpoint-1250	No
checkpoint-1500	No
checkpoint-1750	No
checkpoint-2000	No
checkpoint-2250	No
checkpoint-2500	No
checkpoint-2750	No
checkpoint-3000	No
checkpoint-3250	No
checkpoint-3500	No
checkpoint-3750	No

Run Evaluation

Click to expand auto-evaluation script

This script evaluates the model on the same 8 benchmark datasets used for all arthrod GLiNER models. Results are written to an evals-*/ folder with CSV metrics and plots.

# 1. Clone the evaluation toolkit
git clone https://github.com/arthrod/gliner_review.git && cd gliner_review

# 2. Install dependencies
pip install gliner typer pandas seaborn matplotlib huggingface-hub

# 3. Run evaluation (auto-detects GPU)
python evaluate_gliner.py run \
  --models arthrod/gliner-nightmare --include-checkpoints \
  --datasets arthrod/gliner-canonical-ptbr-pii-v1,Ihor/gliner-post-train,knowledgator/GLINER-multi-task-synthetic-data,knowledgator/gliner-multilingual-synthetic,nvidia/Nemotron-PII,urchade/pile-mistral-v0.1,urchade/pubmed-ner-mistral-v0.1,urchade/synthetic-pii-ner-mistral-v1 \
  --threshold 0.5 \
  --batch-size 8 \
  --token $HF_TOKEN

# Results will be in evals-YYYYMMDD-HHMM/
# - overall_metrics.csv      (P/R/F1 per model)
# - per_entity_metrics.csv   (P/R/F1 per entity type)
# - overall_metrics_per_model.png
# - entity_f1_heatmap.png

To compare with the best arthrod model:

python evaluate_gliner.py run \
  --models arthrod/gliner-nightmare,arthrod/gliner-pii-mmbert-base-token-v1.0 --include-checkpoints \
  --datasets arthrod/gliner-canonical-ptbr-pii-v1,Ihor/gliner-post-train,knowledgator/GLINER-multi-task-synthetic-data,knowledgator/gliner-multilingual-synthetic,nvidia/Nemotron-PII,urchade/pile-mistral-v0.1,urchade/pubmed-ner-mistral-v0.1,urchade/synthetic-pii-ner-mistral-v1 \
  --threshold 0.5 \
  --token $HF_TOKEN

## Usage

from gliner import GLiNER

model = GLiNER.from_pretrained("arthrod/gliner-nightmare")
entities = model.predict_entities(text, labels, threshold=0.5)

Config

gliner_config.json

{
  "class_token_index": 250103,
  "dropout": 0.4,
  "embed_ent_token": true,
  "encoder_config": {
    "_name_or_path": "microsoft/mdeberta-v3-base",
    "add_cross_attention": false,
    "architectures": null,
    "attention_probs_dropout_prob": 0.1,
    "bad_words_ids": null,
    "begin_suppress_tokens": null,
    "bos_token_id": null,
    "chunk_size_feed_forward": 0,
    "cross_attention_hidden_size": null,
    "decoder_start_token_id": null,
    "diversity_penalty": 0.0,
    "do_sample": false,
    "early_stopping": false,
    "encoder_no_repeat_ngram_size": 0,
    "eos_token_id": null,
    "exponential_decay_length_penalty": null,
    "finetuning_task": null,
    "forced_bos_token_id": null,
    "forced_eos_token_id": null,
    "hidden_act": "gelu",
    "hidden_dropout_prob": 0.1,
    "hidden_size": 768,
    "id2label": {
      "0": "LABEL_0",
      "1": "LABEL_1"
    },
    "initializer_range": 0.02,
    "intermediate_size": 3072,
    "is_decoder": false,
    "is_encoder_decoder": false,
    "label2id": {
      "LABEL_0": 0,
      "LABEL_1": 1
    },
    "layer_norm_eps": 1e-07,
    "legacy": true,
    "length_penalty": 1.0,
    "max_length": 20,
    "max_position_embeddings": 512,
    "max_relative_positions": -1,
    "min_length": 0,
    "model_type": "deberta-v2",
    "no_repeat_ngram_size": 0,
    "norm_rel_ebd": "layer_norm",
    "num_attention_heads": 12,
    "num_beam_groups": 1,
    "num_beams": 1,
    "num_hidden_layers": 12,
    "num_return_sequences": 1,
    "output_attentions": false,
    "output_hidden_states": false,
    "output_scores": false,
    "pad_token_id": 0,
    "pooler_dropout": 0,
    "pooler_hidden_act": "gelu",
    "pooler_hidden_size": 768,
    "pos_att_type": [
      "p2c",
      "c2p"
    ],
    "position_biased_input": false,
    "position_buckets": 256,
    "prefix": null,
    "problem_type": null,
    "pruned_heads": {},
    "relative_attention": true,
    "remove_invalid_values": false,
    "repetition_penalty": 1.0,
    "return_dict": true,
    "return_dict_in_generate": false,
    "sep_token_id": null,
    "share_att_key": true,
    "suppress_tokens": null,
    "task_specific_params": null,
    "temperature": 1.0,
    "tf_legacy_loss": false,
    "tie_encoder_decoder": false,
    "tie_word_embeddings": true,
    "tokenizer_class": null,
    "top_k": 50,
    "top_p": 1.0,
    "torch_dtype": null,
    "torchscript": false,
    "type_vocab_size": 0,
    "typical_p": 1.0,
    "use_bfloat16": false,
    "vocab_size": 250105
  },
  "ent_token": "<<ENT>>",
  "eval_every": 5000,
  "fine_tune": true,
  "fuse_layers": false,
  "has_rnn": true,
  "hidden_size": 512,
  "labels_encoder": null,
  "labels_encoder_config": null,
  "lr_encoder": "5e-6",
  "lr_others": "1e-5",
  "max_len": 384,
  "max_neg_type_ratio": 3,
  "max_types": 25,
  "max_width": 12,
  "model_name": "microsoft/mdeberta-v3-base",
  "model_type": "gliner",
  "name": "correct",
  "num_post_fusion_layers": 1,
  "num_steps": 30000,
  "post_fusion_schema": "",
  "random_drop": false,
  "sep_token": "<<SEP>>",
  "shuffle_types": true,
  "size_sup": -1,
  "span_mode": "markerV0",
  "subtoken_pooling": "first",
  "train_batch_size": 8,
  "transformers_version": "4.55.0",
  "vocab_size": 250105,
  "warmup_ratio": 5000,
  "words_splitter_type": "whitespace"
}