gguichard/matching-rh-peft3

SentenceTransformer based on EuroBERT/EuroBERT-210m

This is a sentence-transformers model finetuned from EuroBERT/EuroBERT-210m on the matching_rh_train10 dataset. It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: EuroBERT/EuroBERT-210m
• Maximum Sequence Length: 8192 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • matching_rh_train10

Model Sources

• Documentation: Sentence Transformers Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 8192, 'do_lower_case': False, 'architecture': 'EuroBertModel'})
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("gguichard/matching-rh-peft3")
# Run inference
sentences = [
    '{"type": "opportunity", "customer_code": "", "opportunity_title": ".NET Developer", "opportunity_place": "", "opportunity_expertise_area": "Autres", "opportunity_tools": "", "opportunity_activity_area": "", "opportunity_type": "1", "opportunity_description": ".NET\\nReact", "opportunity_criteria": "", "opportunity_extract": 1}',
    '{"type": "candidate", "customer_code": "", "title": "Agile Back end  Developer", "skills": "", "education": "", "experience": "-1", "tools": "", "languages": "", "mobility": "", "expertise_area": "", "activity_area": "", "list_diplomes": "", "typeOf": "0", "source": "", "informationComments": "", "extract": 1, "experiences": "[]"}',
    '{"type": "candidate", "customer_code": "", "title": "Consultant Data", "skills": "", "education": "", "experience": "-1", "tools": "", "languages": "", "mobility": "mondeeuropefrancerhonealpes", "expertise_area": "", "activity_area": "", "list_diplomes": "", "typeOf": "-1", "source": "3", "informationComments": "pas à l\'écoute", "extract": 1, "experiences": "[]"}',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.8869, 0.1913],
#         [0.8869, 1.0000, 0.2530],
#         [0.1913, 0.2530, 1.0000]])

Training Details

Training Dataset

matching_rh_train10

• Dataset: matching_rh_train10 at 601ef4d
• Size: 297,400 training samples
• Columns: label, sentence1, and sentence2
• Approximate statistics based on the first 1000 samples:

  	label	sentence1	sentence2
  type	float	string	string
  details	min: 0.0 mean: 0.81 max: 1.0	min: 82 tokens mean: 326.44 tokens max: 1277 tokens	min: 95 tokens mean: 1200.82 tokens max: 6900 tokens

• Samples:

  label	sentence1	sentence2
  1.0	{"type": "opportunity", "customer_code": "", "opportunity_title": "SIENNA - DEV DOT NET", "opportunity_place": "", "opportunity_expertise_area": "Banque", "opportunity_tools": "", "opportunity_activity_area": "", "opportunity_type": "1", "opportunity_description": "", "opportunity_criteria": "", "opportunity_extract": 1}	{"type": "candidate", "customer_code": "", "title": "Consultant Sénior Microsoft .NET", "skills": "", "education": "", "experience": "-1", "tools": "", "languages": "", "mobility": "", "expertise_area": "", "activity_area": "", "list_diplomes": "2007 - Master Management des projets informatiques et systèmes d'information, 2004 - Filière Informatique et Réseaux - ENSICAEN", "typeOf": "1", "source": "1", "informationComments": "", "extract": 1, "experiences": "[{'skills': '', 'startMonth': '6', 'endDate': '', 'startYear': '2004', 'description': 'AUTRES MISSIONS\nA\nIngénieur Conception et développement CALCIA\nAnalyste - Responsable d’applications chez EDF\nIngénieur Conception et Développement chez EDF\nIngénieur Conception et Développement chez BNPPARIBAS', 'company': 'AUTRES MISSIONS', 'location': '', 'id': '2536', 'title': 'Ingénieur Conception et développement', 'endMonth': '11', 'endYear': '2008', 'startDate': ''}, {'skills': '.net, .net 2.0, asp.net, c#, front office, gamaweb...
  1.0	{"type": "opportunity", "customer_code": "", "opportunity_title": "Consultant Mainframe - DGFIP - ONEPOINT", "opportunity_place": "", "opportunity_expertise_area": "Autres", "opportunity_tools": "", "opportunity_activity_area": "", "opportunity_type": "1", "opportunity_description": "", "opportunity_criteria": "", "opportunity_extract": 1}	{"type": "candidate", "customer_code": "", "title": "Ingénieur de développement\nPACBASE/COBOL/MAINFRAME\n2 ans et ½ d’expérience", "skills": "", "education": "", "experience": "-1", "tools": "", "languages": "français, anglais", "mobility": "mondeeuropefranceiledefranceparis, mondeeuropefranceiledefranceseineetmarne, mondeeuropefranceiledefranceyvelines, mondeeuropefranceiledefranceessone, mondeeuropefranceiledefrancehautsdeseine92, mondeeuropefranceiledefranceseinesaintdenis, mondeeuropefranceiledefrancevaldemarne, mondeeuropefranceiledefrancevaloise", "expertise_area": "", "activity_area": "", "list_diplomes": "2018 - Formation PACBASE - Banque Populaire Dijon, 2018 - Formation Cobol en alternance appliqué au contexte Descours & Cabaud - Alteca Lyon et Informatique, 2018 - Formation interne VBA EXCEL, 2018 - Formation Mainframe IBM/COBOL et Qualification logiciel - INTI Formation, 2016 - Master international Science de la matière - Université de Rouen", "typeOf": "1", "source": "3",...
  1.0	{"type": "opportunity", "customer_code": "", "opportunity_title": "STIME responsable application adjoint", "opportunity_place": "", "opportunity_expertise_area": "Grande distribution", "opportunity_tools": "", "opportunity_activity_area": "", "opportunity_type": "1", "opportunity_description": "", "opportunity_criteria": "", "opportunity_extract": 1}	{"type": "candidate", "customer_code": "", "title": "Consultant AMOA- Chef de projet SI", "skills": "", "education": "", "experience": "-1", "tools": "", "languages": "anglais, espagnol", "mobility": "", "expertise_area": "", "activity_area": "", "list_diplomes": "2020 - CERTYOU Paris, 2019 - Certification SCRUM Master - Actinuum Paris, 2017 - Cycle Project Management Professional V5 PMP, 2015 - Urbanisation et architecture SI, 2014 - ITIL Fondation", "typeOf": "1", "source": "1", "informationComments": "", "extract": 1, "experiences": "[{'skills': 'crm, oracle parties, mep, dba, infrastructure, crm people soft, uml, power amc, sql query, oracle, hp quality', 'startMonth': '4', 'endDate': '', 'startYear': '2007', 'description': 'INWI\nà\nSynthèse :\nParticipation à la mise en place du CRM pepoleSoft Oracle parties : vue 360°\nclient , facture et réclamations.\nRôle :\nConsultant AMOA homologation\nRéalisation :\n\uf0b7\nCollecte de besoin métier.\n\uf0b7\nRédaction de spéc...

• Loss: CosineSimilarityLoss with these parameters:

  {
      "loss_fct": "torch.nn.modules.loss.MSELoss"
  }
  

Evaluation Dataset

matching_rh_val10

• Dataset: matching_rh_val10 at 16fd0da
• Size: 17,380 evaluation samples
• Columns: label, sentence1, and sentence2
• Approximate statistics based on the first 1000 samples:

  	label	sentence1	sentence2
  type	float	string	string
  details	min: 0.0 mean: 0.84 max: 1.0	min: 80 tokens mean: 352.97 tokens max: 3661 tokens	min: 90 tokens mean: 615.01 tokens max: 6579 tokens

• Samples:

  label	sentence1	sentence2
  1.0	{"type": "opportunity", "customer_code": "", "opportunity_title": "DATA MANAGER - La POSTE", "opportunity_place": "", "opportunity_expertise_area": "Services", "opportunity_tools": "", "opportunity_activity_area": "", "opportunity_type": "1", "opportunity_description": "", "opportunity_criteria": "", "opportunity_extract": 1}	{"type": "candidate", "customer_code": "", "title": "Senior Consultant/Project Manager - Data Management", "skills": "", "education": "", "experience": "-1", "tools": "", "languages": "", "mobility": "", "expertise_area": "", "activity_area": "", "list_diplomes": "BACHELOR - Mathématiques Appliquées - stratégique Université Paris I Panthéon Sorbonne, DEUG - Option Statistique - stratégique Université Paris I Panthéon Sorbonne", "typeOf": "-1", "source": "1", "informationComments": "adresse perso consultant : 99 rue Alfred DININ 92000 Nanterre", "extract": 1, "experiences": "[{'skills': '', 'startMonth': '', 'endDate': '', 'startYear': '', 'description': "Avril ❖Mission : * Automatisation et fiabilisation des calculs de l'inventaire de réassurance sur les produits de prévoyance individuelle commercialisés par les partenaires d'Axa France (SAS/SQL) * Etude de l'efficience et de la rentabilité des traités de réassurance mis en place pour sécuriser le portefeuille de ces produits (SAS/C++...
  1.0	{"type": "opportunity", "customer_code": "", "opportunity_title": "BABILOU - Responsable infra", "opportunity_place": "", "opportunity_expertise_area": "Autres", "opportunity_tools": "", "opportunity_activity_area": "", "opportunity_type": "1", "opportunity_description": "", "opportunity_criteria": "", "opportunity_extract": 1}	{"type": "candidate", "customer_code": "", "title": "CHEF DE PROJET INFRASTRUCTURE", "skills": "", "education": "", "experience": "-1", "tools": "", "languages": "", "mobility": "", "expertise_area": "", "activity_area": "", "list_diplomes": "2020 - Microsoft Azure Artificial Intelligence - Microsoft Azure Fundamentals, 2014 - DEA - Probabilités et Applications - Université, 2003 - Diplôme d'ingénieur - Télécoms ENST ParisTech, 2003 - DEA - Signal et Communications Numériques - Université de Nice Sophia-Antipolis", "typeOf": "-1", "source": "1", "informationComments": "", "extract": 1, "experiences": "[{'skills': '', 'startMonth': '', 'endDate': '', 'startYear': '', 'description': '23 mois Études, architecture, ingénierie et paramétrage des réseaux de signalisation et de transit', 'company': '', 'location': '', 'id': '1947', 'title': 'Ingénieur accès fixe et mobile - Contexte - 01/10/2005 - 01/08/2007', 'endMonth': '', 'endYear': '', 'startDate': ''}, {'skills': '', 'startMonth': '', '...
  1.0	{"type": "opportunity", "customer_code": "", "opportunity_title": "DGFIP - ONEPOINT - Consultant JCL", "opportunity_place": "", "opportunity_expertise_area": "Autres", "opportunity_tools": "", "opportunity_activity_area": "", "opportunity_type": "1", "opportunity_description": "", "opportunity_criteria": "", "opportunity_extract": 1}	{"type": "candidate", "customer_code": "", "title": "analyste developpeur pacbase cobol db2", "skills": "cobol, pacbase, db2, cics", "education": "", "experience": "-1", "tools": "", "languages": "", "mobility": "mondeeuropefranceiledefranceparis, mondeeuropefranceiledefranceseineetmarne, mondeeuropefranceiledefranceyvelines, mondeeuropefranceiledefranceessone, mondeeuropefranceiledefrancehautsdeseine92, mondeeuropefranceiledefranceseinesaintdenis, mondeeuropefranceiledefrancevaldemarne, mondeeuropefranceiledefrancevaloise", "expertise_area": "", "activity_area": "", "list_diplomes": "", "typeOf": "0", "source": "", "informationComments": "Sabrina Kadrie\n06 83 65 01 64\nsabrina20@orange.fr", "extract": 1, "experiences": "[]"}

• Loss: CosineSimilarityLoss with these parameters:

  {
      "loss_fct": "torch.nn.modules.loss.MSELoss"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 4
• per_device_eval_batch_size: 4
• learning_rate: 2e-05
• num_train_epochs: 1
• warmup_ratio: 0.1
• log_level: error
• log_level_replica: passive
• log_on_each_node: False
• logging_nan_inf_filter: False
• bf16: True

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 4
• per_device_eval_batch_size: 4
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 1
• eval_accumulation_steps: None
• torch_empty_cache_steps: None
• learning_rate: 2e-05
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 1
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.1
• warmup_steps: 0
• log_level: error
• log_level_replica: passive
• log_on_each_node: False
• logging_nan_inf_filter: False
• save_safetensors: True
• save_on_each_node: False
• save_only_model: False
• restore_callback_states_from_checkpoint: False
• no_cuda: False
• use_cpu: False
• use_mps_device: False
• seed: 42
• data_seed: None
• jit_mode_eval: False
• use_ipex: False
• bf16: True
• fp16: False
• fp16_opt_level: O1
• half_precision_backend: auto
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• local_rank: 0
• ddp_backend: None
• tpu_num_cores: None
• tpu_metrics_debug: False
• debug: []
• dataloader_drop_last: False
• dataloader_num_workers: 0
• dataloader_prefetch_factor: None
• past_index: -1
• disable_tqdm: False
• remove_unused_columns: True
• label_names: None
• load_best_model_at_end: False
• ignore_data_skip: False
• fsdp: []
• fsdp_min_num_params: 0
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
• fsdp_transformer_layer_cls_to_wrap: None
• accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
• parallelism_config: None
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch_fused
• optim_args: None
• adafactor: False
• group_by_length: False
• length_column_name: length
• ddp_find_unused_parameters: None
• ddp_bucket_cap_mb: None
• ddp_broadcast_buffers: False
• dataloader_pin_memory: True
• dataloader_persistent_workers: False
• skip_memory_metrics: True
• use_legacy_prediction_loop: False
• push_to_hub: False
• resume_from_checkpoint: None
• hub_model_id: None
• hub_strategy: every_save
• hub_private_repo: None
• hub_always_push: False
• hub_revision: None
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• include_inputs_for_metrics: False
• include_for_metrics: []
• eval_do_concat_batches: True
• fp16_backend: auto
• push_to_hub_model_id: None
• push_to_hub_organization: None
• mp_parameters:
• auto_find_batch_size: False
• full_determinism: False
• torchdynamo: None
• ray_scope: last
• ddp_timeout: 1800
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• include_tokens_per_second: False
• include_num_input_tokens_seen: False
• neftune_noise_alpha: None
• optim_target_modules: None
• batch_eval_metrics: False
• eval_on_start: False
• use_liger_kernel: False
• liger_kernel_config: None
• eval_use_gather_object: False
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Click to expand

Epoch	Step	Training Loss	Validation Loss
0.0067	500	0.2078	-
0.0134	1000	0.1805	-
0.0202	1500	0.1644	-
0.0269	2000	0.1455	-
0.0336	2500	0.1326	-
0.0403	3000	0.132	0.1514
0.0471	3500	0.1292	-
0.0538	4000	0.1199	-
0.0605	4500	0.1223	-
0.0672	5000	0.1219	-
0.0740	5500	0.1116	-
0.0807	6000	0.1149	0.1483
0.0874	6500	0.1149	-
0.0941	7000	0.1243	-
0.1009	7500	0.1204	-
0.1076	8000	0.1116	-
0.1143	8500	0.109	-
0.1210	9000	0.111	0.1289
0.1278	9500	0.1168	-
0.1345	10000	0.1121	-
0.1412	10500	0.1054	-
0.1479	11000	0.1031	-
0.1547	11500	0.0994	-
0.1614	12000	0.0968	0.1204
0.1681	12500	0.0932	-
0.1748	13000	0.0978	-
0.1816	13500	0.0996	-
0.1883	14000	0.0974	-
0.1950	14500	0.095	-
0.2017	15000	0.0926	0.1139
0.2085	15500	0.0928	-
0.2152	16000	0.1007	-
0.2219	16500	0.0933	-
0.2286	17000	0.0903	-
0.2354	17500	0.0912	-
0.2421	18000	0.0927	0.1124
0.2488	18500	0.0927	-
0.2555	19000	0.1001	-
0.2623	19500	0.0951	-
0.2690	20000	0.0893	-
0.2757	20500	0.0874	-
0.2824	21000	0.0854	0.1100
0.2892	21500	0.0905	-
0.2959	22000	0.0858	-
0.3026	22500	0.0906	-
0.3093	23000	0.0899	-
0.3161	23500	0.0861	-
0.3228	24000	0.0934	0.1063
0.3295	24500	0.0995	-
0.3362	25000	0.0905	-
0.3430	25500	0.0875	-
0.3497	26000	0.074	-
0.3564	26500	0.0875	-
0.3631	27000	0.0821	0.1043
0.3699	27500	0.0877	-
0.3766	28000	0.0837	-
0.3833	28500	0.0854	-
0.3900	29000	0.0754	-
0.3968	29500	0.0803	-
0.4035	30000	0.0872	0.1029
0.4102	30500	0.0829	-
0.4169	31000	0.0841	-
0.4237	31500	0.0861	-
0.4304	32000	0.0827	-
0.4371	32500	0.0867	-
0.4438	33000	0.0808	0.1028
0.4506	33500	0.081	-
0.4573	34000	0.0789	-
0.4640	34500	0.0774	-
0.4707	35000	0.084	-
0.4775	35500	0.0866	-
0.4842	36000	0.0839	0.1010
0.4909	36500	0.0849	-
0.4976	37000	0.0834	-
0.5044	37500	0.0832	-
0.5111	38000	0.0739	-
0.5178	38500	0.077	-
0.5245	39000	0.0799	0.1016
0.5313	39500	0.0775	-
0.5380	40000	0.0788	-
0.5447	40500	0.0821	-
0.5514	41000	0.0796	-
0.5582	41500	0.0795	-
0.5649	42000	0.0836	0.0976
0.5716	42500	0.0783	-
0.5783	43000	0.082	-
0.5851	43500	0.0788	-
0.5918	44000	0.0849	-
0.5985	44500	0.0754	-
0.6052	45000	0.0764	0.0989
0.6120	45500	0.0736	-
0.6187	46000	0.0805	-
0.6254	46500	0.0788	-
0.6321	47000	0.0724	-
0.6389	47500	0.0833	-
0.6456	48000	0.0752	0.0972
0.6523	48500	0.0733	-
0.6590	49000	0.0686	-
0.6658	49500	0.0802	-
0.6725	50000	0.0817	-
0.6792	50500	0.0772	-
0.6859	51000	0.0746	0.0958
0.6927	51500	0.0742	-
0.6994	52000	0.0732	-
0.7061	52500	0.0711	-
0.7128	53000	0.0773	-
0.7196	53500	0.0782	-
0.7263	54000	0.0774	0.0953
0.7330	54500	0.0788	-
0.7397	55000	0.0667	-
0.7465	55500	0.0721	-
0.7532	56000	0.074	-
0.7599	56500	0.0698	-
0.7666	57000	0.0703	0.0948
0.7734	57500	0.0718	-
0.7801	58000	0.0764	-
0.7868	58500	0.078	-
0.7935	59000	0.0784	-
0.8003	59500	0.0771	-
0.8070	60000	0.0766	0.0937
0.8137	60500	0.0758	-
0.8204	61000	0.0747	-
0.8272	61500	0.0814	-
0.8339	62000	0.0719	-
0.8406	62500	0.067	-
0.8473	63000	0.0717	0.0937
0.8541	63500	0.0732	-
0.8608	64000	0.0755	-
0.8675	64500	0.0749	-
0.8742	65000	0.072	-
0.8810	65500	0.071	-
0.8877	66000	0.0702	0.0923
0.8944	66500	0.0676	-
0.9011	67000	0.0753	-
0.9079	67500	0.0734	-
0.9146	68000	0.0654	-
0.9213	68500	0.073	-
0.9280	69000	0.0703	0.0922
0.9348	69500	0.07	-
0.9415	70000	0.0716	-
0.9482	70500	0.0811	-
0.9549	71000	0.0722	-
0.9617	71500	0.0697	-
0.9684	72000	0.0746	0.0915
0.9751	72500	0.0768	-
0.9818	73000	0.0691	-
0.9886	73500	0.0718	-
0.9953	74000	0.0707	-

Framework Versions

• Python: 3.10.16
• Sentence Transformers: 5.1.1
• Transformers: 4.56.2
• PyTorch: 2.8.0+cu128
• Accelerate: 1.10.1
• Datasets: 4.1.1
• Tokenizers: 0.22.1

Citation

BibTeX

Sentence Transformers

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}