SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: NewModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Normalize()
)

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.

python

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("pj-mathematician/JobGTE-multilingual-base-v1")
# Run inference
sentences = [
    'Volksvertreter',
    'Parlamentarier',
    'Oberbürgermeister',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Information Retrieval

Datasets: full_en, full_es, full_de, full_zh, mix_es, mix_de and mix_zh
Evaluated with InformationRetrievalEvaluator

Metric	full_en	full_es	full_de	full_zh	mix_es	mix_de	mix_zh
cosine_accuracy@1	0.6571	0.1243	0.2956	0.6602	0.728	0.6703	0.1908
cosine_accuracy@20	0.9905	1.0	0.9704	0.9806	0.96	0.9506	1.0
cosine_accuracy@50	0.9905	1.0	0.9852	0.9903	0.9792	0.9776	1.0
cosine_accuracy@100	0.9905	1.0	0.9852	0.9903	0.9943	0.9865	1.0
cosine_accuracy@150	0.9905	1.0	0.9901	0.9903	0.9958	0.9932	1.0
cosine_accuracy@200	0.9905	1.0	0.9901	0.9903	0.9974	0.9948	1.0
cosine_precision@1	0.6571	0.1243	0.2956	0.6602	0.728	0.6703	0.1908
cosine_precision@20	0.5171	0.5719	0.5084	0.4782	0.1243	0.1252	0.1544
cosine_precision@50	0.316	0.3885	0.3654	0.2895	0.0515	0.0523	0.0618
cosine_precision@100	0.189	0.2517	0.2413	0.1757	0.0263	0.0267	0.0309
cosine_precision@150	0.1338	0.1905	0.1804	0.126	0.0176	0.018	0.0206
cosine_precision@200	0.1043	0.1522	0.1447	0.0982	0.0133	0.0135	0.0154
cosine_recall@1	0.0678	0.0037	0.0111	0.0615	0.2813	0.2524	0.0614
cosine_recall@20	0.547	0.3842	0.3221	0.5108	0.9183	0.9096	1.0
cosine_recall@50	0.74	0.5641	0.5025	0.6923	0.9499	0.9482	1.0
cosine_recall@100	0.8453	0.6742	0.6248	0.8004	0.9701	0.9685	1.0
cosine_recall@150	0.8838	0.7464	0.683	0.8465	0.9768	0.9782	1.0
cosine_recall@200	0.9109	0.7825	0.7216	0.8771	0.9818	0.981	1.0
cosine_ndcg@1	0.6571	0.1243	0.2956	0.6602	0.728	0.6703	0.1908
cosine_ndcg@20	0.6954	0.6139	0.5393	0.654	0.8044	0.7736	0.5474
cosine_ndcg@50	0.715	0.5874	0.5267	0.6707	0.813	0.7844	0.5474
cosine_ndcg@100	0.7679	0.6144	0.5579	0.7234	0.8173	0.7889	0.5474
cosine_ndcg@150	0.7857	0.6499	0.588	0.7438	0.8186	0.7909	0.5474
cosine_ndcg@200	0.797	0.6681	0.6071	0.7554	0.8195	0.7914	0.5474
cosine_mrr@1	0.6571	0.1243	0.2956	0.6602	0.728	0.6703	0.1908
cosine_mrr@20	0.8138	0.5581	0.5104	0.8037	0.7969	0.752	0.4093
cosine_mrr@50	0.8138	0.5581	0.511	0.8041	0.7975	0.7529	0.4093
cosine_mrr@100	0.8138	0.5581	0.511	0.8041	0.7977	0.7531	0.4093
cosine_mrr@150	0.8138	0.5581	0.511	0.8041	0.7977	0.7531	0.4093
cosine_mrr@200	0.8138	0.5581	0.511	0.8041	0.7977	0.7531	0.4093
cosine_map@1	0.6571	0.1243	0.2956	0.6602	0.728	0.6703	0.1908
cosine_map@20	0.5579	0.4799	0.401	0.5087	0.7351	0.6968	0.3298
cosine_map@50	0.5471	0.425	0.3588	0.4926	0.7374	0.6996	0.3298
cosine_map@100	0.5796	0.4302	0.3633	0.5217	0.738	0.7003	0.3298
cosine_map@150	0.5875	0.4459	0.3777	0.5299	0.7381	0.7004	0.3298
cosine_map@200	0.5912	0.4533	0.3848	0.5334	0.7382	0.7005	0.3298
cosine_map@500	0.5953	0.4656	0.3978	0.5386	0.7383	0.7006	0.3298

Training Details

Training Datasets

full_en

Dataset: full_en
Size: 28,880 training samples
Columns: anchor and positive
Approximate statistics based on the first 1000 samples:
anchor positive
type string string
details
min: 3 tokens
mean: 5.68 tokens
max: 11 tokens
min: 3 tokens
mean: 5.76 tokens
max: 12 tokens
Samples:
anchor positive
air commodore flight lieutenant
command and control officer flight officer
air commodore command and control officer
Loss: GISTEmbedLoss with these parameters:
json
1{'guide': SentenceTransformer( 2 (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: BertModel 3 (1): Pooling({'word_embedding_dimension': 384, '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}) 4 (2): Normalize() 5), 'temperature': 0.01, 'margin_strategy': 'absolute', 'margin': 0.0}

full_de

Dataset: full_de
Size: 23,023 training samples
Columns: anchor and positive
Approximate statistics based on the first 1000 samples:
anchor positive
type string string
details
min: 3 tokens
mean: 7.99 tokens
max: 30 tokens
min: 3 tokens
mean: 8.19 tokens
max: 30 tokens
Samples:
anchor positive
Staffelkommandantin Kommodore
Luftwaffenoffizierin Luftwaffenoffizier/Luftwaffenoffizierin
Staffelkommandantin Luftwaffenoffizierin
Loss: GISTEmbedLoss with these parameters:
json
1{'guide': SentenceTransformer( 2 (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: BertModel 3 (1): Pooling({'word_embedding_dimension': 384, '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}) 4 (2): Normalize() 5), 'temperature': 0.01, 'margin_strategy': 'absolute', 'margin': 0.0}

full_es

Dataset: full_es
Size: 20,724 training samples
Columns: anchor and positive
Approximate statistics based on the first 1000 samples:
anchor positive
type string string
details
min: 3 tokens
mean: 9.13 tokens
max: 32 tokens
min: 3 tokens
mean: 8.84 tokens
max: 32 tokens
Samples:
anchor positive
jefe de escuadrón instructor
comandante de aeronave instructor de simulador
instructor oficial del Ejército del Aire
Loss: GISTEmbedLoss with these parameters:
json
1{'guide': SentenceTransformer( 2 (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: BertModel 3 (1): Pooling({'word_embedding_dimension': 384, '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}) 4 (2): Normalize() 5), 'temperature': 0.01, 'margin_strategy': 'absolute', 'margin': 0.0}

full_zh

Dataset: full_zh
Size: 30,401 training samples
Columns: anchor and positive
Approximate statistics based on the first 1000 samples:
anchor positive
type string string
details
min: 5 tokens
mean: 7.15 tokens
max: 14 tokens
min: 5 tokens
mean: 7.46 tokens
max: 21 tokens
Samples:
anchor positive
技术总监 技术和运营总监
技术总监 技术主管
技术总监 技术艺术总监
Loss: GISTEmbedLoss with these parameters:
json
1{'guide': SentenceTransformer( 2 (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: BertModel 3 (1): Pooling({'word_embedding_dimension': 384, '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}) 4 (2): Normalize() 5), 'temperature': 0.01, 'margin_strategy': 'absolute', 'margin': 0.0}

mix

Dataset: mix
Size: 21,760 training samples
Columns: anchor and positive
Approximate statistics based on the first 1000 samples:
anchor positive
type string string
details
min: 2 tokens
mean: 6.71 tokens
max: 19 tokens
min: 2 tokens
mean: 7.69 tokens
max: 19 tokens
Samples:
anchor positive
technical manager Technischer Direktor für Bühne, Film und Fernsehen
head of technical directora técnica
head of technical department 技术艺术总监
Loss: GISTEmbedLoss with these parameters:
json
1{'guide': SentenceTransformer( 2 (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: BertModel 3 (1): Pooling({'word_embedding_dimension': 384, '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}) 4 (2): Normalize() 5), 'temperature': 0.01, 'margin_strategy': 'absolute', 'margin': 0.0}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 64
per_device_eval_batch_size: 128
gradient_accumulation_steps: 2
num_train_epochs: 5
warmup_ratio: 0.05
log_on_each_node: False
fp16: True
dataloader_num_workers: 4
ddp_find_unused_parameters: True
batch_sampler: no_duplicates

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: steps
prediction_loss_only: True
per_device_train_batch_size: 64
per_device_eval_batch_size: 128
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 2
eval_accumulation_steps: None
torch_empty_cache_steps: None
learning_rate: 5e-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: 5
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.05
warmup_steps: 0
log_level: passive
log_level_replica: warning
log_on_each_node: False
logging_nan_inf_filter: True
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: False
fp16: True
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: True
dataloader_num_workers: 4
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}
tp_size: 0
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}
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch
optim_args: None
adafactor: False
group_by_length: False
length_column_name: length
ddp_find_unused_parameters: True
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
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
eval_use_gather_object: False
average_tokens_across_devices: False
prompts: None
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	full_en_cosine_ndcg@200	full_es_cosine_ndcg@200	full_de_cosine_ndcg@200	full_zh_cosine_ndcg@200	mix_es_cosine_ndcg@200	mix_de_cosine_ndcg@200	mix_zh_cosine_ndcg@200
-1	-1	-	0.7447	0.6125	0.5378	0.7240	0.7029	0.6345	0.5531
0.0010	1	3.4866	-	-	-	-	-	-	-
0.1027	100	2.5431	-	-	-	-	-	-	-
0.2053	200	1.4536	0.7993	0.6633	0.5974	0.7642	0.7567	0.7011	0.5498
0.3080	300	1.1018	-	-	-	-	-	-	-
0.4107	400	0.9184	0.7925	0.6586	0.6058	0.7587	0.7749	0.7278	0.5486
0.5133	500	0.8973	-	-	-	-	-	-	-
0.6160	600	0.7309	0.7951	0.6671	0.6096	0.7708	0.7793	0.7339	0.5525
0.7187	700	0.7297	-	-	-	-	-	-	-
0.8214	800	0.7281	0.7929	0.6711	0.6088	0.7645	0.7899	0.7444	0.5479
0.9240	900	0.6607	-	-	-	-	-	-	-
1.0267	1000	0.6075	0.7915	0.6659	0.6088	0.7665	0.7968	0.7588	0.5482
1.1294	1100	0.4553	-	-	-	-	-	-	-
1.2320	1200	0.4775	0.7979	0.6696	0.6033	0.7669	0.7959	0.7624	0.5484
1.3347	1300	0.4838	-	-	-	-	-	-	-
1.4374	1400	0.4912	0.7973	0.6757	0.6112	0.7656	0.7978	0.7650	0.5487
1.5400	1500	0.4732	-	-	-	-	-	-	-
1.6427	1600	0.5269	0.8031	0.6723	0.6108	0.7654	0.8008	0.7660	0.5492
1.7454	1700	0.4822	-	-	-	-	-	-	-
1.8480	1800	0.5072	0.7962	0.6668	0.6051	0.7592	0.8001	0.7714	0.5486
1.9507	1900	0.4709	-	-	-	-	-	-	-
2.0544	2000	0.3772	0.7940	0.6647	0.6037	0.7579	0.8064	0.7732	0.5479
2.1571	2100	0.3982	-	-	-	-	-	-	-
2.2598	2200	0.3073	0.7969	0.6652	0.6005	0.7625	0.8054	0.7734	0.5493
2.3624	2300	0.383	-	-	-	-	-	-	-
2.4651	2400	0.3687	0.7925	0.6690	0.5987	0.7583	0.8081	0.7735	0.5477
2.5678	2500	0.3472	-	-	-	-	-	-	-
2.6704	2600	0.3557	0.7956	0.6758	0.6019	0.7659	0.8082	0.7767	0.5491
2.7731	2700	0.3527	-	-	-	-	-	-	-
2.8758	2800	0.3446	0.7945	0.6719	0.6020	0.7616	0.8124	0.7818	0.5496
2.9784	2900	0.3566	-	-	-	-	-	-	-
3.0821	3000	0.3252	0.7948	0.6682	0.6025	0.7617	0.8152	0.7848	0.5516
3.1848	3100	0.2968	-	-	-	-	-	-	-
3.2875	3200	0.2962	0.7953	0.6717	0.6086	0.7613	0.8110	0.7824	0.5482
3.3901	3300	0.3084	-	-	-	-	-	-	-
3.4928	3400	0.2909	0.7940	0.6634	0.6023	0.7615	0.8138	0.7822	0.5457
3.5955	3500	0.2964	-	-	-	-	-	-	-
3.6982	3600	0.3193	0.7960	0.6635	0.6070	0.7534	0.8164	0.7844	0.5467
3.8008	3700	0.3514	-	-	-	-	-	-	-
3.9035	3800	0.3147	0.7973	0.6696	0.6125	0.7616	0.8176	0.7885	0.5469
4.0062	3900	0.2738	-	-	-	-	-	-	-
4.1088	4000	0.2842	0.7960	0.6672	0.6082	0.7536	0.8174	0.7891	0.5479
4.2115	4100	0.2739	-	-	-	-	-	-	-
4.3142	4200	0.2704	0.7979	0.6681	0.6111	0.7540	0.8180	0.7891	0.5476
4.4168	4300	0.2529	-	-	-	-	-	-	-
4.5195	4400	0.272	0.7968	0.6685	0.6087	0.7564	0.8185	0.7901	0.5476
4.6222	4500	0.3	-	-	-	-	-	-	-
4.7248	4600	0.2598	0.7972	0.6675	0.6072	0.7556	0.8190	0.7909	0.5478
4.8275	4700	0.3101	-	-	-	-	-	-	-
4.9302	4800	0.2524	0.7970	0.6681	0.6071	0.7554	0.8195	0.7914	0.5474

Framework Versions

Python: 3.11.11
Sentence Transformers: 4.1.0
Transformers: 4.51.2
PyTorch: 2.6.0+cu124
Accelerate: 1.6.0
Datasets: 3.5.0
Tokenizers: 0.21.1

Citation

BibTeX

Sentence Transformers

bibtex

@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",
}

GISTEmbedLoss

bibtex

@misc{solatorio2024gistembed,
    title={GISTEmbed: Guided In-sample Selection of Training Negatives for Text Embedding Fine-tuning},
    author={Aivin V. Solatorio},
    year={2024},
    eprint={2402.16829},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

	anchor	positive
type	string	string
details	min: 3 tokens mean: 5.68 tokens max: 11 tokens	min: 3 tokens mean: 5.76 tokens max: 12 tokens

anchor	positive
`air commodore`	`flight lieutenant`
`command and control officer`	`flight officer`
`air commodore`	`command and control officer`

anchor	positive
`Staffelkommandantin`	`Kommodore`
`Luftwaffenoffizierin`	`Luftwaffenoffizier/Luftwaffenoffizierin`
`Staffelkommandantin`	`Luftwaffenoffizierin`

anchor	positive
`jefe de escuadrón`	`instructor`
`comandante de aeronave`	`instructor de simulador`
`instructor`	`oficial del Ejército del Aire`

anchor	positive
`技术总监`	`技术和运营总监`
`技术总监`	`技术主管`
`技术总监`	`技术艺术总监`

anchor	positive
`technical manager`	`Technischer Direktor für Bühne, Film und Fernsehen`
`head of technical`	`directora técnica`
`head of technical department`	`技术艺术总监`