coffeepowered/arctic-l-enhanced

SentenceTransformer based on Snowflake/snowflake-arctic-embed-l-v2.0

This is a sentence-transformers model finetuned from Snowflake/snowflake-arctic-embed-l-v2.0 on the json dataset. It maps sentences & paragraphs to a 1024-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: Snowflake/snowflake-arctic-embed-l-v2.0
• Maximum Sequence Length: 8192 tokens
• Output Dimensionality: 1024 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • json

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}) with Transformer model: XLMRobertaModel 
  (1): Pooling({'word_embedding_dimension': 1024, '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.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("coffeepowered/arctic-l-enhanced")
# Run inference
sentences = [
    'The weather is lovely today.',
    "It's so sunny outside!",
    'He drove to the stadium.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

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

Training Details

Training Dataset

json

• Dataset: json
• Columns: sentence1, sentence2, score, and split
• Loss: CosineSimilarityLoss with these parameters:

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

Evaluation Dataset

json

• Dataset: json
• Columns: sentence1, sentence2, score, and split
• Loss: CosineSimilarityLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• learning_rate: 4.000000000000001e-06
• max_steps: 13938
• warmup_ratio: 0.1
• fp16: True
• load_best_model_at_end: 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: 8
• per_device_eval_batch_size: 8
• 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: 4.000000000000001e-06
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 3
• max_steps: 13938
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.1
• warmup_steps: 0
• log_level: passive
• log_level_replica: warning
• log_on_each_node: True
• 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: 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: True
• 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}
• 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: 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
• 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
• dispatch_batches: None
• split_batches: 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: batch_sampler
• multi_dataset_batch_sampler: proportional

Training Logs

Click to expand

Epoch	Step	Training Loss	Validation Loss
0.0007	10	0.1359	-
0.0014	20	0.1202	-
0.0022	30	0.1314	-
0.0029	40	0.1302	-
0.0036	50	0.1158	-
0.0043	60	0.1158	-
0.0050	70	0.1114	-
0.0057	80	0.1316	-
0.0065	90	0.133	-
0.0072	100	0.0958	-
0.0079	110	0.0973	-
0.0086	120	0.0949	-
0.0093	130	0.0892	-
0.0100	140	0.0867	-
0.0108	150	0.0717	0.0712
0.0115	160	0.0762	-
0.0122	170	0.0828	-
0.0129	180	0.0775	-
0.0136	190	0.0614	-
0.0143	200	0.0748	-
0.0151	210	0.0545	-
0.0158	220	0.0725	-
0.0165	230	0.0627	-
0.0172	240	0.0612	-
0.0179	250	0.0508	-
0.0187	260	0.0592	-
0.0194	270	0.0489	-
0.0201	280	0.0545	-
0.0208	290	0.0598	-
0.0215	300	0.0641	0.0566
0.0222	310	0.0554	-
0.0230	320	0.0523	-
0.0237	330	0.0568	-
0.0244	340	0.0594	-
0.0251	350	0.0488	-
0.0258	360	0.06	-
0.0265	370	0.0737	-
0.0273	380	0.0565	-
0.0280	390	0.0456	-
0.0287	400	0.0489	-
0.0294	410	0.0478	-
0.0301	420	0.0481	-
0.0309	430	0.0541	-
0.0316	440	0.0581	-
0.0323	450	0.0549	0.0578
0.0330	460	0.042	-
0.0337	470	0.0451	-
0.0344	480	0.0537	-
0.0352	490	0.0487	-
0.0359	500	0.0444	-
0.0366	510	0.0443	-
0.0373	520	0.0418	-
0.0380	530	0.0447	-
0.0387	540	0.0453	-
0.0395	550	0.0465	-
0.0402	560	0.0623	-
0.0409	570	0.0507	-
0.0416	580	0.0428	-
0.0423	590	0.0494	-
0.0430	600	0.0407	0.0519
0.0438	610	0.0452	-
0.0445	620	0.0439	-
0.0452	630	0.0534	-
0.0459	640	0.0422	-
0.0466	650	0.0514	-
0.0474	660	0.0406	-
0.0481	670	0.0428	-
0.0488	680	0.0346	-
0.0495	690	0.0485	-
0.0502	700	0.0549	-
0.0509	710	0.0399	-
0.0517	720	0.0362	-
0.0524	730	0.0507	-
0.0531	740	0.0434	-
0.0538	750	0.0363	0.0553
0.0545	760	0.0426	-
0.0552	770	0.0375	-
0.0560	780	0.0501	-
0.0567	790	0.043	-
0.0574	800	0.0476	-
0.0581	810	0.037	-
0.0588	820	0.0317	-
0.0595	830	0.0387	-
0.0603	840	0.0348	-
0.0610	850	0.0379	-
0.0617	860	0.0506	-
0.0624	870	0.0419	-
0.0631	880	0.0431	-
0.0639	890	0.0478	-
0.0646	900	0.0394	0.0571
0.0653	910	0.0334	-
0.0660	920	0.0384	-
0.0667	930	0.0381	-
0.0674	940	0.0343	-
0.0682	950	0.0355	-
0.0689	960	0.0384	-
0.0696	970	0.0352	-
0.0703	980	0.0479	-
0.0710	990	0.0362	-
0.0717	1000	0.0357	-
0.0725	1010	0.0414	-
0.0732	1020	0.0346	-
0.0739	1030	0.039	-
0.0746	1040	0.0317	-
0.0753	1050	0.0502	0.0523
0.0761	1060	0.0407	-
0.0768	1070	0.0377	-
0.0775	1080	0.0404	-
0.0782	1090	0.0297	-
0.0789	1100	0.0344	-
0.0796	1110	0.0277	-
0.0804	1120	0.0453	-
0.0811	1130	0.0469	-
0.0818	1140	0.044	-
0.0825	1150	0.0339	-
0.0832	1160	0.0331	-
0.0839	1170	0.0327	-
0.0847	1180	0.0445	-
0.0854	1190	0.0392	-
0.0861	1200	0.048	0.0531
0.0868	1210	0.0375	-
0.0875	1220	0.0311	-
0.0882	1230	0.037	-
0.0890	1240	0.0369	-
0.0897	1250	0.0398	-
0.0904	1260	0.0363	-
0.0911	1270	0.0425	-
0.0918	1280	0.0355	-
0.0926	1290	0.0362	-
0.0933	1300	0.0354	-
0.0940	1310	0.0252	-
0.0947	1320	0.03	-
0.0954	1330	0.03	-
0.0961	1340	0.0218	-
0.0969	1350	0.0248	0.0448
0.0976	1360	0.0287	-
0.0983	1370	0.0322	-
0.0990	1380	0.0359	-
0.0997	1390	0.0287	-
0.1004	1400	0.034	-
0.1012	1410	0.0305	-
0.1019	1420	0.0241	-
0.1026	1430	0.0227	-
0.1033	1440	0.0279	-
0.1040	1450	0.0298	-
0.1047	1460	0.0382	-
0.1055	1470	0.0269	-
0.1062	1480	0.0243	-
0.1069	1490	0.0266	-
0.1076	1500	0.016	0.0456
0.1083	1510	0.0233	-
0.1091	1520	0.0247	-
0.1098	1530	0.0233	-
0.1105	1540	0.0214	-
0.1112	1550	0.0229	-
0.1119	1560	0.0145	-
0.1126	1570	0.0187	-
0.1134	1580	0.0231	-
0.1141	1590	0.0268	-
0.1148	1600	0.027	-
0.1155	1610	0.0354	-
0.1162	1620	0.0398	-
0.1169	1630	0.0431	-
0.1177	1640	0.0391	-
0.1184	1650	0.056	0.0422
0.1191	1660	0.0394	-
0.1198	1670	0.0338	-
0.1205	1680	0.0417	-
0.1213	1690	0.0479	-
0.1220	1700	0.0348	-
0.1227	1710	0.0366	-
0.1234	1720	0.0335	-
0.1241	1730	0.0367	-
0.1248	1740	0.0355	-
0.1256	1750	0.0319	-
0.1263	1760	0.0324	-
0.1270	1770	0.0345	-
0.1277	1780	0.0352	-
0.1284	1790	0.0362	-
0.1291	1800	0.0319	0.0388
0.1299	1810	0.0381	-
0.1306	1820	0.0368	-
0.1313	1830	0.0318	-
0.1320	1840	0.0296	-
0.1327	1850	0.0331	-
0.1334	1860	0.0381	-
0.1342	1870	0.0237	-
0.1349	1880	0.0284	-
0.1356	1890	0.0353	-
0.1363	1900	0.0408	-
0.1370	1910	0.0277	-
0.1378	1920	0.0268	-
0.1385	1930	0.0348	-
0.1392	1940	0.0348	-
0.1399	1950	0.0269	0.0351
0.1406	1960	0.0317	-
0.1413	1970	0.0375	-
0.1421	1980	0.0321	-
0.1428	1990	0.0302	-
0.1435	2000	0.0411	-
0.1442	2010	0.0329	-
0.1449	2020	0.0246	-
0.1456	2030	0.0288	-
0.1464	2040	0.0297	-
0.1471	2050	0.0285	-
0.1478	2060	0.0326	-
0.1485	2070	0.035	-
0.1492	2080	0.0208	-
0.1499	2090	0.0209	-
0.1507	2100	0.0303	0.0364
0.1514	2110	0.028	-
0.1521	2120	0.0366	-
0.1528	2130	0.033	-
0.1535	2140	0.0353	-
0.1543	2150	0.0292	-
0.1550	2160	0.0202	-
0.1557	2170	0.0233	-
0.1564	2180	0.0346	-
0.1571	2190	0.0309	-
0.1578	2200	0.0241	-
0.1586	2210	0.0317	-
0.1593	2220	0.0331	-
0.1600	2230	0.0273	-
0.1607	2240	0.0219	-
0.1614	2250	0.0256	0.0338
0.1621	2260	0.024	-
0.1629	2270	0.0216	-
0.1636	2280	0.0253	-
0.1643	2290	0.0271	-
0.1650	2300	0.0336	-
0.1657	2310	0.0273	-
0.1665	2320	0.0267	-
0.1672	2330	0.0292	-
0.1679	2340	0.0269	-
0.1686	2350	0.0342	-
0.1693	2360	0.0361	-
0.1700	2370	0.0224	-
0.1708	2380	0.016	-
0.1715	2390	0.0209	-
0.1722	2400	0.0162	0.0329
0.1729	2410	0.019	-
0.1736	2420	0.0181	-
0.1743	2430	0.0332	-
0.1751	2440	0.0189	-
0.1758	2450	0.0292	-
0.1765	2460	0.0193	-
0.1772	2470	0.024	-
0.1779	2480	0.0245	-
0.1786	2490	0.02	-
0.1794	2500	0.0187	-
0.1801	2510	0.0224	-
0.1808	2520	0.028	-
0.1815	2530	0.0219	-
0.1822	2540	0.0212	-
0.1830	2550	0.0208	0.0354
0.1837	2560	0.0299	-
0.1844	2570	0.0208	-
0.1851	2580	0.0271	-
0.1858	2590	0.0146	-
0.1865	2600	0.0163	-
0.1873	2610	0.0149	-
0.1880	2620	0.0145	-
0.1887	2630	0.0196	-
0.1894	2640	0.0252	-
0.1901	2650	0.0173	-
0.1908	2660	0.0202	-
0.1916	2670	0.0171	-
0.1923	2680	0.023	-
0.1930	2690	0.02	-
0.1937	2700	0.0171	0.0387
0.1944	2710	0.0239	-
0.1951	2720	0.0261	-
0.1959	2730	0.0197	-
0.1966	2740	0.0195	-
0.1973	2750	0.0132	-
0.1980	2760	0.0226	-
0.1987	2770	0.0172	-
0.1995	2780	0.0214	-
0.2002	2790	0.0173	-
0.2009	2800	0.0179	-
0.2016	2810	0.0184	-
0.2023	2820	0.0162	-
0.2030	2830	0.0126	-
0.2038	2840	0.0188	-
0.2045	2850	0.0177	0.0385
0.2052	2860	0.0153	-
0.2059	2870	0.0137	-
0.2066	2880	0.0174	-
0.2073	2890	0.0148	-
0.2081	2900	0.0194	-
0.2088	2910	0.0133	-
0.2095	2920	0.0152	-
0.2102	2930	0.0179	-
0.2109	2940	0.0187	-
0.2117	2950	0.0161	-
0.2124	2960	0.0185	-
0.2131	2970	0.0151	-
0.2138	2980	0.0194	-
0.2145	2990	0.0179	-
0.2152	3000	0.0115	0.0434
0.2160	3010	0.0183	-
0.2167	3020	0.0168	-
0.2174	3030	0.0137	-
0.2181	3040	0.0196	-
0.2188	3050	0.0145	-
0.2195	3060	0.0164	-
0.2203	3070	0.0132	-
0.2210	3080	0.0133	-
0.2217	3090	0.0154	-
0.2224	3100	0.0145	-
0.2231	3110	0.0135	-
0.2238	3120	0.0143	-
0.2246	3130	0.0197	-
0.2253	3140	0.0163	-
0.2260	3150	0.017	0.0409
0.2267	3160	0.0257	-
0.2274	3170	0.0165	-
0.2282	3180	0.0189	-
0.2289	3190	0.0207	-
0.2296	3200	0.0179	-
0.2303	3210	0.0152	-
0.2310	3220	0.0179	-
0.2317	3230	0.0187	-
0.2325	3240	0.0189	-
0.2332	3250	0.0151	-
0.2339	3260	0.0137	-
0.2346	3270	0.0122	-
0.2353	3280	0.0134	-
0.2360	3290	0.0134	-
0.2368	3300	0.0162	0.0455
0.2375	3310	0.0142	-
0.2382	3320	0.0203	-
0.2389	3330	0.0148	-
0.2396	3340	0.0181	-
0.2404	3350	0.0178	-
0.2411	3360	0.018	-
0.2418	3370	0.012	-
0.2425	3380	0.0201	-
0.2432	3390	0.0193	-
0.2439	3400	0.013	-
0.2447	3410	0.0114	-
0.2454	3420	0.0184	-
0.2461	3430	0.0153	-
0.2468	3440	0.0172	-
0.2475	3450	0.0141	0.0458
0.2482	3460	0.0165	-
0.2490	3470	0.0174	-
0.2497	3480	0.0109	-
0.2504	3490	0.0177	-
0.2511	3500	0.0178	-
0.2518	3510	0.0101	-
0.2525	3520	0.0154	-
0.2533	3530	0.0204	-
0.2540	3540	0.0119	-
0.2547	3550	0.0158	-
0.2554	3560	0.0161	-
0.2561	3570	0.0116	-
0.2569	3580	0.0183	-
0.2576	3590	0.0125	-
0.2583	3600	0.0148	0.0475
0.2590	3610	0.0184	-
0.2597	3620	0.0169	-
0.2604	3630	0.0135	-
0.2612	3640	0.0138	-
0.2619	3650	0.0133	-
0.2626	3660	0.0148	-
0.2633	3670	0.0121	-
0.2640	3680	0.0119	-
0.2647	3690	0.0183	-
0.2655	3700	0.0178	-
0.2662	3710	0.0139	-
0.2669	3720	0.0171	-
0.2676	3730	0.0155	-
0.2683	3740	0.0173	-
0.2690	3750	0.0158	0.0488
0.2698	3760	0.0125	-
0.2705	3770	0.0137	-
0.2712	3780	0.0184	-
0.2719	3790	0.0117	-
0.2726	3800	0.0189	-
0.2734	3810	0.0163	-
0.2741	3820	0.0163	-
0.2748	3830	0.0212	-
0.2755	3840	0.0133	-
0.2762	3850	0.0168	-
0.2769	3860	0.0197	-
0.2777	3870	0.0184	-
0.2784	3880	0.017	-
0.2791	3890	0.0173	-
0.2798	3900	0.0165	0.0511

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.14
• Sentence Transformers: 3.4.1
• Transformers: 4.49.0
• PyTorch: 2.2.2
• Accelerate: 1.4.0
• Datasets: 3.3.2
• Tokenizers: 0.21.0

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",
}