SentenceTransformer based on Snowflake/snowflake-arctic-embed-m-v1.5

This is a sentence-transformers model finetuned from Snowflake/snowflake-arctic-embed-m-v1.5. 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: Snowflake/snowflake-arctic-embed-m-v1.5
Maximum Sequence Length: 512 tokens
Output Dimensionality: 768 dimensions
Similarity Function: Cosine Similarity

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': 512, 'do_lower_case': False}) with Transformer model: BertModel 
  (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.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("sentence_transformers_model_id")
# Run inference
sentences = [
    'How does the operational efficiency of energy systems impact the integration of renewable energy sources into the grid, and what role does this play in improving overall energy efficiency?',
    '(58) Directive 2010/75/EU of the European Parliament and of the Council (21) lays down rules on installations that contribute to energy production or use energy for production purposes, and provides that information on the energy used in or generated by the installation is to be included in applications for integrated permits in accordance with Article 12(1), point (b) of that Directive. Moreover, Article 11 of that Directive provides that efficient use of energy is one of the general principles governing the basic obligations of the operator and one of the criteria for determining best available techniques pursuant to Annex III to that Directive. The operational efficiency of energy systems at any given moment is influenced by the ability to feed power generated from different sources with different degrees of inertia and start-up times into the grid smoothly and flexibly. Improving efficiency will enable better use to be made of renewable energy.\n\n(59) Improvement in energy efficiency can contribute to higher economic output. Member States and the Union should aim to decrease energy consumption regardless of levels of economic growth.\n\n(60) The energy savings obligation established by this Directive should be increased and should also apply after 2030. That ensures stability for investors and thus encourages long-term investments and long-term energy efficiency measures, such as the deep renovation of buildings with the long-term objective of facilitating the cost effective transformation of existing buildings into nearly zero-energy buildings. The energy savings obligation plays an important role in the creation of local growth, jobs, competitiveness and alleviating energy poverty. It should ensure that the Union can achieve its energy and climate objectives by creating further opportunities and by breaking the link between energy consumption and growth. Cooperation with the private sector is important to assess the conditions on which private investment for energy efficiency projects can be unlocked and to develop new revenue models for innovation in the field of energy efficiency.',
    'where:\n\n‘distance’ means the great circle distance between the aerodrome of departure and the aerodrome of arrival plus an additional fixed factor of 95 km; and\n\n‘payload’ means the total mass of freight, mail and passengers carried.\n\nFor the purposes of calculating the payload:\n\nthe number of passengers shall be the number of persons on-board excluding crew members,\n\nan aircraft operator may choose to apply either the actual or standard mass for passengers and checked baggage contained in its mass and balance documentation for the relevant flights or a default value of 100 kg for each passenger and his checked baggage.\n\nReporting of tonne-kilometre data for the purpose of Articles 3e and 3f',
]
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

Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.811
cosine_accuracy@3	0.943
cosine_accuracy@5	0.9704
cosine_accuracy@10	0.9849
cosine_precision@1	0.811
cosine_precision@3	0.3143
cosine_precision@5	0.1941
cosine_precision@10	0.0985
cosine_recall@1	0.811
cosine_recall@3	0.943
cosine_recall@5	0.9704
cosine_recall@10	0.9849
cosine_ndcg@10	0.9068
cosine_mrr@10	0.8808
cosine_map@100	0.8816

Training Details

Training Dataset

Unnamed Dataset

Size: 32,897 training samples
Columns: sentence_0 and sentence_1
Approximate statistics based on the first 1000 samples:
sentence_0 sentence_1
type string string
details
min: 12 tokens
mean: 41.21 tokens
max: 184 tokens

min: 3 tokens
mean: 234.18 tokens
max: 512 tokens

	sentence_0	sentence_1
type	string	string
details	min: 12 tokens mean: 41.21 tokens max: 184 tokens	min: 3 tokens mean: 234.18 tokens max: 512 tokens

Samples:

sentence_0	sentence_1
`What is the maximum allowable concentration of Dimethylfumarate in articles or parts thereof, and what are the implications for market placement if this concentration is exceeded?`	"32011R0366: INSERTED") 60. Acrylamide CAS No 79-06-1 Shall not be placed on the market or used as a substance or constituent of mixtures in a concentration, equal to or greater than 0,1 % by weight for grouting applications after 5 November 2012. ▼M16 61. Dimethylfumarate (DMF) CAS No 624-49-7 EC 210-849-0 Shall not be used in articles or any parts thereof in concentrations greater than 0,1 mg/kg. Articles or any parts thereof containing DMF in concentrations greater than 0,1 mg/kg shall not be placed on the market. ▼M20 62. (a) Phenylmercury acetate EC No: 200-532-5 CAS No:
`The passage highlights the importance of ensuring that hygiene products made partly or wholly from plastics remain safe by avoiding harmful chemical substances. By calling for the assessment of additional restrictions under chemical regulations, it underscores a precautionary stance rooted in safeguarding consumer well-being, particularly for products intimately connected to human health.`	`(19)`
`What responsibilities do Member States have regarding penalties for infringements of national provisions adopted under the directive?`	`Member States shall lay down the rules on penalties applicable to infringements of national provisions adopted pursuant to this Directive and shall take all measures necessary to ensure that they are implemented. The penalties provided for shall be effective, proportionate and dissuasive. Member States shall, by 3 July 2021, notify the Commission of those rules and those measures and shall notify it of any subsequent amendment affecting them.`

Loss: MatryoshkaLoss with these parameters:

{
    "loss": "MultipleNegativesRankingLoss",
    "matryoshka_dims": [
        768,
        512,
        256,
        128,
        64
    ],
    "matryoshka_weights": [
        1,
        1,
        1,
        1,
        1
    ],
    "n_dims_per_step": -1
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 4
per_device_eval_batch_size: 4
num_train_epochs: 4
multi_dataset_batch_sampler: round_robin

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: 5e-05
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1
num_train_epochs: 4
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.0
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: 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}
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: round_robin

Training Logs