Speedsy/pylate-49500

PyLate model based on microsoft/Multilingual-MiniLM-L12-H384

This is a PyLate model finetuned from microsoft/Multilingual-MiniLM-L12-H384 on the train dataset. It maps sentences & paragraphs to sequences of 128-dimensional dense vectors and can be used for semantic textual similarity using the MaxSim operator.

Model Details

Model Description

• Model Type: PyLate model
• Base model: microsoft/Multilingual-MiniLM-L12-H384
• Document Length: 180 tokens
• Query Length: 32 tokens
• Output Dimensionality: 128 tokens
• Similarity Function: MaxSim
• Training Dataset:
  • train
• Language: en

Model Sources

• Documentation: PyLate Documentation
• Repository: PyLate on GitHub
• Hugging Face: PyLate models on Hugging Face

Full Model Architecture

ColBERT(
  (0): Transformer({'max_seq_length': 179, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Dense({'in_features': 384, 'out_features': 128, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity'})
)

Usage

First install the PyLate library:

pip install -U pylate

Retrieval

PyLate provides a streamlined interface to index and retrieve documents using ColBERT models. The index leverages the Voyager HNSW index to efficiently handle document embeddings and enable fast retrieval.

Indexing documents

First, load the ColBERT model and initialize the Voyager index, then encode and index your documents:

from pylate import indexes, models, retrieve

# Step 1: Load the ColBERT model
model = models.ColBERT(
    model_name_or_path=pylate_model_id,
)

# Step 2: Initialize the Voyager index
index = indexes.Voyager(
    index_folder="pylate-index",
    index_name="index",
    override=True,  # This overwrites the existing index if any
)

# Step 3: Encode the documents
documents_ids = ["1", "2", "3"]
documents = ["document 1 text", "document 2 text", "document 3 text"]

documents_embeddings = model.encode(
    documents,
    batch_size=32,
    is_query=False,  # Ensure that it is set to False to indicate that these are documents, not queries
    show_progress_bar=True,
)

# Step 4: Add document embeddings to the index by providing embeddings and corresponding ids
index.add_documents(
    documents_ids=documents_ids,
    documents_embeddings=documents_embeddings,
)

Note that you do not have to recreate the index and encode the documents every time. Once you have created an index and added the documents, you can re-use the index later by loading it:

# To load an index, simply instantiate it with the correct folder/name and without overriding it
index = indexes.Voyager(
    index_folder="pylate-index",
    index_name="index",
)

Retrieving top-k documents for queries

Once the documents are indexed, you can retrieve the top-k most relevant documents for a given set of queries. To do so, initialize the ColBERT retriever with the index you want to search in, encode the queries and then retrieve the top-k documents to get the top matches ids and relevance scores:

# Step 1: Initialize the ColBERT retriever
retriever = retrieve.ColBERT(index=index)

# Step 2: Encode the queries
queries_embeddings = model.encode(
    ["query for document 3", "query for document 1"],
    batch_size=32,
    is_query=True,  #  # Ensure that it is set to False to indicate that these are queries
    show_progress_bar=True,
)

# Step 3: Retrieve top-k documents
scores = retriever.retrieve(
    queries_embeddings=queries_embeddings,
    k=10,  # Retrieve the top 10 matches for each query
)

Reranking

If you only want to use the ColBERT model to perform reranking on top of your first-stage retrieval pipeline without building an index, you can simply use rank function and pass the queries and documents to rerank:

from pylate import rank, models

queries = [
    "query A",
    "query B",
]

documents = [
    ["document A", "document B"],
    ["document 1", "document C", "document B"],
]

documents_ids = [
    [1, 2],
    [1, 3, 2],
]

model = models.ColBERT(
    model_name_or_path=pylate_model_id,
)

queries_embeddings = model.encode(
    queries,
    is_query=True,
)

documents_embeddings = model.encode(
    documents,
    is_query=False,
)

reranked_documents = rank.rerank(
    documents_ids=documents_ids,
    queries_embeddings=queries_embeddings,
    documents_embeddings=documents_embeddings,
)

Training Details

Training Dataset

train

• Dataset: train at b9b0f7f
• Size: 798,036 training samples
• Columns: query_id, document_ids, and scores
• Approximate statistics based on the first 1000 samples:

  	query_id	document_ids	scores
  type	string	list	list
  details	min: 4 tokens mean: 5.82 tokens max: 6 tokens	size: 32 elements	size: 32 elements

• Samples:

  query_id	document_ids	scores
  817836	['2716076', '6741935', '2681109', '5562684', '3507339', ...]	[1.0, 0.7059561610221863, 0.21702419221401215, 0.38270196318626404, 0.20812414586544037, ...]
  1045170	['5088671', '2953295', '8783471', '4268439', '6339935', ...]	[1.0, 0.6493034362792969, 0.0692221149802208, 0.17963139712810516, 0.6697239875793457, ...]
  1154488	['6498614', '3770829', '1060712', '2590533', '7672044', ...]	[0.9497447609901428, 0.6662212610244751, 0.7423420548439026, 1.0, 0.6580896973609924, ...]

• Loss: pylate.losses.distillation.Distillation

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 16
• learning_rate: 3e-05
• num_train_epochs: 1
• fp16: True

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: no
• prediction_loss_only: True
• per_device_train_batch_size: 16
• 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: 3e-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: 0.0
• 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: 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: 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: proportional

Training Logs

Epoch	Step	Training Loss
0.0100	500	0.0305
0.0200	1000	0.027
0.0301	1500	0.026
0.0401	2000	0.0253
0.0501	2500	0.0249
0.0601	3000	0.0239
0.0702	3500	0.0239
0.0802	4000	0.0236
0.0902	4500	0.0236
0.1002	5000	0.0232
0.1103	5500	0.0229
0.1203	6000	0.0228
0.1303	6500	0.0226
0.1403	7000	0.0226
0.1504	7500	0.0222
0.1604	8000	0.0223
0.1704	8500	0.0218
0.1804	9000	0.0219
0.1905	9500	0.0221
0.2005	10000	0.0219
0.2105	10500	0.0216
0.2205	11000	0.0211
0.2306	11500	0.0214
0.2406	12000	0.0211
0.2506	12500	0.0214
0.2606	13000	0.0212
0.2707	13500	0.0209
0.2807	14000	0.0206
0.2907	14500	0.0208
0.3007	15000	0.0204
0.3108	15500	0.0207
0.3208	16000	0.0204
0.3308	16500	0.0204
0.3408	17000	0.0204
0.3509	17500	0.0203
0.3609	18000	0.0199
0.3709	18500	0.02
0.3809	19000	0.0199
0.3910	19500	0.0197
0.4010	20000	0.0197
0.4110	20500	0.0199
0.4210	21000	0.0198
0.4311	21500	0.0196
0.4411	22000	0.02
0.4511	22500	0.0196
0.4611	23000	0.0196
0.4711	23500	0.0192
0.4812	24000	0.0195
0.4912	24500	0.0197
0.5012	25000	0.0194
0.5112	25500	0.0191
0.5213	26000	0.019
0.5313	26500	0.019
0.5413	27000	0.0192
0.5513	27500	0.0193
0.5614	28000	0.0189
0.5714	28500	0.019
0.5814	29000	0.0189
0.5914	29500	0.0189
0.6015	30000	0.0189
0.6115	30500	0.0187
0.6215	31000	0.0187
0.6315	31500	0.0186
0.6416	32000	0.0187
0.6516	32500	0.0187
0.6616	33000	0.0186
0.6716	33500	0.0185
0.6817	34000	0.0188
0.6917	34500	0.0186
0.7017	35000	0.0185
0.7117	35500	0.0182
0.7218	36000	0.0184
0.7318	36500	0.0183
0.7418	37000	0.0184
0.7518	37500	0.0183
0.7619	38000	0.0182
0.7719	38500	0.0184
0.7819	39000	0.0183
0.7919	39500	0.0184
0.8020	40000	0.0184
0.8120	40500	0.0182
0.8220	41000	0.0182
0.8320	41500	0.0182
0.8421	42000	0.0183
0.8521	42500	0.0183
0.8621	43000	0.0179
0.8721	43500	0.0178
0.8822	44000	0.0178
0.8922	44500	0.018
0.9022	45000	0.0181
0.9122	45500	0.0178
0.9223	46000	0.018
0.9323	46500	0.018
0.9423	47000	0.0178
0.9523	47500	0.0178
0.9623	48000	0.0179
0.9724	48500	0.018
0.9824	49000	0.0181
0.9924	49500	0.018

Framework Versions

• Python: 3.11.11
• Sentence Transformers: 3.4.1
• PyLate: 1.1.7
• Transformers: 4.48.2
• PyTorch: 2.6.0+cu124
• Accelerate: 1.5.2
• Datasets: 3.5.0
• Tokenizers: 0.21.1

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"
}

PyLate

@misc{PyLate,
title={PyLate: Flexible Training and Retrieval for Late Interaction Models},
author={Chaffin, Antoine and Sourty, Raphaël},
url={https://github.com/lightonai/pylate},
year={2024}
}