PyLate model based on Alibaba-NLP/gte-modernbert-base

This is a PyLate model trained on the ms-marco-en-bge-gemma 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: Alibaba-NLP/gte-modernbert-base
Document Length: 300 tokens
Query Length: 32 tokens
Output Dimensionality: 128 dimensions
Similarity Function: MaxSim
Training Dataset:
- ms-marco-en-bge-gemma
Language: English
License: Apache 2.0

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': 299, 'do_lower_case': False}) with Transformer model: ModernBertModel 
  (1): Dense({'in_features': 768, '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,
)

Evaluation

Metrics

BEIR Benchmark

GTE-ModernColBERT is the first model to outpeform ColBERT-small on the BEIR benchmark. As reproduction in the IR domain is challenging, we worked closely with Benjamin Clavié, the author of ColBERT-small to reproduce the evaluation setup of this model. Despite all these efforts and reducing to the maximum the difference in scores in most of the datasets, some are still a bit different. For this reason, we also report the results of ColBERT-small in the same setup we used to evaluate GTE-ModernColBERT for completness and fair comparison.

Model	Average	FiQA2018	NFCorpus	TREC-COVID	Touche2020	ArguAna	QuoraRetrieval	SCIDOCS	SciFact	NQ	ClimateFEVER	HotpotQA	DBPedia	CQADupstack	FEVER	MSMARCO
GTE-ModernColBERT	54.89	48.51	37.93	83.59	31.23	48.51	86.61	19.06	76.34	61.8	30.62	77.32	48.03	41	87.44	45.32
ColBERT-small (reported)	53.79	41.15	37.3	84.59	25.69	50.09	87.72	18.42	74.77	59.1	33.07	76.11	45.58	38.75	90.96	43.5
JinaColBERT-v2		40.8	34.6	83.4	27.4	36.6	88.7	18.6	67.8	64	23.9	76.6	47.1		80.5
ColBERT-small (rerunned)	53.35	41.01	36.86	83.14	24.95	46.76	87.89	18.72	74.02	59.42	32.83	76.88	46.36	39.36	88.66	43.44

PyLate Information Retrieval

Datasets: NanoClimateFEVER, NanoDBPedia, NanoFEVER, NanoFiQA2018, NanoHotpotQA, NanoMSMARCO, NanoNFCorpus, NanoNQ, NanoQuoraRetrieval, NanoSCIDOCS, NanoArguAna, NanoSciFact and NanoTouche2020
Evaluated with pylate.evaluation.pylate_information_retrieval_evaluator.PyLateInformationRetrievalEvaluator

Metric	NanoClimateFEVER	NanoDBPedia	NanoFEVER	NanoFiQA2018	NanoHotpotQA	NanoMSMARCO	NanoNFCorpus	NanoNQ	NanoQuoraRetrieval	NanoSCIDOCS	NanoArguAna	NanoSciFact	NanoTouche2020
MaxSim_accuracy@1	0.36	0.88	0.92	0.56	0.92	0.54	0.56	0.64	0.96	0.48	0.3	0.74	0.7755
MaxSim_accuracy@3	0.62	0.94	0.98	0.66	1.0	0.68	0.68	0.82	1.0	0.74	0.62	0.86	0.9388
MaxSim_accuracy@5	0.78	0.96	0.98	0.74	1.0	0.74	0.74	0.86	1.0	0.78	0.7	0.9	0.9796
MaxSim_accuracy@10	0.86	0.98	1.0	0.8	1.0	0.92	0.76	0.9	1.0	0.84	0.82	0.94	0.9796
MaxSim_precision@1	0.36	0.88	0.92	0.56	0.92	0.54	0.56	0.64	0.96	0.48	0.3	0.74	0.7755
MaxSim_precision@3	0.2333	0.7133	0.36	0.3267	0.58	0.2267	0.4333	0.2867	0.4	0.4	0.2067	0.3	0.6599
MaxSim_precision@5	0.208	0.656	0.216	0.256	0.36	0.148	0.392	0.18	0.256	0.292	0.14	0.196	0.6571
MaxSim_precision@10	0.128	0.572	0.11	0.152	0.186	0.092	0.304	0.1	0.134	0.194	0.082	0.104	0.5184
MaxSim_recall@1	0.1833	0.118	0.8567	0.3092	0.46	0.54	0.0664	0.61	0.8473	0.1007	0.3	0.715	0.0518
MaxSim_recall@3	0.289	0.2307	0.96	0.4784	0.87	0.68	0.102	0.78	0.9453	0.2467	0.62	0.83	0.1362
MaxSim_recall@5	0.4157	0.2962	0.96	0.5752	0.9	0.74	0.1284	0.82	0.9693	0.2997	0.7	0.885	0.2193
MaxSim_recall@10	0.4957	0.4146	0.98	0.6412	0.93	0.92	0.1566	0.88	0.9893	0.3967	0.82	0.93	0.334
MaxSim_ndcg@10	0.4148	0.7296	0.9452	0.567	0.9012	0.7089	0.3957	0.7645	0.9691	0.3987	0.5609	0.8372	0.5927
MaxSim_mrr@10	0.5266	0.9169	0.9522	0.6359	0.96	0.6447	0.627	0.739	0.9767	0.6137	0.4775	0.8117	0.8629
MaxSim_map@100	0.3347	0.5884	0.9271	0.5032	0.8592	0.6496	0.1918	0.7239	0.9552	0.3163	0.4824	0.8049	0.4257

Nano BEIR

Dataset: NanoBEIR_mean
Evaluated with pylate.evaluation.nano_beir_evaluator.NanoBEIREvaluator

Metric	Value
MaxSim_accuracy@1	0.6643
MaxSim_accuracy@3	0.8107
MaxSim_accuracy@5	0.8584
MaxSim_accuracy@10	0.9077
MaxSim_precision@1	0.6643
MaxSim_precision@3	0.3943
MaxSim_precision@5	0.3044
MaxSim_precision@10	0.2059
MaxSim_recall@1	0.3968
MaxSim_recall@3	0.5514
MaxSim_recall@5	0.6084
MaxSim_recall@10	0.6837
MaxSim_ndcg@10	0.6758
MaxSim_mrr@10	0.7496
MaxSim_map@100	0.5971

Training Details

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 16
learning_rate: 3e-05
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: 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: 3
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: True
fp16: False
fp16_opt_level: O1
half_precision_backend: auto
bf16_full_eval: False
fp16_full_eval: False
tf32: None
local_rank: 6
ddp_backend: None
tpu_num_cores: None
tpu_metrics_debug: False
debug: []
dataloader_drop_last: True
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