langcache-colbert-v1-4gpu

Fine-tuned ColBERT model for semantic caching

This is a PyLate model finetuned from colbert-ir/colbertv2.0 on the LangCache Sentence Pairs (subsets=['all'], train+val=True) 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: colbert-ir/colbertv2.0 <!-- at revision c1e84128e85ef755c096a95bdb06b47793b13acf -->
• Document Length: 128 tokens
• Query Length: 128 tokens
• Output Dimensionality: 128 tokens
• Similarity Function: MaxSim
• Training Dataset:
  • LangCache Sentence Pairs (subsets=['all'], train+val=True)
• Language: en
• 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': 127, 'do_lower_case': False, 'architecture': 'BertModel'})
  (1): Dense({'in_features': 768, 'out_features': 128, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity', 'use_residual': False})
)

Usage

First install the PyLate library:

pip install -U pylate

Retrieval

Use this model with PyLate to index and retrieve documents. The index uses FastPLAID for efficient similarity search.

Indexing documents

Load the ColBERT model and initialize the PLAID 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="aditeyabaral/langcache-colbert-v1-4gpu",
)

# Step 2: Initialize the PLAID index
index = indexes.PLAID(
    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.PLAID(
    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="aditeyabaral/langcache-colbert-v1-4gpu",
)

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,
)

<!-- ### Direct Usage (Transformers) <details><summary>Click to see the direct usage in Transformers</summary> </details> --> <!-- ### Downstream Usage (Sentence Transformers) You can finetune this model on your own dataset. <details><summary>Click to expand</summary> </details> --> <!-- ### Out-of-Scope Use *List how the model may foreseeably be misused and address what users ought not to do with the model.* -->

Evaluation

Metrics

Col BERTTriplet

• Dataset: test_triplet
• Evaluated with <code>pylate.evaluation.colbert_triplet.ColBERTTripletEvaluator</code>

<!-- ## Bias, Risks and Limitations *What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.* --> <!-- ### Recommendations *What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.* -->

Training Details

Training Dataset

LangCache Sentence Pairs (subsets=['all'], train+val=True)

• Dataset: LangCache Sentence Pairs (subsets=['all'], train+val=True)
• Size: 1,452,533 training samples
• Columns: <code>anchor</code>, <code>positive</code>, and <code>negative_1</code>
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative_1
  type	string	string	string
  details	<ul><li>min: 9 tokens</li><li>mean: 29.49 tokens</li><li>max: 73 tokens</li></ul>	<ul><li>min: 8 tokens</li><li>mean: 29.18 tokens</li><li>max: 58 tokens</li></ul>	<ul><li>min: 4 tokens</li><li>mean: 23.79 tokens</li><li>max: 52 tokens</li></ul>

• Samples:

  anchor	positive	negative_1
  <code> Any Canadian teachers (B.Ed. holders) teaching in U.S. schools?</code>	<code> Any Canadian teachers (B.Ed. holders) teaching in U.S. schools?</code>	<code>Are there many Canadians living and working illegally in the United States?</code>
  <code> Are there any underlying psychological tricks/tactics that are used when designing the lines for rides at amusement parks?</code>	<code> Are there any underlying psychological tricks/tactics that are used when designing the lines for rides at amusement parks?</code>	<code>Is there any tricks for straight lines mcqs?</code>
  <code> Can I pay with a debit card on PayPal?</code>	<code> Can I pay with a debit card on PayPal?</code>	<code>Can you transfer PayPal funds onto a debit card/credit card?</code>

• Loss: <code>pylate.losses.contrastive.Contrastive</code>

Evaluation Dataset

LangCache Sentence Pairs (split=test)

• Dataset: LangCache Sentence Pairs (split=test)
• Size: 110,066 evaluation samples
• Columns: <code>anchor</code>, <code>positive</code>, and <code>negative_1</code>
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative_1
  type	string	string	string
  details	<ul><li>min: 5 tokens</li><li>mean: 28.57 tokens</li><li>max: 121 tokens</li></ul>	<ul><li>min: 5 tokens</li><li>mean: 28.01 tokens</li><li>max: 121 tokens</li></ul>	<ul><li>min: 7 tokens</li><li>mean: 20.73 tokens</li><li>max: 65 tokens</li></ul>

• Samples:

  anchor	positive	negative_1
  <code> What high potential jobs are there other than computer science?</code>	<code> What high potential jobs are there other than computer science?</code>	<code>Why IT or Computer Science jobs are being over rated than other Engineering jobs?</code>
  <code> Would India ever be able to develop a missile system like S300 or S400 missile?</code>	<code> Would India ever be able to develop a missile system like S300 or S400 missile?</code>	<code>Should India buy the Russian S400 air defence missile system?</code>
  <code> water from the faucet is being drunk by a yellow dog</code>	<code>A yellow dog is drinking water from the faucet</code>	<code>Do you get more homework in 9th grade than 8th?</code>

• Loss: <code>pylate.losses.contrastive.Contrastive</code>

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 48
• num_train_epochs: 5
• learning_rate: 0.0002
• warmup_steps: 0.1
• optim: adamw_torch
• weight_decay: 0.001
• eval_strategy: steps
• per_device_eval_batch_size: 48
• eval_on_start: True
• push_to_hub: True
• hub_model_id: aditeyabaral/langcache-colbert-v1-4gpu
• load_best_model_at_end: True
• ddp_find_unused_parameters: True
• batch_sampler: no_duplicates

All Hyperparameters

<details><summary>Click to expand</summary>

• per_device_train_batch_size: 48
• num_train_epochs: 5
• max_steps: -1
• learning_rate: 0.0002
• lr_scheduler_type: linear
• lr_scheduler_kwargs: None
• warmup_steps: 0.1
• optim: adamw_torch
• optim_args: None
• weight_decay: 0.001
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• optim_target_modules: None
• gradient_accumulation_steps: 1
• average_tokens_across_devices: True
• max_grad_norm: 1.0
• label_smoothing_factor: 0.0
• bf16: False
• fp16: False
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• use_liger_kernel: False
• liger_kernel_config: None
• use_cache: False
• neftune_noise_alpha: None
• torch_empty_cache_steps: None
• auto_find_batch_size: False
• log_on_each_node: True
• logging_nan_inf_filter: True
• include_num_input_tokens_seen: no
• log_level: passive
• log_level_replica: warning
• disable_tqdm: False
• project: huggingface
• trackio_space_id: trackio
• eval_strategy: steps
• per_device_eval_batch_size: 48
• prediction_loss_only: True
• eval_on_start: True
• eval_do_concat_batches: True
• eval_use_gather_object: False
• eval_accumulation_steps: None
• include_for_metrics: []
• batch_eval_metrics: False
• save_only_model: False
• save_on_each_node: False
• enable_jit_checkpoint: False
• push_to_hub: True
• hub_private_repo: None
• hub_model_id: aditeyabaral/langcache-colbert-v1-4gpu
• hub_strategy: every_save
• hub_always_push: False
• hub_revision: None
• load_best_model_at_end: True
• ignore_data_skip: False
• restore_callback_states_from_checkpoint: False
• full_determinism: False
• seed: 42
• data_seed: None
• use_cpu: False
• accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
• parallelism_config: None
• dataloader_drop_last: True
• dataloader_num_workers: 0
• dataloader_pin_memory: True
• dataloader_persistent_workers: False
• dataloader_prefetch_factor: None
• remove_unused_columns: True
• label_names: None
• train_sampling_strategy: random
• length_column_name: length
• ddp_find_unused_parameters: True
• ddp_bucket_cap_mb: None
• ddp_broadcast_buffers: False
• ddp_backend: None
• ddp_timeout: 1800
• fsdp: []
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
• deepspeed: None
• debug: []
• skip_memory_metrics: True
• do_predict: False
• resume_from_checkpoint: None
• warmup_ratio: None
• local_rank: -1
• prompts: None
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

</details>

Training Logs

Framework Versions

• Python: 3.12.12
• Sentence Transformers: 5.3.0
• PyLate: 1.5.0
• Transformers: 5.3.0
• PyTorch: 2.9.0+cu130
• Accelerate: 1.13.0
• Datasets: 4.8.5
• Tokenizers: 0.22.2

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

@inproceedings{DBLP:conf/cikm/ChaffinS25,
  author       = {Antoine Chaffin and
                  Rapha{"{e}}l Sourty},
  editor       = {Meeyoung Cha and
                  Chanyoung Park and
                  Noseong Park and
                  Carl Yang and
                  Senjuti Basu Roy and
                  Jessie Li and
                  Jaap Kamps and
                  Kijung Shin and
                  Bryan Hooi and
                  Lifang He},
  title        = {PyLate: Flexible Training and Retrieval for Late Interaction Models},
  booktitle    = {Proceedings of the 34th {ACM} International Conference on Information
                  and Knowledge Management, {CIKM} 2025, Seoul, Republic of Korea, November
                  10-14, 2025},
  pages        = {6334--6339},
  publisher    = {{ACM}},
  year         = {2025},
  url          = {https://github.com/lightonai/pylate},
  doi          = {10.1145/3746252.3761608},
}

<!-- ## Glossary *Clearly define terms in order to be accessible across audiences.* --> <!-- ## Model Card Authors *Lists the people who create the model card, providing recognition and accountability for the detailed work that goes into its construction.* --> <!-- ## Model Card Contact *Provides a way for people who have updates to the Model Card, suggestions, or questions, to contact the Model Card authors.* -->