Train Sentence Transformers

Hugging FaceOfficialPublished by Hugging Face

Skill
Workflow
Data and AI
Apache-2.0
Medium risk
Has scripts

From the author

Train or fine-tune sentence-transformers models across `SentenceTransformer` (bi-encoder, dense or static embedding model for retrieval, similarity, clustering, classification, paraphrase mining, dedup, multimodal), `CrossEncoder` (reranker, pair scoring for two-stage retrieval / pair classification), `SparseEncoder` (SPLADE, sparse embedding model for learned-sparse retrieval), and `MultiVectorEncoder` (ColBERT / late-interaction, per-token embeddings scored with MaxSim). Covers loss selection, hard-negative mining, evaluators, distillation, LoRA, Matryoshka, and Hugging Face Hub publishing. Use for any sentence-transformers training task.

— the description by Hugging Face

AutoGineer summary

Written by AutoGineer from the skill’s own files, not by the author

This skill provides a standardized approach to training various embedding and ranking models using the sentence-transformers framework. It acts as a routing system that directs users to specific production templates and technical references based on the chosen model architecture. The workflow covers the entire lifecycle from model selection and data preparation to loss function configuration, evaluation, and publishing to the Hugging Face Hub.

What it does

Routes users to appropriate training templates for bi-encoders, cross-encoders, sparse encoders, and multi-vector encoders.
Provides technical references for loss functions, evaluators, and model architectures.
Offers guidance on hard-negative mining and dataset formatting for training.
Includes instructions for hardware-specific configurations and Hugging Face Hub integration.
Defines mandatory script requirements such as smoke testing, logging, and verdict reporting.

When to use it

When you need to train or fine-tune a sentence-transformers model for retrieval, similarity, or ranking tasks.
When you are unsure which model architecture (e.g., bi-encoder vs. cross-encoder) is best suited for your specific use case.
When you require a structured, repeatable process for training and evaluating embedding models.
When you need to integrate advanced training techniques like distillation, LoRA, or Matryoshka embeddings.

Advantages

Enforces consistent training practices through standardized production templates.
Reduces common errors by providing specific references for loss-to-data-shape mapping.
Facilitates systematic experimentation with built-in logging and verdict reporting.
Supports a wide range of model types including modern late-interaction and sparse architectures.

Usage examples

“I need to train a bi-encoder for semantic search on my internal documentation.”: The agent will identify the [SentenceTransformer] type, provide the relevant loss and evaluator references, and guide you to the bi-encoder production template.
“How do I implement a two-stage retrieval system using a reranker?”: The agent will direct you to the [CrossEncoder] documentation and template, explaining how to use it to score pairs for two-stage retrieval.
“I want to train a SPLADE model for sparse retrieval.”: The agent will provide the [SparseEncoder] references, including the required SpladeLoss wrapper and specific evaluators for sparse retrieval.
“How can I improve my retrieval model using hard negatives?”: The agent will guide you to the hard-negative mining script and explain how to generate n-tuple datasets for training.

Install

Download the skill as a ZIP:
https://mp-staging.autogineer.com/agent-skills/huggingface/train-sentence-transformers/download

Unzip it into the cross-client skills folder of your project:

mkdir -p .agents/skills && unzip train-sentence-transformers.zip -d .agents/skills

Documentation by Hugging Face

Train a sentence-transformers Model

This SKILL.md is a router, not a manual. It tells you which references and example scripts to load for your task. The actual content (recommended losses, evaluators, training-script structure, model selection, training-arg knobs, troubleshooting) lives in references/ and scripts/.

Do not synthesize a training script from this file alone. Open the per-type production template (scripts/train_<type>_example.py) and copy it as your starting point. The templates contain load-bearing scaffolding (autocast helper, model-card class, logger silencing list, force=True, seed, TF32, version-compatible imports, named-evaluator metric handling) that prior agent runs have repeatedly missed when rolling their own from a synthesized snippet.

1. Identify the model type

Tag	Class	What it does	When to pick
[SentenceTransformer]	`SentenceTransformer` (bi-encoder)	Maps each input to a fixed-dim dense vector	Retrieval, similarity, clustering, classification, paraphrase mining, dedup
[CrossEncoder]	`CrossEncoder` (reranker)	Scores `(query, passage)` pairs jointly	Two-stage retrieval (rerank top-100 from bi-encoder), pair classification
[SparseEncoder]	`SparseEncoder` (SPLADE)	Sparse vectors over the vocabulary	Learned-sparse retrieval, inverted-index backends (Elasticsearch / OpenSearch / Lucene)
[MultiVectorEncoder]	`MultiVectorEncoder` (ColBERT)	One embedding per token, scored with MaxSim	Late-interaction retrieval, recall gains over bi-encoders at higher storage cost, multimodal (ColPali / ColQwen2)

Tiebreakers when the request is ambiguous: "embedding model" / "vector search" / "similarity" → [SentenceTransformer]. "rerank" / "ranker" / "two-stage" → [CrossEncoder]. "SPLADE" / "sparse" / "inverted index" → [SparseEncoder]. "ColBERT" / "late interaction" / "multi-vector" / "MaxSim" / "ColPali" / "ColQwen" → [MultiVectorEncoder]. If still unclear, ask.

2. Required reading

Read these in full before writing any code. Do not triage by perceived relevance.

Per-type: always required

[SentenceTransformer]

references/losses_sentence_transformer.md: loss-to-data-shape mapping, BatchSamplers.NO_DUPLICATES requirement for MNRL-family, Cached* ↔ gradient_checkpointing incompatibility.
references/evaluators_sentence_transformer.md: evaluator-to-task mapping, metric_for_best_model key construction (named vs unnamed), per-evaluator primary_metric values.
references/model_architectures.md: encoder vs decoder vs static vs Router pipelines, pooling rules (mean / cls / lasttoken), auto-mean-pooling behavior for fresh-start MLM bases.
scripts/train_sentence_transformer_example.py: production template. Copy this as your starting point.

[CrossEncoder]

references/losses_cross_encoder.md: pointwise / pairwise / listwise / distillation, pos_weight derivation, activation_fn=Identity() mandatory for non-BCE losses (silent eval-rank collapse otherwise).
references/evaluators_cross_encoder.md: CrossEncoderRerankingEvaluator recipe, named-evaluator key format eval_{name}_{primary_metric}.
scripts/train_cross_encoder_example.py: production template. Copy this as your starting point.

[SparseEncoder]

references/losses_sparse_encoder.md: SpladeLoss wrapper requirement, FLOPS regularizer weights, smoke-test active-dim ramp behavior.
references/evaluators_sparse_encoder.md: SparseNanoBEIREvaluator (English-only) and the in-domain alternative, eval_{name}_{primary_metric} key format.
scripts/train_sparse_encoder_example.py: production template. Copy this as your starting point.

[MultiVectorEncoder]

references/losses_multi_vector_encoder.md: MaxSim scoring, scale choice per scoring mode (scale=1.0 for MaxSim, roughly the average query length for MeanMaxSim), MNRL / CachedMNRL / MarginMSE / DistillKLDiv, XTR-vs-ColBERT scoring, CachedMNRL ↔ gradient_checkpointing incompatibility.
references/evaluators_multi_vector_encoder.md: MultiVectorNanoBEIREvaluator (English-only) and the in-domain alternative, eval_NanoBEIR_mean_maxsim_ndcg@10 key format, distillation-eval spearman variant.
scripts/train_multi_vector_encoder_example.py: production template. Copy this as your starting point.

Cross-cutting: always required (regardless of task)

references/training_args.md: TrainingArguments knobs, precision rules (load fp32 + autocast bf16/fp16, never torch_dtype=bfloat16), warmup_steps (float) vs deprecated warmup_ratio, save_steps must be a multiple of eval_steps for load_best_model_at_end, schedulers, HPO, tracker, resume, hub-push variants.
references/dataset_formats.md: column-matching rules (label name auto-detection, column-order-not-name), reshaping recipes, hard-negative mining options.
references/base_model_selection.md: discovery commands, per-type model namespaces, ModernBERT-family max_seq_length=8192 trap, datasets >= 4 script-loader rejection, non-English starting-point shortcuts.
references/troubleshooting.md: symptom-indexed failure recipes. Skim the section headings on every run, even a healthy one. The "Metrics don't improve" and "Hub push fails" entries cover bugs that bite frequently and are cheaper to recognize before they fire than to debug after.

Cross-cutting: load when applicable

references/hardware_guide.md: VRAM sizing, multi-GPU, FSDP / DeepSpeed, HF Jobs flavors. Required for >24GB models, multi-GPU, or HF Jobs runs.
references/hf_jobs_execution.md: required when running on HF Jobs.
references/prompts_and_instructions.md: required when using prompt-tuned bases (E5, BGE, GTE, Qwen3-Embedding, Instructor, Nomic, etc.) or adding query: / passage: style prefixes.

Variant scripts (open when the task matches)

[SentenceTransformer] scripts/train_sentence_transformer_<matryoshka|multi_dataset|with_lora|distillation|make_multilingual|static_embedding>_example.py.
[CrossEncoder] scripts/train_cross_encoder_<distillation|listwise>_example.py.
[SparseEncoder] scripts/train_sparse_encoder_distillation_example.py.
Hard-negative mining CLI: scripts/mine_hard_negatives.py.

3. Defaults

Override only if the user specifies otherwise:

Local execution. Pitch HF Jobs only if local hardware can't fit the job.
Single run. After it completes, propose experimentation if the user would benefit (weak/marginal verdict, "see how high you can push it" framing, etc.). Iteration rules in references/training_args.md (Experimentation section).
Public Hub push at end-of-run, wrapped in try-except. On HF Jobs (ephemeral env) ALSO enable in-trainer push (push_to_hub=True + hub_strategy="every_save"). Details in references/hf_jobs_execution.md.

4. Constraints the produced script must satisfy

These are non-negotiable contracts. Implementation lives in the production templates and references. Do not reinvent.

Capture the pre-training evaluator score as baseline_eval before trainer.train().
Emit a single end-of-run line: VERDICT: WIN|MARGINAL|REGRESSION | score=... | baseline=... | delta=.... A monitor scrapes for this.
Silence httpx, httpcore, huggingface_hub, urllib3, filelock, fsspec to WARNING (otherwise HF download URLs flood the agent's context).
Tee logs to logs/{RUN_NAME}.log.
End with model.push_to_hub(...) wrapped in try/except.
Smoke-test before any long run (max_steps=1 + tiny dataset slice). The production templates show one common pattern (SMOKE_TEST env var).
[CrossEncoder] Include EarlyStoppingCallback(patience>=3). CE rerankers often peak mid-training and regress.
[SparseEncoder] Log query_active_dims / corpus_active_dims on the verdict line. High nDCG with collapsed sparsity is not a win. The keys come back name-prefixed (e.g. ..._query_active_dims). Use suffix matching to pluck them. See the SPARSE production template for the exact pattern.
[MultiVectorEncoder] Match scale to the scoring mode on any MNRL-family loss: near 1.0 for unnormalized MaxSim (do not copy scale=20.0 from bi-encoder MNRL), roughly the average query length with length-normalized MeanMaxSim, since each score is divided by its query's token count. XTRScores is a train-only similarity_fct: the evaluators reject it, so evaluation always scores with MaxSim, including for XTR-trained models.

5. Workflow

Identify the model type (§1). Ask if ambiguous.
Load the §2 required-reading files for that type.
Open scripts/train_<type>_example.py and copy it as your starting point.
Replace MODEL_NAME, DATASET_NAME, RUN_NAME, the loss, and the evaluator with the user's task. Cross-check loss/data-shape match against references/losses_<type>.md. Cross-check the metric_for_best_model key against references/evaluators_<type>.md (named evaluators format the key as eval_{name}_{primary_metric}).
Smoke-test (max_steps=1).
Run.
After the run, append to logs/experiments.md and propose iteration if the verdict is weak/marginal.

Prerequisites

pip install "sentence-transformers[train]>=5.0"        # add [train,image] / [audio] / [video] for [SentenceTransformer] multimodal
                                                       # [MultiVectorEncoder] requires >=6.0
pip install trackio                                    # optional tracker (or wandb / tensorboard / mlflow)
hf auth login                                          # or set HF_TOKEN with write scope (for Hub push)

GPU strongly recommended. CPU works only for demos and [SentenceTransformer] StaticEmbedding.

Files

ATTRIBUTION.mdAdded by AutoGineer898 B
LICENSELicence file added by AutoGineer11.1 KiB
SKILL.md10.3 KiB
references/base_model_selection.md7.5 KiB
references/dataset_formats.md6.3 KiB
references/evaluators_cross_encoder.md5.5 KiB
references/evaluators_multi_vector_encoder.md6.6 KiB
references/evaluators_sentence_transformer.md5.8 KiB
references/evaluators_sparse_encoder.md4.9 KiB
references/hardware_guide.md5.9 KiB
references/hf_jobs_execution.md7.1 KiB
references/losses_cross_encoder.md9.5 KiB
references/losses_multi_vector_encoder.md10.5 KiB
references/losses_sentence_transformer.md10.7 KiB
references/losses_sparse_encoder.md5.1 KiB
references/model_architectures.md9.0 KiB
references/prompts_and_instructions.md6.4 KiB
references/training_args.md15.0 KiB
references/troubleshooting.md12.9 KiB
scripts/mine_hard_negatives.pyScript8.3 KiB
scripts/train_cross_encoder_distillation_example.pyScript9.2 KiB
scripts/train_cross_encoder_example.pyScript9.0 KiB
scripts/train_cross_encoder_listwise_example.pyScript10.2 KiB
scripts/train_multi_vector_encoder_example.pyScript8.8 KiB
scripts/train_sentence_transformer_distillation_example.pyScript11.9 KiB
scripts/train_sentence_transformer_example.pyScript7.5 KiB
scripts/train_sentence_transformer_make_multilingual_example.pyScript11.5 KiB
scripts/train_sentence_transformer_matryoshka_example.pyScript7.1 KiB
scripts/train_sentence_transformer_multi_dataset_example.pyScript9.9 KiB
scripts/train_sentence_transformer_static_embedding_example.pyScript10.3 KiB
scripts/train_sentence_transformer_with_lora_example.pyScript10.3 KiB
scripts/train_sparse_encoder_distillation_example.pyScript9.4 KiB
scripts/train_sparse_encoder_example.pyScript8.1 KiB

references/hardware_guide.md

Hardware Guide

Training embedding models is memory-bound more often than compute-bound.

If you hit OOM

Try in this order:

Reduce per_device_train_batch_size. Raise gradient_accumulation_steps to keep the effective batch size for regression losses. (For MNRL, effective batch via grad-accum is not equivalent. See point 3.)
Enable gradient_checkpointing=True. ~30% slower, ~40% less activation memory. Incompatible with Cached* losses.
Switch to a Cached* loss:
- CachedMultipleNegativesRankingLoss(model, mini_batch_size=32): forwards in mini-batches, accumulates the contrastive loss over the full batch. Can simulate batch sizes of 1024+ on a 24GB GPU.
- CachedSpladeLoss(model, loss=..., mini_batch_size=16): same trick for sparse.
- CachedGISTEmbedLoss(model, guide_model, mini_batch_size=32): GIST variant.
Enable PEFT / LoRA for decoder models >1B. LoraConfig(r=64, lora_alpha=128, task_type="FEATURE_EXTRACTION"). See ../scripts/train_sentence_transformer_with_lora_example.py (docstring covers when to use, hyperparams, QLoRA, sharing).
Move to multi-GPU. See below.
Shorten sequences. If truncating to 128 is already sufficient for your task, set max_seq_length on the transformer module.

Multi-GPU

sentence-transformers uses accelerate under the hood. Distributed training works without code changes.

Data parallel (DDP)

Launch:

accelerate launch train.py

# or explicitly:
accelerate launch --multi_gpu --num_processes=4 train.py

per_device_train_batch_size stays per-GPU. Effective batch size scales linearly. MNRL's in-batch negatives remain per-device, not global, unless you pass gather_across_devices=True to losses that support it (MultipleNegativesRankingLoss, CachedMultipleNegativesRankingLoss, the symmetric variants, GISTEmbedLoss, CachedGISTEmbedLoss, SparseMultipleNegativesRankingLoss).

FSDP / DeepSpeed

For models >3B, use accelerate config to enable FSDP or DeepSpeed ZeRO. Both are supported: sentence-transformers doesn't require any code changes, only the launch config.

accelerate config                   # interactive; choose FSDP or DeepSpeed
accelerate launch train.py

With FSDP full-shard: a 7B model trains on 4×24GB GPUs that would OOM on any single one of them.

FSDP caveats (from the distributed training docs):

Evaluators don't run under FSDP as of writing: the eval hooks call model.encode() which FSDP-wrapped modules can't service mid-training. Plan to evaluate post-training on a single-GPU load of the final checkpoint instead, or train with DDP if you need mid-training evaluation.
Layer wrapping must be specified, e.g. fsdp_config={"transformer_layer_cls_to_wrap": "BertLayer"} (substitute the right layer class for your model: BertLayer, LlamaDecoderLayer, Qwen2DecoderLayer, etc.). Without this FSDP sharding can silently misbehave.
Slower than DDP for models that fit on a single GPU. Only reach for FSDP when you actually need the memory savings.

DeepSpeed ZeRO-2/3 is an alternative with its own config. It works identically at the accelerate config level.

Effective batch size for contrastive losses

For MultipleNegativesRankingLoss and its variants, batch size is a quality knob, not just a speed knob. Bigger batches = more in-batch negatives = richer gradients.

The effective pool of in-batch negatives per anchor:

Setup	In-batch negatives per anchor
Single GPU, batch 64	63
4× DDP, per-device batch 64	63 local only by default. 255 with `MultipleNegativesRankingLoss(model, gather_across_devices=True)`
Single GPU, CachedMNRL, mini_batch 32, batch 256	255
4× DDP, CachedMNRL, per-device 256	255 local. 1023 with `gather_across_devices=True`

For large corpora (retrieval), push toward 512+ effective negatives. For small, clean datasets (STS), 64 is plenty.

Precision choice by GPU

GPU generation	Recommended
T4, V100, GTX 1xxx, RTX 2xxx	`fp16=True`
RTX 3xxx, A10G, A100, L4	`bf16=True`
RTX 4xxx, H100, B200	`bf16=True` (or fp8 on H100 via specific kernels, not default)
Apple M-series / ROCm	MPS/ROCm support is variable. `fp16` or `fp32` most reliable

bf16 is more numerically stable and almost always preferred when available.

Hugging Face Jobs flavor guide

Hugging Face Jobs requires a Pro/Team/Enterprise plan. Pricing is approximate and subject to change. See the Jobs pricing page.

Flavor	Memory	Typical use	Est. $/hr
`cpu-basic`	~2 GB	Dataset prep, validation, hard-neg mining (small)	<$0.10
`cpu-upgrade`	~4 GB	Same, slightly bigger	$0.10
`t4-small`	16 GB	Demos, MiniLM/DistilBERT with small batches	~$0.75
`t4-medium`	16 GB	MiniLM / DistilBERT with larger batch	~$1.50
`l4x1`	24 GB	BERT-base, MPNet, ModernBERT-base	~$2.50
`a10g-small`	24 GB	BERT-base to BERT-large	~$3.50
`a10g-large`	48 GB	ModernBERT-large, Qwen3-0.6B	~$5.00
`a10g-largex2`	96 GB (2× 48GB)	Mid-size multi-GPU	~$10
`a100-large`	80 GB	Large models or big contrastive batches	~$10-12
`h100`	80 GB	Biggest single-GPU	~$12
`h100x8`	640 GB	LLM-scale distributed	~$96

Defaults by base model:

MiniLM / DistilBERT -> t4-small
BERT-base / MPNet / ModernBERT-base -> a10g-small or l4x1
BERT-large / ModernBERT-large -> a10g-large
Qwen3-0.6B decoder base -> a10g-large
1B+ decoder bases with LoRA -> a10g-large or a100-large

Always start one flavor smaller than you think you need: OOM on Jobs is cheap ($0.50-$5 for a failed run). Underprovisioned is better than overprovisioned for the first attempt. When budgeting timeout, add 20-30% buffer for model loading, checkpoint saving, and Hub push.

Requirements and compatibility

Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, OpenCode, Goose, Amp, Roo Code, JetBrains Junie, Factory Droid, Devin (ex-Windsurf), Mistral Vibe, OpenHands, Letta

“Works with” means the tool reads the open Agent Skills format; a skill written for one tool may use features others lack.

Needs: Needs Python

Security

What our scanner noticed· Medium risk

Instructions that use the network
Downloads piped into a shell
Scripts that your agent may run

This skill includes scripts that your agent may run. Read them before you use it.

A static scan is not a guarantee. Skills are third-party content: review a skill before you let an agent use it.

Attribution and licence

Published by

Hugging Face

Publisher

Hugging Face (Official)

Source

huggingface/skills · skills/train-sentence-transformers · commit 25910b3

Licence

Apache License 2.0 (redistribution allowed)

LICENSE (added by AutoGineer from LICENSE)
ATTRIBUTION.md (written by AutoGineer)

Synced

2026-11-08

AutoGineer is not affiliated with or endorsed by Hugging Face.

Download ZIP

The ZIP holds the skill’s files, its licence files and an ATTRIBUTION.md. Download train-sentence-transformers.zip