zyc-zju/qwen3-embedding-4b_qwen3-8b_hotpotqa_lsr

SentenceTransformer based on Qwen/Qwen3-Embedding-4B

This is a sentence-transformers model finetuned from Qwen/Qwen3-Embedding-4B on the flash_rag_datasets dataset. It maps sentences & paragraphs to a 2560-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: Qwen/Qwen3-Embedding-4B
• Maximum Sequence Length: 40960 tokens
• Output Dimensionality: 2560 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • flash_rag_datasets
• Language: en

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': 40960, 'do_lower_case': False, 'architecture': 'PeftModelForFeatureExtraction'})
  (1): Pooling({'word_embedding_dimension': 2560, 'pooling_mode_cls_token': False, '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': True, '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("qwen3-embedding-4b_qwen3-8b_hotpotqa_lsr")
# Run inference
queries = [
    "Huddersfield Giants R.L.F.C. are an English professional rugby league club from Huddersfield, West Yorkshire, the birthplace of rugby league, who play in the Super League competition, they play their home games at the Kirklees Stadium which is shared with Huddersfield Town F.C., is a multi-use sports stadium in Huddersfield in West Yorkshire, in which country?",
]
documents = [
    'England',
    'Alan Menken',
    'lead singer',
]
query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(documents)
print(query_embeddings.shape, document_embeddings.shape)
# [1, 2560] [3, 2560]

# Get the similarity scores for the embeddings
similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities)
# tensor([[0.2828, 0.1562, 0.1402]])

Training Details

Training Dataset

flash_rag_datasets

• Dataset: flash_rag_datasets at bcafb8d
• Size: 90,447 training samples
• Columns: query and response
• Approximate statistics based on the first 1000 samples:

  	query	response
  type	string	string
  details	min: 8 tokens mean: 25.45 tokens max: 147 tokens	min: 2 tokens mean: 4.93 tokens max: 73 tokens

• Samples:

  query	response
  Which magazine was started first Arthur's Magazine or First for Women?	Arthur's Magazine
  The Oberoi family is part of a hotel company that has a head office in what city?	Delhi
  Musician and satirist Allie Goertz wrote a song about the "The Simpsons" character Milhouse, who Matt Groening named after who?	President Richard Nixon

• Loss: fed_rag.loss.pytorch.lsr.LSRLoss

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 1
• gradient_accumulation_steps: 16
• learning_rate: 1e-05
• max_steps: 100
• lr_scheduler_type: constant
• remove_unused_columns: False
• dataloader_pin_memory: False
• push_to_hub: 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: 1
• per_device_eval_batch_size: 8
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 16
• eval_accumulation_steps: None
• torch_empty_cache_steps: None
• learning_rate: 1e-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: 3.0
• max_steps: 100
• lr_scheduler_type: constant
• 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
• 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: False
• 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}
• parallelism_config: None
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch_fused
• optim_args: None
• adafactor: False
• group_by_length: False
• length_column_name: length
• project: huggingface
• trackio_space_id: trackio
• ddp_find_unused_parameters: None
• ddp_bucket_cap_mb: None
• ddp_broadcast_buffers: False
• dataloader_pin_memory: False
• dataloader_persistent_workers: False
• skip_memory_metrics: True
• use_legacy_prediction_loop: False
• push_to_hub: True
• resume_from_checkpoint: None
• hub_model_id: None
• hub_strategy: every_save
• hub_private_repo: None
• hub_always_push: False
• hub_revision: None
• 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
• include_tokens_per_second: False
• include_num_input_tokens_seen: no
• neftune_noise_alpha: None
• optim_target_modules: None
• batch_eval_metrics: False
• eval_on_start: False
• use_liger_kernel: False
• liger_kernel_config: None
• eval_use_gather_object: False
• average_tokens_across_devices: True
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss
0.0002	1	0.0009
0.0004	2	0.0009
0.0005	3	0.0006
0.0007	4	0.0005
0.0009	5	0.001
0.0011	6	0.0004
0.0012	7	0.0008
0.0014	8	0.0009
0.0016	9	0.0011
0.0018	10	0.0014
0.0019	11	0.0012
0.0021	12	0.0015
0.0023	13	0.0009
0.0025	14	0.0005
0.0027	15	0.0005
0.0028	16	0.0013
0.0030	17	0.001
0.0032	18	0.0005
0.0034	19	0.0005
0.0035	20	0.001
0.0037	21	0.0007
0.0039	22	0.0015
0.0041	23	0.0013
0.0042	24	0.0014
0.0044	25	0.0009
0.0046	26	0.0012
0.0048	27	0.0011
0.0050	28	0.0011
0.0051	29	0.0007
0.0053	30	0.0007
0.0055	31	0.0011
0.0057	32	0.0014
0.0058	33	0.0006
0.0060	34	0.0008
0.0062	35	0.0012
0.0064	36	0.0006
0.0065	37	0.0008
0.0067	38	0.0006
0.0069	39	0.0008
0.0071	40	0.0005
0.0073	41	0.0008
0.0074	42	0.0008
0.0076	43	0.0013
0.0078	44	0.0005
0.0080	45	0.0009
0.0081	46	0.0007
0.0083	47	0.001
0.0085	48	0.0009
0.0087	49	0.001
0.0088	50	0.001
0.0090	51	0.0015
0.0092	52	0.0006
0.0094	53	0.0009
0.0096	54	0.0009
0.0097	55	0.0012
0.0099	56	0.0007
0.0101	57	0.0006
0.0103	58	0.0006
0.0104	59	0.0006
0.0106	60	0.0005
0.0108	61	0.0004
0.0110	62	0.0008
0.0111	63	0.001
0.0113	64	0.0012
0.0115	65	0.0011
0.0117	66	0.001
0.0119	67	0.0011
0.0120	68	0.0011
0.0122	69	0.0012
0.0124	70	0.0008
0.0126	71	0.0006
0.0127	72	0.0006
0.0129	73	0.0006
0.0131	74	0.0015
0.0133	75	0.0011
0.0134	76	0.0009
0.0136	77	0.0012
0.0138	78	0.0008
0.0140	79	0.0011
0.0142	80	0.0008
0.0143	81	0.0009
0.0145	82	0.0009
0.0147	83	0.0007
0.0149	84	0.0007
0.0150	85	0.0008
0.0152	86	0.0012
0.0154	87	0.001
0.0156	88	0.0003
0.0157	89	0.0003
0.0159	90	0.001
0.0161	91	0.0008
0.0163	92	0.0006
0.0165	93	0.0009
0.0166	94	0.0012
0.0168	95	0.0012
0.0170	96	0.0007
0.0172	97	0.0007
0.0173	98	0.001
0.0175	99	0.0008
0.0177	100	0.0007

Framework Versions

• Python: 3.11.14
• Sentence Transformers: 5.2.0
• Transformers: 4.57.2
• PyTorch: 2.9.1+cu128
• Accelerate: 1.12.0
• Datasets: 4.5.0
• 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",
}