metadata

language:
  - code
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - dense
  - generated_from_trainer
  - dataset_size:197351
  - loss:MultipleNegativesRankingLoss
base_model: Qwen/Qwen3-Embedding-0.6B
widget:
  - source_sentence: ABCB7
    sentences:
      - >-
        This gene encodes a tetrameric mitochondrial flavoprotein, which is a
        member of the acyl-CoA dehydrogenase family. This enzyme catalyzes the
        initial step of the mitochondrial fatty acid beta-oxidation pathway.
        Mutations in this gene have been associated with short-chain acyl-CoA
        dehydrogenase (SCAD) deficiency. Alternative splicing results in two
        variants which encode different isoforms. [provided by RefSeq, Oct 2014]
      - >-
        The membrane-associated protein encoded by this gene is a member of the
        superfamily of ATP-binding cassette (ABC) transporters. ABC proteins
        transport various molecules across extra- and intra-cellular membranes.
        ABC genes are divided into seven distinct subfamilies (ABC1, MDR/TAP,
        MRP, ALD, OABP, GCN20, White). This protein is a member of the MDR/TAP
        subfamily. Members of the MDR/TAP subfamily are involved in multidrug
        resistance as well as antigen presentation. This gene encodes a
        half-transporter involved in the transport of heme from the mitochondria
        to the cytosol. With iron/sulfur cluster precursors as its substrates,
        this protein may play a role in metal homeostasis. Mutations in this
        gene have been associated with mitochondrial iron accumulation and
        isodicentric (X)(q13) and sideroblastic anemia. Alternatively spliced
        transcript variants encoding multiple isoforms have been observed for
        this gene. [provided by RefSeq, Nov 2012]
      - >-
        The membrane-associated protein encoded by this gene is a member of the
        superfamily of ATP-binding cassette (ABC) transporters. ABC proteins
        transport various molecules across extra- and intracellular membranes.
        ABC genes are divided into seven distinct subfamilies (ABC1, MDR/TAP,
        MRP, ALD, OABP, GCN20, and White). This encoded protein is a member of
        the ABC1 subfamily. Members of the ABC1 subfamily comprise the only
        major ABC subfamily found exclusively in multicellular eukaryotes. This
        gene is clustered among 4 other ABC1 family members on 17q24, but
        neither the substrate nor the function of this gene is known.
        Alternative splicing of this gene results in several transcript
        variants; however, not all variants have been fully described. [provided
        by RefSeq, Jul 2008]
  - source_sentence: ABCC8
    sentences:
      - >-
        The protein encoded by this gene is a member of the superfamily of
        ATP-binding cassette (ABC) transporters. ABC proteins transport various
        molecules across extra- and intra-cellular membranes. ABC genes are
        divided into seven distinct subfamilies (ABC1, MDR/TAP, MRP, ALD, OABP,
        GCN20, White). This protein is a member of the MRP subfamily which is
        involved in multi-drug resistance. This protein functions as a modulator
        of ATP-sensitive potassium channels and insulin release. Mutations in
        the ABCC8 gene and deficiencies in the encoded protein have been
        observed in patients with hyperinsulinemic hypoglycemia of infancy, an
        autosomal recessive disorder of unregulated and high insulin secretion.
        Mutations have also been associated with non-insulin-dependent diabetes
        mellitus type II, an autosomal dominant disease of defective insulin
        secretion. Alternatively spliced transcript variants have been found for
        this gene. [provided by RefSeq, Jul 2020]
      - >-
        Predicted to enable GTPase activator activity and zinc ion binding
        activity. Predicted to be involved in protein transport. Located in
        membrane. [provided by Alliance of Genome Resources, Jul 2025]
      - >-
        The protein encoded by this gene is a member of the superfamily of
        ATP-binding cassette (ABC) transporters. ABC proteins transport various
        molecules across extra- and intra-cellular membranes. ABC genes are
        divided into seven distinct subfamilies (ABC1, MDR/TAP, MRP, ALD, OABP,
        GCN20, White). This ABC full transporter is a member of the MRP
        subfamily which is involved in multi-drug resistance. The product of
        this gene participates in physiological processes involving bile acids,
        conjugated steroids, and cyclic nucleotides. In addition, a SNP in this
        gene is responsible for determination of human earwax type. This gene
        and family member ABCC12 are determined to be derived by duplication and
        are both localized to chromosome 16q12.1. Multiple alternatively spliced
        transcript variants have been described for this gene. [provided by
        RefSeq, Jul 2008]
  - source_sentence: >-
      MALAT1 TMSB4X ACTB TPT1 EEF1A1 S100A10 LGALS1 VIM SH3BGRL3 S100A4 FTL PTMA
      SRGN TMSB10 CYBA GAPDH CD74 TAGLN2 FTH1 S100A6 UBA52 YBX1 MYL6 OAZ1 CST3
      NACA FAU ARPC2 GSTP1 PFN1 HSP90AA1 COTL1 PPIA ARPC3 UQCRB MYL12A CD63 EIF1
      NEAT1 RACK1 MACROH2A1 ATP6V0E1 ATP5F1E SRP14 ENO1 SLC25A3 CTSH PRDX1 VAMP8
      COX4I1 CAP1 BTF3 DBI HNRNPA3 GNAS DDX5 H3-3B TPM3 LAPTM5 ZEB2 GNG5 FLNA
      CALM1 CD44
    sentences:
      - >-
        MALAT1 PTMA TMSB10 LGALS1 ACTB PRDX1 S100A4 H3-3B TMSB4X VIM TPT1 LMO4
        HNRNPA2B1 SH3BGRL3 TAGLN2 HNRNPU DDIT4 PFN1 IGFBP7 HMGB1 FTH1 CFL1 CD74
        SOX4 KLF2 BST2 S100A11 RACK1 PSMA4 DDX5 NCL RSRP1 IRF1 SERF2 EEF1A1
        CALM1 UBA52 CYBA HSP90AA1 MYL12A AHNAK ITM2B SRP14 EMP3 CALM2 TSC22D3
        YWHAZ SELENOW PPIA S100A6 TSPO IRAG2 TPM3 UBC ARPC2 HNRNPA3 UBB EIF1 JUN
        IFITM2 PRR13 N4BP2L2 LAPTM4A CDC42
      - >-
        This measurement was conducted with 10x 3' v3. This sample is derived
        from a 3-month-old male patient with KMT2A-rearranged (KMT2A-r) infant
        acute lymphoblastic leukemia (ALL) with a CD8_Cytotoxic T cell type,
        specifically T/NK cells, and a presumed MLL-AF4 fusion.
      - >-
        This measurement was conducted with 10x 3' v3. Blast cells derived from
        a 1-month-old human with a presumed MLL-AF10 fusion, projected as
        cDC-like cells.
  - source_sentence: >-
      MALAT1 CXCL14 EEF1A1 VIM IGFBP7 COL1A2 FTH1 TPT1 S100A6 TMSB4X A2M APOE
      DCN PTGDS TMSB10 LGALS1 ACTB FBLN1 FTL RARRES2 CD81 CALD1 CD63 COL6A2 MYL6
      SPARCL1 NEAT1 IGFBP5 PTMA CST3 FAU SERF2 SPARC IFITM3 EIF1 S100A4 NACA
      JUND COL6A1 GSN C1S CFH HSP90AA1 PDLIM1 H3-3B EDIL3 UBA52 VCAN LTBP4 TIMP3
      CTSC ITM2B IGFBP4 UBC UBB RACK1 TIMP1 ACTA2 ZFP36L2 PLPP3 TUBA1A FILIP1L
      FOS S100A10
    sentences:
      - >-
        MALAT1 TMSB10 A2M FABP5 PTMA VIM ACTB CAV1 SPARCL1 CD74 EEF1A1 KLF2
        IFITM3 CLDN5 TMSB4X TPT1 ENPP2 TM4SF1 FOS EIF1 S100A6 CALM1 CD81 HES1
        SRGN ID1 GNG11 IGFBP4 STOM GSN TAGLN2 IGFBP7 CD320 FTH1 MCAM HSP90AA1
        GNAS MYL6 TIMP3 EPAS1 TNFSF10 PODXL ITM2B SRP14 UBC TGFBR2 KCTD12 GIMAP7
        UBA52 RHOA CD59 FTL PCSK5 MYH9 MYL12A FLT1 CXCL12 LIFR TUBA1B DSTN
        ARPC1B JUND H3-3B TMBIM6
      - >-
        This measurement was conducted with 10x 3' v3. Fibroblasts derived from
        the terminal ileum of a female individual in her fourth decade,
        exhibiting Crohn's disease (CD) related changes.
      - >-
        This measurement was conducted with 10x 3' v3. Glial cells derived from
        the ileal epithelium of a female in her fourth decade.
  - source_sentence: >-
      MALAT1 DCN MGP APOD GSN LAMA2 CST3 SPARCL1 IGFBP7 TIMP1 VIM EEF1A1 ITM2B
      FBLN1 C3 IFITM3 FBN1 FTH1 TPT1 ABCA8 C1S TXNIP FTL TIMP3 FN1 CD63 RBMS3
      ABCA6 ZBTB20 CEBPD NEAT1 CFH VCAN PTN PTGDS CD81 SERF2 COL6A1 COL6A2
      ABI3BP ABCA10 EBF1 COL1A2 PRKG1 S100A6 MGST1 TMSB10 TIMP2 CELF2 LAPTM4A
      RORA ACTB LTBP4 MYL6 LGALS1 DDX5 SPTBN1 EFEMP1 BICC1 LRP1 H3-3B SCN7A
      IGFBP4 FAU
    sentences:
      - >-
        This measurement was conducted with 10x 3' v3. CD4+T naive lymphocyte
        cells derived from the right cardiac atrium of a European male in his
        sixties.
      - >-
        This measurement was conducted with 10x multiome. Fibroblast cell sample
        taken from the right ventricle of a European female donor in her fifth
        decade, who is a DCD donor. The sample is in nucleus form.
      - >-
        MALAT1 NEAT1 LINC00486 SLC8A1 VMP1 SAT1 PIK3R5 DIRC3 FMN1 PMP22 RBM47
        AGFG1 DIP2B RBMS1 GNAQ TBC1D14 RAB1A ARHGAP24 DAPK1 SLC1A3 RHOQ SH3BGRL
        DOCK10 SLCO2B1 RUNX1 ENOX2 LDLRAD4 RNF150 PIAS1 DDX5 WSB1 TSHZ3 SBF2
        DOCK2 LRP4 DENND4C FCHSD2 EXOC6B AFF3 ARHGAP26 DIAPH2 MGAT5 TMEM163
        NSMCE2 RBPJ ZEB2 TANC2 BPTF SH3RF3 MFSD14CP TCF4 RORA-AS1 NOP58 MEF2A
        EPN2 PICALM ARHGAP15 MEF2C ANKRD12 FCGRT DOCK8 SETX TBC1D9 KLHL2
datasets:
  - jo-mengr/cellxgene_pseudo_bulk_100k_multiplets_natural_language_annotation
  - jo-mengr/descriptions_genes
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - cosine_accuracy
model-index:
  - name: SentenceTransformer based on Qwen/Qwen3-Embedding-0.6B
    results:
      - task:
          type: triplet
          name: Triplet
        dataset:
          name: >-
            cellxgene pseudo bulk 100k multiplets natural language annotation
            cell sentence 2
          type: >-
            cellxgene_pseudo_bulk_100k_multiplets_natural_language_annotation_cell_sentence_2
        metrics:
          - type: cosine_accuracy
            value: 0.8204416632652283
            name: Cosine Accuracy
      - task:
          type: triplet
          name: Triplet
        dataset:
          name: gene description
          type: gene_description
        metrics:
          - type: cosine_accuracy
            value: 0.9559999704360962
            name: Cosine Accuracy

SentenceTransformer based on Qwen/Qwen3-Embedding-0.6B

This is a sentence-transformers model finetuned from Qwen/Qwen3-Embedding-0.6B on the cellxgene_pseudo_bulk_100k_multiplets_natural_language_annotation and gene_description datasets. It maps sentences & paragraphs to a 1024-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-0.6B
Maximum Sequence Length: 32768 tokens
Output Dimensionality: 1024 dimensions
Similarity Function: Cosine Similarity
Training Datasets:
- cellxgene_pseudo_bulk_100k_multiplets_natural_language_annotation
- gene_description
Language: code

Model Sources

Documentation: Sentence Transformers Documentation
Repository: Sentence Transformers on GitHub
Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): MMContextEncoder(
    (text_encoder): Qwen3Model(
      (embed_tokens): Embedding(151669, 1024)
      (layers): ModuleList(
        (0-27): 28 x Qwen3DecoderLayer(
          (self_attn): Qwen3Attention(
            (q_proj): Linear(in_features=1024, out_features=2048, bias=False)
            (k_proj): Linear(in_features=1024, out_features=1024, bias=False)
            (v_proj): Linear(in_features=1024, out_features=1024, bias=False)
            (o_proj): Linear(in_features=2048, out_features=1024, bias=False)
            (q_norm): Qwen3RMSNorm((128,), eps=1e-06)
            (k_norm): Qwen3RMSNorm((128,), eps=1e-06)
          )
          (mlp): Qwen3MLP(
            (gate_proj): Linear(in_features=1024, out_features=3072, bias=False)
            (up_proj): Linear(in_features=1024, out_features=3072, bias=False)
            (down_proj): Linear(in_features=3072, out_features=1024, bias=False)
            (act_fn): SiLU()
          )
          (input_layernorm): Qwen3RMSNorm((1024,), eps=1e-06)
          (post_attention_layernorm): Qwen3RMSNorm((1024,), eps=1e-06)
        )
      )
      (norm): Qwen3RMSNorm((1024,), eps=1e-06)
      (rotary_emb): Qwen3RotaryEmbedding()
    )
    (pooling): Pooling({'word_embedding_dimension': 1024, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  )
)

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("jo-mengr/mmcontext-qwen-scvi_fm")
# Run inference
sentences = [
    'MALAT1 DCN MGP APOD GSN LAMA2 CST3 SPARCL1 IGFBP7 TIMP1 VIM EEF1A1 ITM2B FBLN1 C3 IFITM3 FBN1 FTH1 TPT1 ABCA8 C1S TXNIP FTL TIMP3 FN1 CD63 RBMS3 ABCA6 ZBTB20 CEBPD NEAT1 CFH VCAN PTN PTGDS CD81 SERF2 COL6A1 COL6A2 ABI3BP ABCA10 EBF1 COL1A2 PRKG1 S100A6 MGST1 TMSB10 TIMP2 CELF2 LAPTM4A RORA ACTB LTBP4 MYL6 LGALS1 DDX5 SPTBN1 EFEMP1 BICC1 LRP1 H3-3B SCN7A IGFBP4 FAU',
    'This measurement was conducted with 10x multiome. Fibroblast cell sample taken from the right ventricle of a European female donor in her fifth decade, who is a DCD donor. The sample is in nucleus form.',
    "This measurement was conducted with 10x 3' v3. CD4+T naive lymphocyte cells derived from the right cardiac atrium of a European male in his sixties.",
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.6280, 0.0951],
#         [0.6280, 1.0000, 0.2002],
#         [0.0951, 0.2002, 1.0000]])

Evaluation

Metrics

Triplet

Datasets: cellxgene_pseudo_bulk_100k_multiplets_natural_language_annotation_cell_sentence_2 and gene_description
Evaluated with TripletEvaluator

Metric	cellxgene_pseudo_bulk_100k_multiplets_natural_language_annotation_cell_sentence_2	gene_description
cosine_accuracy	0.8204	0.956

Training Details

Training Datasets

cellxgene_pseudo_bulk_100k_multiplets_natural_language_annotation

Dataset: cellxgene_pseudo_bulk_100k_multiplets_natural_language_annotation at d518eb2
Size: 81,143 training samples
Columns: anchor, positive, negative_1, and negative_2

Approximate statistics based on the first 1000 samples:

	anchor	positive	negative_1	negative_2
type	string	string	string	string
details	min: 356 characters mean: 385.24 characters max: 450 characters	min: 92 characters mean: 216.13 characters max: 900 characters	min: 103 characters mean: 212.72 characters max: 1186 characters	min: 353 characters mean: 384.82 characters max: 433 characters

Samples:

anchor	positive	negative_1	negative_2
`TMSB4X TMSB10 ACTB MALAT1 GNLY NKG7 IFITM2 LGALS1 GZMA EEF1A1 PFN1 HMGB2 FTH1 PTMA HSP90AA1 GZMB ARHGDIB HNRNPA2B1 PLAAT4 FAU CMC1 VIM MYL12A CBX3 ATP5F1E HCST IFI44L KLRF1 H3-3A COX6C ARL6IP1 CFL1 ISG15 HMGB1 S100A4 ATP5MF RORA MYL6 CORO1A OAZ1 KLRB1 ID2 HMGN3 CCNI RBM39 CAP1 SERF2 ELOC FCER1G S100A9 IFI16 YWHAZ EIF1 CALR HMGN2 SKAP2 SLC25A5 ZZZ3 YBX1 NUCB2 CDC42 GSTP1 FTL ATP5F1D`	`This measurement was conducted with 10x 3' v2. A proliferating lymphocyte cell sample, obtained from a 34-year-old female Asian individual, derived from peripheral blood mononuclear cells.`	`This measurement was conducted with 10x 3' v2. Sample is a CD8-positive, alpha-beta T cell derived from a 31-year-old Asian female's peripheral blood mononuclear cells.`	`MALAT1 TMSB4X EEF1A1 TMSB10 FAU TPT1 PTMA EIF1 UBA52 ACTB FTH1 RACK1 FTL H3-3B JUNB ATP5F1E BTG1 CD52 NACA MYL12A PFN1 COX7C COX4I1 SERF2 UQCRB TOMM7 IL32 YBX1 PABPC1 MYL6 EIF3E OAZ1 NOP53 ARHGDIB LDHB HCST SARAF ITM2B ATP6V1G1 SRP14 UBC H3-3A COX6C HINT1 UBB COMMD6 S100A4 S100A6 CALM1 VIM CYBA ENO1 HSP90AA1 FXYD5 HSP90AB1 CIRBP SRSF5 NFKBIA CORO1A LEPROTL1 TLE5 CHCHD2 DDX5 CD69`
`EEF1A1 MALAT1 FTH1 JUNB TPT1 FOS TMSB10 BTG1 TMSB4X ZFP36L2 NACA PABPC1 ACTB FAU VIM H3-3B EIF1 ZFP36 SARAF PTMA IL7R JUN RACK1 EEF2 UBA52 GAPDH FTL FXYD5 DUSP1 S100A4 CD69 CXCR4 UBC TSC22D3 CFL1 KLF6 ARHGDIB KLF2 BTG2 CITED2 IER2 TUBB4B CD3E EEF1G SLC2A3 NFKBIA PFN1 SRGN SNX9 COX4I1 DNAJB1 SERF2 CD8A PCBP2 IL32 BIRC3 SMAP2 FUS GADD45B MYL12A OAZ1 ATP5F1E TUBA4A PNRC1`	`This measurement was conducted with 10x 5' v1. Sample is a cell from the omentum tissue, specifically an effector memory CD4-positive, alpha-beta T cell, from a female in her sixth decade.`	`This measurement was conducted with 10x 5' v1. Sample is a CD4-positive helper T cell, specifically Trm_Th1/Th17 subset, derived from the duodenum tissue of a male individual in his sixth decade.`	`MALAT1 TPT1 EEF1A1 VIM JUND TMSB4X PTMA FTH1 CRIP1 ANXA1 EIF1 UBC H3-3B ACTB SRGN FTL FAU KLF6 IL7R CALM1 UBA52 BTG1 SARAF IL32 TMSB10 PABPC1 HSP90AB1 DDX5 GAPDH TAGLN2 NACA CD44 HSPA5 RORA HSP90AA1 KLRB1 TNFAIP3 ATP5F1E PNRC1 ZFP36L2 H3-3A UBB FOS RACK1 FYN FAM107B GNAS EZR MYL6 CREM NFKBIA PFN1 ARHGDIB SRSF7 CD2 CCNI HNRNPA2B1 COX7C ITM2B SERF2 SH3BGRL3 TSC22D3 LMNA YWHAZ`
`MALAT1 GRIK1 SYT1 PCDH9 RORA NRG1 CADPS ZFPM2 LRRC4C LINGO2 RALYL PTPRD SPHKAP CNTNAP5 SLC8A1 CCSER1 HDAC9 CELF2 R3HDM1 CNTN4 RBMS3 PCDH7 GALNT13 UNC5D ROBO1 SYNPR SNAP25 GPM6A ANK3 FRMPD4 CHRM2 RYR2 KHDRBS2 CADM1 CACNA1D RGS6 PDE4D DOCK4 UNC13C CDH18 FAT3 MEG3 NR2F2-AS1 HMCN1 GULP1 CAMK2D ZEB1 SYN2 DYNC1I1 OXR1 DPP10 OSBPL6 FRAS1 PPP3CA ZNF385D ZMAT4 PCBP3 HS6ST3 ERC2 PLEKHA5 CDK14 MAP2 NCOA1 ATP8A2`	`This measurement was conducted with 10x 3' v3. Neuron cell type from a 29-year-old male, specifically from the thalamic complex, specifically the thalamus (THM) - posterior nuclear complex of thalamus (PoN) - medial geniculate nuclei (MG).`	`This measurement was conducted with 10x 3' v3. Astrocyte cell type from the thalamic complex, specifically from the thalamus (THM) - posterior nuclear complex of thalamus (PoN) - medial geniculate nuclei (MG) region, of a 42-year-old male.`	`MALAT1 PCDH9 PLP1 MBP ST18 QKI PDE4B RNF220 PTPRD SEPTIN7 TTLL7 NCKAP5 GPM6B PIP4K2A MOBP SLC44A1 PTGDS PLCL1 MAP7 ELMO1 SIK3 FTH1 ZBTB20 MAN2A1 TMEM165 DOCK10 TCF12 EDIL3 ZEB2 DPYD MAP4K4 PHLPP1 TF GAB1 TRIM2 FRMD4B DNAJC6 MARCHF1 ANK3 DST AGAP1 TMEM144 NEAT1 PLEKHH1 DLG1 CRYAB ERBIN RTN4 SPP1 ATP8A1 DOCK4 SLAIN1 APP DOCK5 APBB2 SAMD12 SHTN1 ZNF536 ZFYVE16 ARAP2 LIMCH1 HIPK2 BCAS1 FAM107B`

Loss: MultipleNegativesRankingLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "cos_sim"
}

gene_description

Dataset: gene_description at dd22363
Size: 116,208 training samples
Columns: anchor, positive, and negative_1

Approximate statistics based on the first 1000 samples:

	anchor	positive	negative_1
type	string	string	string
details	min: 3 characters mean: 5.88 characters max: 12 characters	min: 16 characters mean: 367.09 characters max: 1375 characters	min: 13 characters mean: 167.33 characters max: 1375 characters

Samples:

anchor	positive	negative_1
`A1BG`	`The protein encoded by this gene is a plasma glycoprotein of unknown function. The protein shows sequence similarity to the variable regions of some immunoglobulin supergene family member proteins. [provided by RefSeq, Jul 2008]`	`A1BG antisense RNA 1`
`A1BG`	`The protein encoded by this gene is a plasma glycoprotein of unknown function. The protein shows sequence similarity to the variable regions of some immunoglobulin supergene family member proteins. [provided by RefSeq, Jul 2008]`	`G antigen 12D`
`A1BG`	`The protein encoded by this gene is a plasma glycoprotein of unknown function. The protein shows sequence similarity to the variable regions of some immunoglobulin supergene family member proteins. [provided by RefSeq, Jul 2008]`	`G antigen 12B`

Loss: MultipleNegativesRankingLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "cos_sim"
}

Evaluation Datasets

cellxgene_pseudo_bulk_100k_multiplets_natural_language_annotation

Dataset: cellxgene_pseudo_bulk_100k_multiplets_natural_language_annotation at d518eb2
Size: 9,011 evaluation samples
Columns: anchor, positive, negative_1, and negative_2

Approximate statistics based on the first 1000 samples:

	anchor	positive	negative_1	negative_2
type	string	string	string	string
details	min: 347 characters mean: 386.7 characters max: 437 characters	min: 99 characters mean: 209.99 characters max: 941 characters	min: 101 characters mean: 208.8 characters max: 728 characters	min: 356 characters mean: 386.56 characters max: 434 characters

Samples:

anchor	positive	negative_1	negative_2
`MALAT1 EEF1A1 FTH1 TMSB4X ACTB FTL RTN4 ATP6V0B TPT1 FAU S100A6 NDUFA4 ATP5F1E COX7C ITM2B IGFBP7 EIF1 C12orf75 CD9 COX7B SERF2 ATP1B1 COX8A TXNIP NDUFB2 MYL6 PPDPF COX6B1 UQCR11 APOE COX4I1 CALM2 UQCRB S100A11 UQCRQ COX6C ATP5MG BSG ATP6AP2 UQCR10 PTMA NACA UBL5 UBA52 TMSB10 ADGRF5 HSP90AA1 GSTP1 ATP5F1D CHCHD2 GAPDH COX7A2 SKP1 HSPE1 PRDX1 CYSTM1 LGALS3 CD63 ATP5MJ CKB NDUFS5 ATP5ME UBB MAL`	`This measurement was conducted with 10x 3' v3. Cell sample from the cortex of kidney, taken from a 43-year-old male of European ethnicity with a reported history of kidney cancer. The cell type is identified as a kidney collecting duct intercalated cell.`	`This measurement was conducted with 10x 3' v3. Cell sample from the cortex of kidney, taken from a 72-year-old male of European ethnicity, identified as a kidney collecting duct intercalated cell, and preserved through cryopreservation.`	`MALAT1 TMSB4X TMSB10 ACTB TXNIP EEF1A1 TPT1 PFN1 BTG1 FAU PTMA S100A4 ATP5F1E EIF1 FTL CFL1 CYBA MYL12A SRGN SERF2 SH3BGRL3 CALM1 TYROBP MYL6 ZFP36 KLRD1 UBB NACA S100A6 UBA52 HSP90AA1 H3-3B LCP1 FTH1 DDIT4 FOS PPIA CD247 RACK1 TMA7 CORO1A OAZ1 TLE5 ARPC3 GAPDH KLF2 UBC ZFP36L2 TSC22D3 ITGB2 ARPC2 ATP5MG HOPX IFITM2 HMGB1 OST4 EEF1G PRDM1 CDC42 GSTP1 NDUFB2 CIRBP LGALS1 CHCHD2`
`MALAT1 KCND2 NRXN1 CDH18 NRXN3 ZNF385D CADM2 RALYL NKAIN2 CADPS2 RIMS1 FSTL5 GRID2 TRPM3 CHN2 DPP6 JMJD1C RORA PDE1A UNC13C TIAM1 NRG1 SNAP25 ZFPM2 CALN1 LSAMP CNTN1 ABLIM1 SYNE1 ANK3 CA10 NFIA ZBTB20 NTM CADM1 OPCML RELN DNM3 NEBL ERC1 SCN2A PPP3CA CACNA1A GALNT13 LRRC4C GPM6A RABGAP1L RIT2 CAMK4 GRIA4 PTPRD RBFOX3 MCTP1 LHFPL6 PCLO MEG3 PDE10A NOVA1 RTN1 ZNF385B CNTN4 GABRB2 SPOCK1 OXR1`	`This measurement was conducted with 10x 3' v3. Neuron cell type from a 29-year-old male cerebellum, specifically from the Cerebellar Vermis - CBV region, with European self-reported ethnicity, analyzed at the nucleus level.`	`This measurement was conducted with 10x 3' v3. Sample is an oligodendrocyte precursor cell taken from the cerebellum tissue of a 42-year-old human male, specifically from the Cerebellum (CB) - Cerebellar Vermis - CBV dissection.`	`MALAT1 NRXN3 SNTG1 UNC5C GRIA4 NRG1 RORA INPP4B CLSTN2 NKAIN2 FRMD4A DPP6 GRID2 NRXN1 LSAMP JMJD1C HS6ST3 NXPH1 MIR99AHG LRRC4C NTM CCNH NFIA ZFPM2 AFF3 OPCML PTPRT CADM2 ZBTB20 OLFM3 SLC22A3 CNTNAP5 CACNA2D3 CNTN4 KCND2 ADARB2 XKR4 GPM6A IL1RAPL1 ALK ANKRD36C UBE2E2 SYN3 GARNL3 PTPRG DAB1 TCF4 LINC00461 PRANCR GRIN2B TNRC6B MAPK10 NOVA1 NFIB ANK3 KCNMA1 KCNQ5 SPON1 TRIM9 VWA8 GDAP1 GABRG2 AHI1 ATP1B1`
`EEF1A1 ACTB GAPDH HMGN2 PTMA SERF2 TMSB4X CD74 PABPC1 FTH1 TMSB10 FAU PFN1 HMGN1 OAZ1 HMGB1 TPT1 PPIA NACA BTF3 MALAT1 MYL6 ATP5MG CFL1 RACK1 ODC1 ATP5F1E TMA7 SLC25A5 ELOB ARPC3 NPM1 COX7C ANP32B C4orf3 EIF1 PCBP2 KLF6 LAPTM5 COX8A RHOA HSPA8 H3-3B PTP4A2 UBA52 OST4 CIRBP LGALS1 EIF3L STMN1 PPDPF COX4I1 RAN EIF3F PPP1CC COMMD6 NDUFA4 YBX1 PEBP1 COTL1 COX7A2 HSPE1 CCNI TRIR`	`This measurement was conducted with 10x 5' v1. Cell sample from the tonsil of a 9-year-old female with recurrent tonsillitis, characterized as a centroblast B cell with IGLC2, IGLV7-43, IGLJ3 immunoglobulin genes expressed.`	`This measurement was conducted with 10x 5' v1. Germinal center B cell derived from the tonsil tissue of a 3-year-old male with recurrent tonsillitis.`	`CD74 MALAT1 EEF1A1 SSR4 TPT1 UBC EEF2 SAT1 RACK1 SEC11C ATP5MG FAU TSC22D3 PPIB XBP1 FTL GAPDH HLA-DRB5 HERPUD1 RGS2 HSPA8 TMSB4X HSP90B1 EIF1 PTMA SERP1 SERF2 NACA SEC61B GSTP1 UBA52 HSPA5 BTF3 LAPTM5 HSPE1 H3-3B ATP5F1A SEC61G CD38 EDF1 FTH1 IL16 NPM1 OST4 CIRBP EIF3E OAZ1 CYTIP PCBP2 MYDGF COX6B1 ZFP36 CSDE1 PABPC1 REXO2 KDELR1 PFN1 PTP4A1 TMBIM6 H1-10 PSAP UBE2J1 VIM MYL6`

Loss: MultipleNegativesRankingLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "cos_sim"
}

gene_description

Dataset: gene_description at dd22363
Size: 1,000 evaluation samples
Columns: anchor, positive, and negative_1

Approximate statistics based on the first 1000 samples:

	anchor	positive	negative_1
type	string	string	string
details	min: 3 characters mean: 5.88 characters max: 12 characters	min: 16 characters mean: 367.09 characters max: 1375 characters	min: 13 characters mean: 167.33 characters max: 1375 characters

Samples:

anchor	positive	negative_1
`A1BG`	`The protein encoded by this gene is a plasma glycoprotein of unknown function. The protein shows sequence similarity to the variable regions of some immunoglobulin supergene family member proteins. [provided by RefSeq, Jul 2008]`	`A1BG antisense RNA 1`
`A1BG`	`The protein encoded by this gene is a plasma glycoprotein of unknown function. The protein shows sequence similarity to the variable regions of some immunoglobulin supergene family member proteins. [provided by RefSeq, Jul 2008]`	`G antigen 12D`
`A1BG`	`The protein encoded by this gene is a plasma glycoprotein of unknown function. The protein shows sequence similarity to the variable regions of some immunoglobulin supergene family member proteins. [provided by RefSeq, Jul 2008]`	`G antigen 12B`

Loss: MultipleNegativesRankingLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "cos_sim"
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 128
per_device_eval_batch_size: 128
learning_rate: 2e-05
num_train_epochs: 4
warmup_ratio: 0.1
bf16: True
gradient_checkpointing: 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: 128
per_device_eval_batch_size: 128
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: 2e-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: 4
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.1
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: 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
hub_revision: None
gradient_checkpointing: True
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: False
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: False
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Click to expand

Epoch	Step	Training Loss	cellxgene pseudo bulk 100k multiplets natural language annotation loss	gene description loss	cellxgene_pseudo_bulk_100k_multiplets_natural_language_annotation_cell_sentence_2_cosine_accuracy	gene_description_cosine_accuracy
0.0324	50	9.3314	12.6479	6.6616	0.5052	0.2570
0.0649	100	7.9528	10.8869	6.0596	0.5078	0.2660
0.0973	150	7.0084	7.0423	5.4704	0.5075	0.3020
0.1297	200	5.6925	6.0263	5.2950	0.5024	0.5200
0.1621	250	5.381	5.8141	4.7323	0.5367	0.6520
0.1946	300	4.3736	5.4432	4.3565	0.5518	0.7060
0.2270	350	3.8184	5.1966	4.1283	0.5836	0.7690
0.2594	400	3.6181	5.0588	3.9594	0.6064	0.7650
0.2918	450	3.1076	4.9406	3.7824	0.6218	0.8030
0.3243	500	3.127	4.8376	3.6785	0.6369	0.8230
0.3567	550	3.1702	4.8230	3.6029	0.6532	0.8410
0.3891	600	2.992	5.1160	3.6091	0.6240	0.8310
0.4215	650	2.606	4.5652	3.5555	0.6679	0.8490
0.4540	700	2.9473	4.5831	3.5215	0.6846	0.8600
0.4864	750	2.369	4.4464	3.4824	0.6930	0.8800
0.5188	800	2.5923	4.4542	3.4372	0.6983	0.8820
0.5512	850	2.9167	4.4572	3.4915	0.6984	0.8730
0.5837	900	2.5716	4.2259	3.4390	0.7126	0.8630
0.6161	950	2.375	4.2200	3.4250	0.7143	0.8740
0.6485	1000	2.4105	4.2001	3.3524	0.7187	0.8890
0.6809	1050	2.4014	4.0744	3.2688	0.7243	0.8950
0.7134	1100	2.7474	4.1131	3.3046	0.7270	0.8850
0.7458	1150	2.1615	4.2206	3.2392	0.7202	0.8860
0.7782	1200	2.4409	4.4682	3.1664	0.7106	0.8870
0.8106	1250	2.5041	4.0881	3.1417	0.7277	0.9030
0.8431	1300	2.4221	3.8777	3.2302	0.7409	0.8940
0.8755	1350	2.189	3.8482	3.1316	0.7441	0.9050
0.9079	1400	2.3055	3.8571	3.1550	0.7451	0.9030
0.9403	1450	2.0945	3.8233	3.1269	0.7530	0.9020
0.9728	1500	2.0217	3.7722	3.0707	0.7527	0.9070
1.0052	1550	2.2443	3.8285	3.0799	0.7459	0.9190
1.0376	1600	1.9441	3.8292	3.0957	0.7470	0.9090
1.0700	1650	1.8771	3.6837	3.0190	0.7555	0.9290
1.1025	1700	1.9489	3.6946	3.0298	0.7570	0.9210
1.1349	1750	2.0622	3.7221	3.0001	0.7574	0.9140
1.1673	1800	1.7275	3.7806	2.9919	0.7530	0.9090
1.1997	1850	2.0068	3.6648	2.9490	0.7584	0.9230
1.2322	1900	1.9126	3.7416	2.9131	0.7603	0.9160
1.2646	1950	1.9513	3.5770	2.9362	0.7625	0.9230
1.2970	2000	1.8021	3.6660	2.8868	0.7670	0.9360
1.3294	2050	1.9685	3.7318	2.8669	0.7587	0.9390
1.3619	2100	1.7835	3.5471	2.8356	0.7712	0.9350
1.3943	2150	1.826	3.5666	2.7893	0.7707	0.9340
1.4267	2200	1.9708	3.5630	2.7570	0.7741	0.9290
1.4591	2250	2.0131	3.5586	2.8239	0.7742	0.9360
1.4916	2300	1.856	3.5155	2.7658	0.7779	0.9410
1.5240	2350	1.9354	3.7959	2.7921	0.7622	0.9380
1.5564	2400	1.8961	3.5166	2.7456	0.7790	0.9430
1.5888	2450	1.6347	3.4784	2.7911	0.7800	0.9470
1.6213	2500	1.9176	3.4388	2.7349	0.7829	0.9440
1.6537	2550	2.0475	3.6968	2.7456	0.7754	0.9390
1.6861	2600	1.7946	3.4758	2.7046	0.7848	0.9470
1.7185	2650	1.9581	3.3828	2.7022	0.7867	0.9430
1.7510	2700	1.8475	3.3631	2.6706	0.7903	0.9470
1.7834	2750	1.836	3.5622	2.6512	0.7857	0.9450
1.8158	2800	2.051	3.3523	2.6542	0.7926	0.9390
1.8482	2850	1.829	3.3676	2.6730	0.7925	0.9390
1.8807	2900	1.7557	3.3632	2.6536	0.7954	0.9470
1.9131	2950	1.7725	3.3448	2.6437	0.7946	0.9470
1.9455	3000	1.7373	3.2736	2.6562	0.7987	0.9440
1.9780	3050	1.886	3.3404	2.6456	0.7958	0.9450
2.0104	3100	1.7217	3.2570	2.6893	0.7988	0.9400
2.0428	3150	1.6235	3.2331	2.6132	0.8004	0.9430
2.0752	3200	1.6678	3.2466	2.5904	0.8030	0.9470
2.1077	3250	1.6784	3.2339	2.5956	0.8008	0.9480
2.1401	3300	1.8422	3.2286	2.5997	0.8039	0.9480
2.1725	3350	1.4859	3.2163	2.5924	0.8049	0.9470
2.2049	3400	1.6165	3.3246	2.6167	0.7989	0.9440
2.2374	3450	1.65	3.2184	2.5864	0.8039	0.9460
2.2698	3500	1.5071	3.2274	2.5788	0.8019	0.9460
2.3022	3550	1.5238	3.2032	2.5608	0.8075	0.9480
2.3346	3600	1.568	3.2409	2.5649	0.8081	0.9470
2.3671	3650	1.4644	3.1937	2.5841	0.8079	0.9430
2.3995	3700	1.5782	3.2033	2.5909	0.8065	0.9450
2.4319	3750	1.6976	3.1905	2.5690	0.8073	0.9470
2.4643	3800	1.4682	3.2078	2.5610	0.8052	0.9490
2.4968	3850	1.7414	3.1822	2.5650	0.8072	0.9500
2.5292	3900	1.654	3.1890	2.5566	0.8110	0.9490
2.5616	3950	1.5187	3.1843	2.5508	0.8090	0.9470
2.5940	4000	1.4893	3.1855	2.5527	0.8067	0.9470
2.6265	4050	1.6716	3.1520	2.5432	0.8093	0.9480
2.6589	4100	1.4914	3.1868	2.5466	0.8099	0.9500
2.6913	4150	1.6231	3.1702	2.5235	0.8112	0.9500
2.7237	4200	1.6058	3.1561	2.5171	0.8096	0.9520
2.7562	4250	1.5753	3.1660	2.5068	0.8111	0.9530
2.7886	4300	1.4654	3.1507	2.5156	0.8138	0.9510
2.8210	4350	1.5901	3.1960	2.4917	0.8115	0.9540
2.8534	4400	1.5034	3.1491	2.4960	0.8116	0.9550
2.8859	4450	1.4088	3.1505	2.5086	0.8133	0.9530
2.9183	4500	1.5527	3.1671	2.5154	0.8112	0.9540
2.9507	4550	1.5344	3.1329	2.5016	0.8141	0.9530
2.9831	4600	1.4156	3.1439	2.4858	0.8146	0.9550
3.0156	4650	1.8602	3.1056	2.4799	0.8163	0.9550
3.0480	4700	1.4472	3.1387	2.4539	0.8126	0.9540
3.0804	4750	1.3582	3.1220	2.4676	0.8159	0.9530
3.1128	4800	1.5408	3.1309	2.4722	0.8142	0.9540
3.1453	4850	1.3755	3.1227	2.4624	0.8171	0.9530
3.1777	4900	1.4571	3.1284	2.4410	0.8162	0.9560
3.2101	4950	1.5657	3.0882	2.4486	0.8167	0.9550
3.2425	5000	1.5325	3.0980	2.4339	0.8178	0.9540
3.2750	5050	1.4671	3.0961	2.4625	0.8169	0.9550
3.3074	5100	1.4808	3.1176	2.4578	0.8180	0.9550
3.3398	5150	1.4172	3.1338	2.4515	0.8168	0.9550
3.3722	5200	1.4953	3.1047	2.4425	0.8174	0.9540
3.4047	5250	1.6419	3.1081	2.4317	0.8180	0.9540
3.4371	5300	1.5425	3.0910	2.4481	0.8210	0.9560
3.4695	5350	1.5598	3.1049	2.4365	0.8198	0.9560
3.5019	5400	1.4086	3.1036	2.4352	0.8198	0.9550
3.5344	5450	1.6057	3.1076	2.4269	0.8197	0.9560
3.5668	5500	1.6735	3.0792	2.4291	0.8200	0.9550
3.5992	5550	1.401	3.0959	2.4364	0.8211	0.9550
3.6316	5600	1.2475	3.0909	2.4324	0.8202	0.9570
3.6641	5650	1.2495	3.0686	2.4148	0.8210	0.9550
3.6965	5700	1.4457	3.0837	2.4123	0.8197	0.9570
3.7289	5750	1.5794	3.0877	2.4171	0.8191	0.9560
3.7613	5800	1.5696	3.0936	2.4153	0.8186	0.9560
3.7938	5850	1.5947	3.0778	2.4173	0.8190	0.9560
3.8262	5900	1.4517	3.0760	2.4242	0.8202	0.9560
3.8586	5950	1.553	3.0897	2.4222	0.8188	0.9580
3.8911	6000	1.2109	3.0683	2.4233	0.8211	0.9550
3.9235	6050	1.4384	3.0756	2.4221	0.8208	0.9560
3.9559	6100	1.4945	3.0755	2.4179	0.8202	0.9560
3.9883	6150	1.4597	3.0686	2.4183	0.8204	0.9560

Framework Versions

Python: 3.11.6
Sentence Transformers: 5.0.0
Transformers: 4.55.0.dev0
PyTorch: 2.5.1+cu121
Accelerate: 1.9.0
Datasets: 2.19.1
Tokenizers: 0.21.4

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",
}

MultipleNegativesRankingLoss

@misc{henderson2017efficient,
    title={Efficient Natural Language Response Suggestion for Smart Reply},
    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
    year={2017},
    eprint={1705.00652},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}