Hugging Face's logo

handwoven8588
/
CodeRankEmbed-flash-attn

sentence-transformers
Safetensors
English
nomic_bert
flash-attention
code-retrieval
nomic-bert
bf16
custom_code
Model card Files
xet
 Community
2
CodeRankEmbed-flash-attn / README.md

Commit History

28 days ago

  1. Card: long-format tables (model / attention / compiled); cornstack only

    handwoven8588
    Verified signature
    2ea1968

  2. Card: one GPU (RTX 3090 Ti); tables without the 5090 columns

    handwoven8588
    Verified signature
    c417c7a

29 days ago

  1. Card: every comparison against nomic-ai/CodeRankEmbed; plain wording; citation fixed

    handwoven8588
    Verified signature
    7e68637

  2. Padding-free encoder, flattening opt-in, torch.compile recipe; card measured on public corpora (#1)⁠

    handwoven8588
    540e0bb

30 days ago

  1. Card: parity table at 1954de4 and real-document results at the model defaults

    handwoven8588
    Verified signature
    5265f03

  1. Card: table generated from idle-bracketed gate runs at 1954de4; state what encode peak measures and how much wall time varies

    handwoven8588
    Verified signature
    d1021c8

  2. Pin weights to the requested revision; harden the varlen probe; card: all tiers measured on one GPU with stated batch sizes

    handwoven8588
    Verified signature
    1954de4

  1. Auto mode probes the varlen kernel and falls back; eager accepts a 2-D mask

    handwoven8588
    Verified signature
    4c04d49

  2. Three-tier attention: torch varlen, flash_attn, eager (inline mask)

    handwoven8588
    Verified signature
    8187d6a

  1. README: drop pipeline_tag (feature-extraction)

    handwoven8588
    Verified signature
    e361c6f

  2. README: reframe bf16 weights — reason is flash_attn half-precision + runtime-bf16 (not download size)

    handwoven8588
    Verified signature
    aa2548a

  3. README pass: base_model_relation=quantized, fix CPU-eager wording (bf16, not bit-identical), drop internal refs + first-person plural

    handwoven8588
    Verified signature
    7961156

  4. README: fill 3090 Ti perf table (bf16 flash 2.1GB/162k tok/s vs fp32 eager 6.7GB/52k tok/s; cosine 0.9986)

    handwoven8588
    Verified signature
    b283422

  5. v2: bf16 weights (547->274MB) + from_pretrained torch_dtype fix (loads bf16 natively) + corrected model tree (base_model=nomic-ai/CodeRankEmbed only) + bf16-derivative README

    handwoven8588
    Verified signature
    581207a

  1. CodeRankEmbed with native flash-attn varlen forward (derivative of nomic-ai/CodeRankEmbed; identical weights; flash-vs-fp32 parity cosine 0.99999, eager fallback bit-identical)

    handwoven8588
    Verified signature
    22d2b3c