Dataset Viewer
Auto-converted to Parquet Duplicate
stages
dict
eval
dict
storage
dict
samples
dict
{ "model_load": { "seconds": 54.8586, "embedder": "Qwen/Qwen3-VL-Embedding-2B" }, "ingest_text": { "seconds": 8.0462, "chunks": 38 }, "ingest_images": { "seconds": 2.7866, "images": 12 }, "ingest_3d_views": { "seconds": 0.2765, "views": 6 }, "ingest_video": { "secon...
{ "text_recall": { "n": 10, "ann_binary_top200_recall@5": 1, "float_mrl512_top20_recall@5": 1, "cross_encoder_recall@5": 1 }, "text_to_image_top5": { "n": 4, "hits": 4 }, "text_to_audio_top1": { "n": 4, "hits": 4 }, "asset_view_grouping": { "views_in_top6": 4, "expe...
{ "chunks": 68, "binary_index_bytes": 4352, "float_store_bytes": 139264, "full_float_would_be_bytes": 557056, "storage_compression_vs_full_float": 3.88 }
{ "audio_sr": 44100, "stt_transcript": "Mr. Quilter is the apostle of the middle classes and we are glad to welcome his gospel.", "text_qa_example": { "question": "Which planet has the hottest surface in the solar system and why?", "gold": "text:corpus:0", "ann_top1": "text:corpus:0", "reranked_to...

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

AnyModal RAG — validated end-to-end prototype

Reference run: job 6ab1136251992417dfccf413 (a10g-small, 4m35s wall-clock, inference only).

Stack (all open weights)

  • Unified embedder: Qwen/Qwen3-VL-Embedding-2B — one 2048-d space for text, images, video keyframes, 3D-proxy views
  • Reranker: Qwen/Qwen3-Reranker-0.6B (cross-encoder)
  • Audio: laion/larger_clap_general (Apache-2.0 audio/text space), faster-whisper tiny for STT
  • Generator: Qwen/Qwen3-1.7B (decoder-only, citations pinned to chunk ids)
  • Index: sqlite-vec, binary (1-bit/dim) top-200 → float MRL-512 top-20 → cross-encoder top-5

Measured results

Metric Value
Text retrieval: binary ANN top-200 recall@5 1.0
Float MRL-512 top-20 recall@5 1.0
Cross-encoder recall@5 1.0
Text→image hits@5 4/4
Text→audio hits@1 4/4
Storage compression vs full float 3.88× (68 chunks)
Query pipeline latency (embed→ANN→float→cross-encoder) 146 ms total (binary ANN 1.1 ms, float rerank 9.7 ms, cross-encoder 135 ms)
Query embedding 6.4 ms
STT (faster-whisper tiny) correct LibriSpeech transcript
RAG answers 3/3 correct, grounded, with citations

Files

  • anyrag_demo.py — full pipeline (ingest → embed → index → retrieve → rerank → generate), all modalities in one script
  • metrics.json — raw metrics from the passing v6 run

Full design doc: see conversation blueprint artifact (stage × modality matrix, architecture rationale, cost analysis).

Downloads last month
61