Books from the Survivor Library (mostly ~1920s & earlier) OCR'd with recent VLMs
BEEspoke Data
community
AI & ML interests
'an LLM is only as good as the dataset it was trained on' - Sun Tzu
smol_llama 220M fine-tunes we did
-
BEE-spoke-data/smol_llama-220M-openhermes
Text Generation • 0.2B • Updated • 1.16k • 5 -
BEE-spoke-data/smol_llama-220M-open_instruct
Text Generation • 0.2B • Updated • 25 • 2 -
BEE-spoke-data/beecoder-220M-python
Text Generation • 0.2B • Updated • 28 • 3 -
BEE-spoke-data/zephyr-220m-sft-full
Text Generation • 0.2B • Updated • 1.09k • 1
models fine-tuned to be knowledgeable about apiary practice
trained and adapted tokenizers - various
🚧"raw" pretrained smol_llama checkpoints - WIP 🚧
-
BEE-spoke-data/smol_llama-101M-GQA
Text Generation • 0.1B • Updated • 2.98k • 30 -
BEE-spoke-data/smol_llama-81M-tied
Text Generation • 81.3M • Updated • 1.17k • 9 -
BEE-spoke-data/smol_llama-220M-GQA
Text Generation • 0.2B • Updated • 3.3k • 13 -
BEE-spoke-data/verysmol_llama-v11-KIx2
Text Generation • 58.1M • Updated • 1.15k • 4
Pretrained encoder (fill-mask) models we made
text classification models for book genres
-
BEE-spoke-data/albert-xxlarge-v2-description2genre
Text Classification • 0.2B • Updated • 24 • 2 -
BEE-spoke-data/mobilebert-uncased-title2genre
Text Classification • 24.6M • Updated • 34 • 1 -
BEE-spoke-data/roberta-large-title2genre
Text Classification • 0.4B • Updated • 33 • 1 -
BEE-spoke-data/roberta-base-description2genre
Text Classification • 0.1B • Updated • 19
concept datasets extracted from fineweb
Books from the Survivor Library (mostly ~1920s & earlier) OCR'd with recent VLMs
🚧"raw" pretrained smol_llama checkpoints - WIP 🚧
-
BEE-spoke-data/smol_llama-101M-GQA
Text Generation • 0.1B • Updated • 2.98k • 30 -
BEE-spoke-data/smol_llama-81M-tied
Text Generation • 81.3M • Updated • 1.17k • 9 -
BEE-spoke-data/smol_llama-220M-GQA
Text Generation • 0.2B • Updated • 3.3k • 13 -
BEE-spoke-data/verysmol_llama-v11-KIx2
Text Generation • 58.1M • Updated • 1.15k • 4
smol_llama 220M fine-tunes we did
-
BEE-spoke-data/smol_llama-220M-openhermes
Text Generation • 0.2B • Updated • 1.16k • 5 -
BEE-spoke-data/smol_llama-220M-open_instruct
Text Generation • 0.2B • Updated • 25 • 2 -
BEE-spoke-data/beecoder-220M-python
Text Generation • 0.2B • Updated • 28 • 3 -
BEE-spoke-data/zephyr-220m-sft-full
Text Generation • 0.2B • Updated • 1.09k • 1
Pretrained encoder (fill-mask) models we made
models fine-tuned to be knowledgeable about apiary practice
text classification models for book genres
-
BEE-spoke-data/albert-xxlarge-v2-description2genre
Text Classification • 0.2B • Updated • 24 • 2 -
BEE-spoke-data/mobilebert-uncased-title2genre
Text Classification • 24.6M • Updated • 34 • 1 -
BEE-spoke-data/roberta-large-title2genre
Text Classification • 0.4B • Updated • 33 • 1 -
BEE-spoke-data/roberta-base-description2genre
Text Classification • 0.1B • Updated • 19
trained and adapted tokenizers - various
concept datasets extracted from fineweb