numind/NuExtract-1.5
Full fine-tune vs microsoft/Phi-3.5-mini-instruct. Fine-tune: 100% of the weights changed, across 32 of 32 layers.
Tensors changed: Embeddings 1/1, Norms 17/65, MLP 64/64, Attention 64/64, lm_head 1/1
4 derivatives checked
- lmstudio-community/NuExtract-v1.5-GGUF: Quantization
- bartowski/NuExtract-v1.5-GGUF: Same weights, converted
- gradients-io-tournaments/tournament-tourn_c8068e356b7c22c8_20260928-880b38e9-b34b-46c3-b3c5-7cc5b08c18b9-5E6p1x3e: Partial fine-tune
- DevQuasar-3/numind.NuExtract-v1.5-GGUF: Quantization
3 other repos hold byte-for-byte identical weight files.