matbee
/

kimodo-llm2vec-nf4

Feature Extraction

sentence-similarity

4-bit precision

Model card Files Files and versions

kimodo-llm2vec-nf4 / qbench_export.json

matbee's picture

Initial NF4-quantized LLM2Vec export for Kimodo

c9fb27c verified 12 days ago

history blame contribute delete

471 Bytes

	{
	"quantization": "nf4",
	"source_base": "McGill-NLP/LLM2Vec-Meta-Llama-3-8B-Instruct-mntp",
	"source_supervised": "McGill-NLP/LLM2Vec-Meta-Llama-3-8B-Instruct-mntp-supervised",
	"torch": "2.6.0+cu124",
	"cuda": "12.4",
	"method": "Snapshot-downloaded base WITHOUT adapter, bnb-quantized via LlamaBiModel.from_pretrained, save_pretrained. MNTP adapter files copied into base/ so reload mirrors Hub flow. Supervised adapter in sibling dir for LLM2VEC_LOCAL_PEFT."
	}