Exported in GGUF format, INT4 quantization
Refined version of my previous models, with new training data and methodology. This should produce more natural reponses in Spanish.
Base model Mistral-7b
Based on the excelent job of senseable/WestLake-7B-v2 and Eric Hartford's cognitivecomputations/WestLake-7B-v2-laser
Fine-tuned in Spanish with a collection of poetry, books, wikipedia articles, phylosophy texts and alpaca-es datasets.
Trained using Lora and PEFT and INT8 quantization on 2 GPUs for several days.

Usage:

Use in llamacpp or other framework that supports GGUF format.

ecastera/ecastera-eva-westlake-7b-spanish-int4-gguf