Upload BROKEN-TEST-ONLY-DONT-DOWNLOAD-MiniMax-M2.7-iq1_s_q4_K.gguf with huggingface_hub

Upload folder using huggingface_hub

uploading smol-IQ4_KSS and KLD data and logs

add perplexity data

add speculative decode tips and upload smol-IQ3_KS

Upload folder using huggingface_hub

Update best working examples in quickstart

Upload folder using huggingface_hub

add reference to ik_llama.cpp PR 1624 gibberish

upload IQ2_KS for full offload in 96GB VRAM

add some early logs

Upload folder using huggingface_hub

dont' use -muge for right now, but looks okay without

Upload imatrix-MiniMax-M2.7-BF16.dat with huggingface_hub

initial commit