Revert "Add tensor split support for llama.cpp (oobabooga#3171)"

This reverts commit 031fe72.

Add llama-2-70b GGML support (oobabooga#3285)

Merge branch 'dev'

Merge pull request oobabooga#3195 from oobabooga/dev

v1.3

Merge pull request oobabooga#3163 from oobabooga/dev

v1.2

Fix lora download folder

Merge pull request oobabooga#3116 from oobabooga/dev

v1.1

Bump cpp llama version (oobabooga#3081)

Bump cpp llama version to 0.1.70

auto-assign gpu memory with --auto-devices alone