I'd assume that 33B model should fit with this(only repo that I know of that implements SparseGPT and GPTQ for LLaMa), I, personally, haven't tried though. But

I'd assume that 33B model should fit with this(only repo that I know of that implements SparseGPT and GPTQ for LLaMa), I, personally, haven't tried though. But you can try your luck https://github.com/lachlansneff/sparsellama https://github.com/lachlansneff/sparsellama