Is FlexGen+GPTQ 4bit possible?

[BarfingLemurs]

Just a curious question I suppose!
GPTQ 4bit - https://github.com/qwopqwop200/GPTQ-for-LLaMa
Suppose someone eventually finetunes 175B OPT model, with loras or regular finetunng. or perhaps the BLOOM or BLOOMZ model, would running inference be possible with GPTQ to allow the model to be run on 4gbvram and 50gb dram?

[Ying1123]

1. FlexGen has support for 4-bit compression, see sec 5 in paper, and weights compression

   FlexGen/flexgen/flex_opt.py

   Line 1306 in bbc9ea9

   parser.add_argument("--compress-weight", action="store_true",

   cache compression

   FlexGen/flexgen/flex_opt.py

   Line 1308 in bbc9ea9

   parser.add_argument("--compress-cache", action="store_true",

2. The compression in FlexGen has computation overhead, so it is not always better to turn it on. For large models like 175B which involves disk swap, it is usually better to turn on both weights and cache compression.
3. GPTQ 4bit has not been implemented in FlexGen.
4. Even you use 4bit, the weights of an 175B model need to occupy ~90G memory. 4GB vram and 50GB dram is not sufficient.