how big is your RAM

[Sorezza]

I trained on 4V100 32g and got OOM . If i set offload to cpu ,the process will be killed because of out of RAM，So how large is your RAM ? And is it possible to train llama 7B with 4v100? Really thanks for reply.

[AetherCortex]

500GB

[huiyangzhou]

what if i only have 300G, how could i do the traning

[huiyangzhou]

500GB

i have 8 v100 300G , i try to training 13B llama, what should i do ? can i adjust the config to do the training?

[cameronbergh]

i cant get past the tokenizing part its using 98gb of ram and one cpu core.

[huiyangzhou]

500GB

i have 8 v100 totaly 300G , i try to training 13B llama but get oom , what should i do ? can i adjust the config to do the training?@AetherCortex

[TZWwww]

I trained on 4_V100 32g and got OOM . If i set offload to cpu ,the process will be killed because of out of RAM，So how large is your RAM ? And is it possible to train llama 7B with 4_v100? Really thanks for reply.

Hello, I've got the same computation resources as you. Do you figure out any solutions?