Closed zhaowujie closed 5 years ago
+1 It trains for a few hours and then it runs out of memory. Is there some memory leak happening?
Hi, you can try changing "timestep_per_batch" in training config (config.lua) to a smaller value to reduce memory consumption.
I'll try , thanks a lot!
I only have a 12GB gpu, how to modify the parameters to avoid the problem of out of memory @fanyix