Add qk256fa5b checkpoint, step 50000 (continuation of the released model, our DVLT-codebase run)
#2
by aburzio-nv - opened
plain model.state_dict() of da3loop.model.LoopModel (546 tensors, fp32); strict-loads into the qk256fa5b branch model. No optimizer state. sha256 2204dc9eebab5862b9f6f12079446be50f1fadfbc728561ce43f203abff37334
Yusin changed pull request status to merged