Alexander Korolev 71163f27e6 kindlingai glm-5.3-flash-gx10: vendor their stack (mode B — image+patches under our rendezvous) + 3 transcribed recipes (TP4/TP2/TP6): entrypoint exec-args + all 9 experimental overlay envs 1:1, per-TP mimics of their case block (pins 26/8/26 GiB, MML 524288/163840/524288, batched 16384/8192/16384), DFlash2 spec + AdaptiveK + tp6 padded checkpoint; build.sh bakes the 39 overlay files into the image (no per-node mounts) 1 месяц назад
..
bench 71163f27e6 kindlingai glm-5.3-flash-gx10: vendor their stack (mode B — image+patches under our rendezvous) + 3 transcribed recipes (TP4/TP2/TP6): entrypoint exec-args + all 9 experimental overlay envs 1:1, per-TP mimics of their case block (pins 26/8/26 GiB, MML 524288/163840/524288, batched 16384/8192/16384), DFlash2 spec + AdaptiveK + tp6 padded checkpoint; build.sh bakes the 39 overlay files into the image (no per-node mounts) 1 месяц назад
texts 71163f27e6 kindlingai glm-5.3-flash-gx10: vendor their stack (mode B — image+patches under our rendezvous) + 3 transcribed recipes (TP4/TP2/TP6): entrypoint exec-args + all 9 experimental overlay envs 1:1, per-TP mimics of their case block (pins 26/8/26 GiB, MML 524288/163840/524288, batched 16384/8192/16384), DFlash2 spec + AdaptiveK + tp6 padded checkpoint; build.sh bakes the 39 overlay files into the image (no per-node mounts) 1 месяц назад
conc.py 71163f27e6 kindlingai glm-5.3-flash-gx10: vendor their stack (mode B — image+patches under our rendezvous) + 3 transcribed recipes (TP4/TP2/TP6): entrypoint exec-args + all 9 experimental overlay envs 1:1, per-TP mimics of their case block (pins 26/8/26 GiB, MML 524288/163840/524288, batched 16384/8192/16384), DFlash2 spec + AdaptiveK + tp6 padded checkpoint; build.sh bakes the 39 overlay files into the image (no per-node mounts) 1 месяц назад
conc_workload.py 71163f27e6 kindlingai glm-5.3-flash-gx10: vendor their stack (mode B — image+patches under our rendezvous) + 3 transcribed recipes (TP4/TP2/TP6): entrypoint exec-args + all 9 experimental overlay envs 1:1, per-TP mimics of their case block (pins 26/8/26 GiB, MML 524288/163840/524288, batched 16384/8192/16384), DFlash2 spec + AdaptiveK + tp6 padded checkpoint; build.sh bakes the 39 overlay files into the image (no per-node mounts) 1 месяц назад
count.py 71163f27e6 kindlingai glm-5.3-flash-gx10: vendor their stack (mode B — image+patches under our rendezvous) + 3 transcribed recipes (TP4/TP2/TP6): entrypoint exec-args + all 9 experimental overlay envs 1:1, per-TP mimics of their case block (pins 26/8/26 GiB, MML 524288/163840/524288, batched 16384/8192/16384), DFlash2 spec + AdaptiveK + tp6 padded checkpoint; build.sh bakes the 39 overlay files into the image (no per-node mounts) 1 месяц назад
decode.py 71163f27e6 kindlingai glm-5.3-flash-gx10: vendor their stack (mode B — image+patches under our rendezvous) + 3 transcribed recipes (TP4/TP2/TP6): entrypoint exec-args + all 9 experimental overlay envs 1:1, per-TP mimics of their case block (pins 26/8/26 GiB, MML 524288/163840/524288, batched 16384/8192/16384), DFlash2 spec + AdaptiveK + tp6 padded checkpoint; build.sh bakes the 39 overlay files into the image (no per-node mounts) 1 месяц назад
decode_stream.py 71163f27e6 kindlingai glm-5.3-flash-gx10: vendor their stack (mode B — image+patches under our rendezvous) + 3 transcribed recipes (TP4/TP2/TP6): entrypoint exec-args + all 9 experimental overlay envs 1:1, per-TP mimics of their case block (pins 26/8/26 GiB, MML 524288/163840/524288, batched 16384/8192/16384), DFlash2 spec + AdaptiveK + tp6 padded checkpoint; build.sh bakes the 39 overlay files into the image (no per-node mounts) 1 месяц назад
nll-ref.json 71163f27e6 kindlingai glm-5.3-flash-gx10: vendor their stack (mode B — image+patches under our rendezvous) + 3 transcribed recipes (TP4/TP2/TP6): entrypoint exec-args + all 9 experimental overlay envs 1:1, per-TP mimics of their case block (pins 26/8/26 GiB, MML 524288/163840/524288, batched 16384/8192/16384), DFlash2 spec + AdaptiveK + tp6 padded checkpoint; build.sh bakes the 39 overlay files into the image (no per-node mounts) 1 месяц назад
nll.py 71163f27e6 kindlingai glm-5.3-flash-gx10: vendor their stack (mode B — image+patches under our rendezvous) + 3 transcribed recipes (TP4/TP2/TP6): entrypoint exec-args + all 9 experimental overlay envs 1:1, per-TP mimics of their case block (pins 26/8/26 GiB, MML 524288/163840/524288, batched 16384/8192/16384), DFlash2 spec + AdaptiveK + tp6 padded checkpoint; build.sh bakes the 39 overlay files into the image (no per-node mounts) 1 месяц назад
prefill.py 71163f27e6 kindlingai glm-5.3-flash-gx10: vendor their stack (mode B — image+patches under our rendezvous) + 3 transcribed recipes (TP4/TP2/TP6): entrypoint exec-args + all 9 experimental overlay envs 1:1, per-TP mimics of their case block (pins 26/8/26 GiB, MML 524288/163840/524288, batched 16384/8192/16384), DFlash2 spec + AdaptiveK + tp6 padded checkpoint; build.sh bakes the 39 overlay files into the image (no per-node mounts) 1 месяц назад
prompts.py 71163f27e6 kindlingai glm-5.3-flash-gx10: vendor their stack (mode B — image+patches under our rendezvous) + 3 transcribed recipes (TP4/TP2/TP6): entrypoint exec-args + all 9 experimental overlay envs 1:1, per-TP mimics of their case block (pins 26/8/26 GiB, MML 524288/163840/524288, batched 16384/8192/16384), DFlash2 spec + AdaptiveK + tp6 padded checkpoint; build.sh bakes the 39 overlay files into the image (no per-node mounts) 1 месяц назад
rigmark-metadata.json 71163f27e6 kindlingai glm-5.3-flash-gx10: vendor their stack (mode B — image+patches under our rendezvous) + 3 transcribed recipes (TP4/TP2/TP6): entrypoint exec-args + all 9 experimental overlay envs 1:1, per-TP mimics of their case block (pins 26/8/26 GiB, MML 524288/163840/524288, batched 16384/8192/16384), DFlash2 spec + AdaptiveK + tp6 padded checkpoint; build.sh bakes the 39 overlay files into the image (no per-node mounts) 1 месяц назад
rigprose.py 71163f27e6 kindlingai glm-5.3-flash-gx10: vendor their stack (mode B — image+patches under our rendezvous) + 3 transcribed recipes (TP4/TP2/TP6): entrypoint exec-args + all 9 experimental overlay envs 1:1, per-TP mimics of their case block (pins 26/8/26 GiB, MML 524288/163840/524288, batched 16384/8192/16384), DFlash2 spec + AdaptiveK + tp6 padded checkpoint; build.sh bakes the 39 overlay files into the image (no per-node mounts) 1 месяц назад
rigsum.py 71163f27e6 kindlingai glm-5.3-flash-gx10: vendor their stack (mode B — image+patches under our rendezvous) + 3 transcribed recipes (TP4/TP2/TP6): entrypoint exec-args + all 9 experimental overlay envs 1:1, per-TP mimics of their case block (pins 26/8/26 GiB, MML 524288/163840/524288, batched 16384/8192/16384), DFlash2 spec + AdaptiveK + tp6 padded checkpoint; build.sh bakes the 39 overlay files into the image (no per-node mounts) 1 месяц назад
run.sh 71163f27e6 kindlingai glm-5.3-flash-gx10: vendor their stack (mode B — image+patches under our rendezvous) + 3 transcribed recipes (TP4/TP2/TP6): entrypoint exec-args + all 9 experimental overlay envs 1:1, per-TP mimics of their case block (pins 26/8/26 GiB, MML 524288/163840/524288, batched 16384/8192/16384), DFlash2 spec + AdaptiveK + tp6 padded checkpoint; build.sh bakes the 39 overlay files into the image (no per-node mounts) 1 месяц назад