| .. |
|
_fa_common.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)
|
пре 3 недеља |
|
attention.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)
|
пре 3 недеља |
|
causal_conv1d.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)
|
пре 3 недеља |
|
flashinfer_mla_sparse_sm90.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)
|
пре 3 недеља |
|
gb10_sparse_mla.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)
|
пре 3 недеља |
|
gdn_attn.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)
|
пре 3 недеља |
|
kda.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)
|
пре 3 недеља |
|
kv_cache_coordinator.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)
|
пре 3 недеља |
|
kv_cache_interface.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)
|
пре 3 недеља |
|
kv_cache_utils.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)
|
пре 3 недеља |
|
mhc_tilelang.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)
|
пре 3 недеља |
|
model.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)
|
пре 3 недеља |
|
recoverssm.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)
|
пре 3 недеља |
|
tilelang_kernels.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)
|
пре 3 недеља |
|
warmup.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)
|
пре 3 недеља |
|
worker_utils.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)
|
пре 3 недеља |
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)