All issues

Search results

[Vision] Toward ModelExpress 1.0

Status: Open.

#594 In ai-dynamo/modelexpress;

· athreesh opened on Aug 6, 2026

feat(huggingface): support pinned revisions for ModelExpress downloads

Status: Open.

#586 In ai-dynamo/modelexpress;

· tmonty12 opened on Aug 3, 2026

feat: integrate server-side model-cache fallback with P2P loaders

enhancementNew feature or requestNew feature or request

Status: Open.

#569 In ai-dynamo/modelexpress;

· zhengluo-nv opened on Jul 28, 2026

Torch compile cache artifacts transfer between workers with incompatible compile configurations

Status: Open.

#560 In ai-dynamo/modelexpress;

· kangclzjc opened on Jul 28, 2026

Pulling both model weights and MTP from s3:// leads to a crash

Status: Open.

#537 In ai-dynamo/modelexpress;

· gshtrasberg opened on Jul 23, 2026

auth: bound TokenReview traffic from unauthenticated callers

Status: Open.

#530 In ai-dynamo/modelexpress;

· wseaton opened on Jul 22, 2026

[RFC] Delta weight sync requirements

Status: Open.

#524 In ai-dynamo/modelexpress;

· athreesh opened on Jul 21, 2026

Clarify usage of model-streamer with s3 compatible object-storage

Status: Open.

#500 In ai-dynamo/modelexpress;

· strigazi opened on Jul 9, 2026

RDMA transfer to a stale donor aborts the process at libfabric with no fallback

Status: Open.

#499 In ai-dynamo/modelexpress;

· emricksini-h opened on Jul 9, 2026

Validate P2P transfers across different GPU SKUs

enhancementNew feature or requestNew feature or request

Status: Open.

#476 In ai-dynamo/modelexpress;

· zhengluo-nv opened on Jul 1, 2026

Topology-aware source metadata for MoE and P/D scaleout

Status: Open.

#471 In ai-dynamo/modelexpress;

· athreesh opened on Jul 1, 2026

perf: multi-NIC parallel RDMA pull — MX is single-NIC-pinned by default, leaves 75-87% of allocated bandwidth idle

Status: Open.

#449 In ai-dynamo/modelexpress;

· KavinKrishnan opened on Jun 23, 2026