chore: Remove unused file - #24
Merged
Merged
Conversation
fzyzcjy
added a commit
that referenced
this pull request
Mar 19, 2026
JD-ETH
pushed a commit
to JensenFire/miles
that referenced
this pull request
Apr 11, 2026
GuanxingLu
pushed a commit
to GuanxingLu/miles
that referenced
this pull request
Apr 21, 2026
XinyuJiangCMU
added a commit
to XinyuJiangCMU/miles
that referenced
this pull request
Jun 28, 2026
…ap, E19, E20)
Three more rollout-forward blockers on ROCm colocate, all env-level:
- E18 gap: AITER_CONFIG_FMOE was missing from the single-file pinning, so tuned_fmoe.csv
still hit the aiter config-merge baton deadlock. Pin it too.
- E19: DSv4 c4 indexer compressor_v2 emits a tvm c4_v2.cuh kernel whose TensorMatcher wants
a 2D kv_input but the caller passes 3D ("Tensor match failed"). SGLANG_OPT_USE_COMPRESSOR_V2=false
routes to the v1 triton compressor.
- E20: aiter paged-MQA-logits preshuffle path needs Triton>=3.5 or AITER_ENABLE_AOT_GLUON_PA_MQA_LOGITS=1
on ROCm, else it drops to a legacy page_size=1/KVBlockSize=1 path that asserts against DSv4's
page_size=256 (torch fallback asserts on seq_lens too). Enable the AOT gluon kernel.
Also rewrote the bring-up doc into one consistent ledger: corrected the stale E15/E16 "OPEN"
markers (E15 solved via fork PR radixark#24, E16 sidestepped), and added E17/E18/E19/E20.
Co-authored-by: Zhiyao Jiang <jessicajiang324@gmail.com>
XinyuJiangCMU
added a commit
to XinyuJiangCMU/miles
that referenced
this pull request
Jun 28, 2026
…r path) rollout forward (prefill+decode) now runs end-to-end on ROCm/gfx950. The indexer paged-MQA-logits goes through the torch fp8_paged_mqa_logits_torch fn instead of aiter: container triton is 3.4 (ROCm 7.0's pinned pytorch-triton-rocm), but aiter's AOT-gluon path needs triton>=3.6 (compiles to "CDNA_VERSION not in list") and the aiter legacy path asserts KVBlockSize==1 against DSv4's page. Per agent investigation, upgrading triton would drag torch up too and risk the whole image, and the v2 c4_v2 tvm kernel is a semantic ABI mismatch -- both heavier than the torch fn. Dockerfile: AITER_INDEXER=false, dropped AITER_ENABLE_AOT_GLUON_PA_MQA_LOGITS (torch fn doesn't use aiter paged-MQA). The torch fn path needs three sglang-side fixes, carried in fork PR radixark#24 (commit 734521ac): seq_lens squeeze, c4_sparse_raw_indices getattr, _c128_decode_kernel broadcast. Next blocker is actor-side E2 (Megatron DSv4 spec hard-imports CUDA-only tile_kernels -> "specialize for NoneType"). Doc updated into one consistent ledger (E10-E23 + E2). Co-authored-by: Zhiyao Jiang <jessicajiang324@gmail.com>
XinyuJiangCMU
added a commit
to XinyuJiangCMU/miles
that referenced
this pull request
Jun 28, 2026
GitHub renders bare #N in markdown as a cross-ref; replace the leftover "PR radixark#24" and "sgl#26359" with the branch name and a plain "PR 26359" so the doc never points at other repos. Co-authored-by: Zhiyao Jiang <jessicajiang324@gmail.com>
yueming-yuan
added a commit
that referenced
this pull request
Jul 26, 2026
…ge#24 Validates the TE 2.17 grouped-linear contract fix end to end. The `not enough values to unpack (expected 21, got 16)` failures in the test_glm5_*_lora_ci tests come from megatron-bridge's `_forward_te_grouped_linear` calling TE's private `_GroupedLinear` with TE <=2.14's positional layout; miles picked that up when it bumped TransformerEngine 2.12 -> 2.17 in #1781. Revert to `@bridge` -- ideally pinned to the merged commit rather than a branch, as line 82 does for mbridge -- once #24 lands. Note this does not address `window_size_left`, which is a separate TE 2.17 vs flash-attn skew inside TE's own context-parallel path.
yueming-yuan
added a commit
that referenced
this pull request
Jul 27, 2026
radixark/Megatron-Bridge#24 landed on @bridge as 7f0fb345 (squash). Its src/megatron/bridge/peft/utils.py is byte-identical to 36d20288, the revision this PR already exercised, so the pin moves off the PR branch without changing what runs. Kept as a SHA rather than @bridge. Line 82 pins mbridge the same way, and the alternative is the failure this PR already hit once: buildkit keys the layer on the instruction text, so a moving branch silently reuses the cached layer and ships the previous revision.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.