Skip to content

Actions: aws-neuron/aws-neuron-sdk

Actions

Acknowledge New Issue

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
acknowledge-new-issue.yml
20 workflow runs
20 workflow runs

Filter by Event

⌥ + click/return to exclude

Filter by Status

⌥ + click/return to exclude

Filter by Branch

⌥ + click/return to exclude

Filter by Actor

⌥ + click/return to exclude
[torch-xla 2.9] bf16 model + fp32 intermediate arithmetic causes dtype mismatch crash after XLA_USE_BF16 migration Acknowledge New Issue #20: Issue #1318 opened by hokindeng
9s
9s
[torch-xla 2.9] torch.utils.checkpoint.checkpoint() crashes on XLA: AttributeError module 'torch' has no attribute 'xla' Acknowledge New Issue #19: Issue #1317 opened by hokindeng
8s
8s
[torch-xla 2.9] Training without torch_xla.sync() silently produces no weight updates -- silent correctness failure Acknowledge New Issue #18: Issue #1316 opened by hokindeng
7s
7s
[torch-xla 2.9] torch.save() on XLA model state_dict hangs or produces corrupt checkpoints Acknowledge New Issue #17: Issue #1315 opened by hokindeng
10s
10s
[neuronx-cc 2.24] NCC_INLA001: XLA_USE_BF16=1 + --auto-cast=none causes compiler crash on transformer backward pass (activation table limit) Acknowledge New Issue #16: Issue #1314 opened by hokindeng
7s
7s
NKI: trn1 verifier rejects shared-memory instruction in fused XLA graph spanning ao_to_mo_transform + mp2_energy Acknowledge New Issue #15: Issue #1311 opened by scttfrdmn
7s
7s
Add support for PJRT_Client_CreateUninitializedBuffer Acknowledge New Issue #14: Issue #1310 opened by steeve
9s
9s
XLA crash on stablehlo.gather on multi-core Neuron Acknowledge New Issue #13: Issue #1309 opened by elogir
8s
8s
nl.multiply and related ops on uint32 tiles lose precision above 2^24 via float32 activation path Acknowledge New Issue #12: Issue #1308 opened by scttfrdmn
11s
11s
VSCode Extension Acknowledge New Issue #11: Issue #1307 opened by eshau
9s
9s
Custom NKI kernel causes OOB scatter/gather error when compiled inside NxDI model graph (SDK 2.29) Acknowledge New Issue #10: Issue #1306 opened by jimburtoft
9s
9s
NCC_ILLC059: nkc.collectives.all_reduce (SBUF or HBM) fails to compile when tp_degree >= 2 and NEURON_LOGICAL_NC_CONFIG=2 Acknowledge New Issue #9: Issue #1305 opened by KevGomes1403
7s
7s
[neuronx-cc 2.23] NCC_ITEN404 SimplifyNeuronTensor crash with paged attention HLO patterns Acknowledge New Issue #8: Issue #1304 opened by dmvevents
9s
9s
[BUG] NCC_IXCG856 (MATCH_REPLACE8) compilation failure on --target=inf2 for Qwen2.5-14B and Qwen3 dense models — Neuron SDK 2.28.0 Acknowledge New Issue #7: Issue #1303 opened by taktakpeops
10s
10s
[libneuronxla] Add support for PJRT_Client_CreateUninitializedBuffer Acknowledge New Issue #6: Issue #1302 opened by steeve
6s
6s
nisa.dma_transpose fails during tracing on minimal legal 2D 128x128 repro on Trn2 Acknowledge New Issue #5: Issue #1301 opened by jcarrots
8s
8s
Neuron scheduler extension initializes device usage map as [true] after operator uninstall/reinstall Acknowledge New Issue #4: Issue #1300 opened by shapirov103
12s
12s
neuron_parallel_compile reports NoneType path error during artifact collection despite compiler PASS Acknowledge New Issue #3: Issue #1299 opened by samhita-alla
7s
7s
NCC_IXLV002: Core barrier name mismatch when LNC2 NKI kernel with sendrecv is called 2+ times in a traced graph Acknowledge New Issue #2: Issue #1298 opened by binaryks
6s
6s
nisa.activation with nl.copy does not add bias Acknowledge New Issue #1: Issue #1296 opened by benlimpa
8s
8s