fix test_setitem segment fault#53

Closed

lynex wants to merge 2026 commits intomicrosoft:mainfrom

msrasys:yomia/fix_test_setitem

This pull request is big! We're only showing the most recent 250 commits

Commits on May 29, 2024

Merged PR 2150: fix fullslice
Yilei Yang
committed

Commits on Jun 3, 2024

Merged PR 2153: Add dedup_attrs to reduce model size in checkpointing
Yi Zhu
committed

Commits on Jun 4, 2024

Commits on Jun 5, 2024

Merged PR 2166: rename dynamic_shape to constant_folding
0xWJ
committed

Commits on Jun 13, 2024

Merged PR 2173: Refine infer batch dim in autodist
Yi Zhu
committed
Merged PR 2175: change use_reentrant to False
Ning Shang
committed

Commits on Jun 14, 2024

Merged PR 2172: refine IRObject handling
0xWJ
committed

Commits on Jun 15, 2024

Commits on Jun 17, 2024

Merged PR 2176: add a mini-trainer
0xWJ
committed

Commits on Jun 18, 2024

Merged PR 2040: add interface for cube integration test and script for collect huggingface mo...
Yilei Yang
committed
Merged PR 2181: quick fix compile huggingface
Yilei Yang
committed

Commits on Jun 19, 2024

Merged PR 2111: refine optimizer state dict merge
0xWJ
committed

Commits on Jun 21, 2024

Merged PR 2180: lightning: fix gradient sync and gradient averaging
0xWJ
committed

Commits on Jun 24, 2024

Merged PR 2183: fix cache dir
Yilei Yang
committed
Merged PR 2179: Refine follow logic in autodist
Yi Zhu
committed

Commits on Jun 25, 2024

Commits on Jun 26, 2024

Merged PR 2184: parser: never fold getattr node 'self.training'
0xWJ
committed

Commits on Jun 28, 2024

Merged PR 2188: self.training in submodules: hotfix for nightly test
0xWJ
committed

Commits on Jul 2, 2024

Commits on Jul 3, 2024

Merged PR 2196: never fold nnscaler runtime functions
0xWJ
committed

Commits on Jul 4, 2024

Merged PR 2171: Refine dp solver in autodist
Yi Zhu
committed

Commits on Jul 8, 2024

Merged PR 2198: Fix bugs in autodist
Yi Zhu
committed

Commits on Jul 11, 2024

Merged PR 2169: support conv1d-2d
Xin Ji (CSI Interfusion Co Ltd)
committed

Commits on Jul 12, 2024

Merged PR 2200: TensorBase adaption in torch>=2.3
Ning Shang
committed
Merged PR 2197: update doc for v0.1
lynex
committed

Commits on Jul 15, 2024

Merged PR 2191: nanoGPT example
Zhe Liu
committed
Merged PR 2199: fix ifexpr warning
0xWJ
committed

Commits on Jul 16, 2024

Commits on Jul 17, 2024

Merged PR 2204: fix embedding padding index
Ning Shang
committed

Commits on Jul 22, 2024

Merged PR 2192: minitrainer: refine config
0xWJ
committed

Commits on Jul 24, 2024

Merged PR 2209: Nanogpt with mini-trainer
0xWJ
committed

Commits on Jul 25, 2024

Merged PR 2210: Minitrainer: refine names / precision support
0xWJ
committed

Commits on Jul 29, 2024

Merged PR 2212: add mixed precision f16 optimizer
0xWJ
committed

Commits on Jul 30, 2024

Commits on Jul 31, 2024

Merged PR 2217: scalar tensor: use right shape in generated module init code
0xWJ
committed

Commits on Aug 2, 2024

Commits on Aug 6, 2024

Merged PR 2203: Refine pipeline implementations
Yi Zhu
committed

Commits on Aug 7, 2024

Merged PR 2221: minitrainer: add document
0xWJ
committed

Commits on Aug 8, 2024

Merged PR 2222: parallel module: remove pipeline related config from global compute config
0xWJ
committed

Commits on Aug 14, 2024

Merged PR 2225: Add missing annotation for cross_entropy
Zhe Liu
committed

Commits on Aug 15, 2024

Merged PR 2230: minitrainer: Add model/optimizer/lr_scheduler hook support.
0xWJ
committed
Merged PR 2231: Align grad computation with fairseq
Yi Zhu
committed
Merged PR 2223: refine pytree
Ning Shang
committed
Merged PR 2233: Minitrainer logging: log tag support
0xWJ
committed
Merged PR 2211: support batchnorm2d
Xin Ji (CSI Interfusion Co Ltd)
committed

Commits on Aug 16, 2024

Commits on Aug 21, 2024

Commits on Aug 22, 2024

Merged PR 2243: Fix parallel module when loading from a merged checkpoint
Yi Zhu
committed
Merged PR 2240: replace _orig_xxx with orig_func
Ning Shang
committed

Commits on Aug 23, 2024

Merged PR 2242: Refine requires_grad of input tensors handling
0xWJ
committed

Commits on Aug 26, 2024

Merged PR 2245: Hotfix for reducer generation.
0xWJ
committed

Commits on Aug 28, 2024

Commits on Aug 29, 2024

Merged PR 2248: Minitrainer: refine progress bar and load_type
0xWJ
committed
Merged PR 2208: auto op partition testing
Xin Ji (CSI Interfusion Co Ltd)
committed

Commits on Sep 2, 2024

Merged PR 2249: Llama3 128K finetuning for v0.3 release
Yi Zhu
committed

Commits on Sep 3, 2024

Commits on Sep 6, 2024

Merged PR 2258: remove unpolished examples
0xWJ
committed
Merged PR 2253: refine tracer utils
Ning Shang
committed

Commits on Sep 10, 2024

Merged PR 2263: minitrainer: empty dataset train_args support
0xWJ
committed

Commits on Sep 18, 2024

Merged PR 2265: Add help message for cpp module when install from source
0xWJ
committed

Commits on Sep 22, 2024

Merged PR 2266: Fix np repr break change from latest np
0xWJ
committed

Commits on Sep 23, 2024

Merged PR 2256: update readme
lynex
committed
Merged PR 2260: modularize run target logic to strategy
Ning Shang
committed
Merged PR 2262: fix module creation during trace
Ning Shang
committed
Merged PR 2264: fix proxy in tensor metadata
Ning Shang
committed
Merged PR 2244: Llama 3 8B-8K example (recreated PR)
Zhe Liu
committed

Commits on Sep 24, 2024

Merged PR 2269: UT refine: add model to cuda
Ning Shang
committed

Commits on Sep 27, 2024

Commits on Oct 8, 2024

Merged PR 2267: Refine Async Support
0xWJ
committed

Commits on Oct 9, 2024

Merged PR 2277: Update pyproject meta fields

Zhe Liu
authored and
0xWJ
committed

Commits on Oct 11, 2024

Merged PR 2280: add options to use reduce scatter when zero is on
0xWJ
committed

Commits on Oct 14, 2024

Merged PR 2283: Support constant pad
Yi Zhu
committed

Commits on Oct 16, 2024

Merged PR 2281: Refine code for MoE
Yi Zhu
committed

Commits on Oct 17, 2024

Commits on Oct 22, 2024

Merged PR 2282: add option to use reduce-scatter adapter
0xWJ
committed

Commits on Oct 23, 2024

Merged PR 2292: rvd: retry without reduce-scatter adapter when any adapters are invalid
0xWJ
committed

Commits on Oct 24, 2024

Merged PR 2293: add dis test
0xWJ
committed
Merged PR 2291: update llama 128k example args
Ning Shang
committed

Commits on Oct 25, 2024

Commits on Oct 30, 2024

Merged PR 2297: fix unpartitionable identifiers in annotation

Yilei Yang
authored and
0xWJ
committed

Commits on Oct 31, 2024

Merged PR 2295: add grad check in trainer
Ning Shang
committed

Commits on Nov 1, 2024

Merged PR 2302: [Bugfix] grad check config
Ning Shang
committed
Merged PR 2299: refine tracer hierarchy
Ning Shang
committed

Commits on Nov 5, 2024

Merged PR 2298: [BugFix]: infer grad correctly
Yi Zhu
committed

Commits on Nov 11, 2024

Merged PR 2304: Transformer (vit) example
0xWJ
committed

Commits on Nov 12, 2024

Merged PR 2305: add unit test for depthwise conv2d
Yilei Yang
committed

Commits on Nov 13, 2024

Commits on Nov 14, 2024

Merged PR 2290: Resume RNG state and add test case
Zhe Liu
committed
Merged PR 2288: make all parameters in reducer buffer to be aligned to 16 bytes
Yilei Yang
committed

Commits on Nov 15, 2024

Commits on Nov 19, 2024

Commits on Nov 21, 2024

Commits on Nov 22, 2024

Merged PR 2285: Add max train steps arg to llama 128k example

Zhe Liu
authored and
Zhe Liu
committed
Merged PR 2322: trust remote code for load dataset
Yilei Yang
committed
Merged PR 2315: [Model Example] Support Llama3 70B 8k by pipeline parallelism on mi300
Yi Zhu
committed
Merged PR 2287: add dagan example to nnscaler examples
Yilei Yang
committed

Commits on Nov 25, 2024

Merged PR 2320: [Tracer] add new jump op name

Ning Shang
authored and
Ning Shang
committed
Merged PR 2313: [Doc] doc update
lynex
committed

Commits on Nov 26, 2024

Merged PR 2325: [BugFix] detach correctly when multiple outputs
Yi Zhu
committed

Commits on Nov 28, 2024

Merged PR 2328: Merge changes from github release

Zhe Liu
authored and
Zhe Liu
committed
Merged PR 2331: Minor fixes for llama3 demo and misc changes

Zhe Liu
authored and
Zhe Liu
committed
Merged PR 2329: [CI/Build] attn_implementation as an option in example/llama for V100 not support flash_attention_2
Yilei Yang
committed

Commits on Nov 29, 2024

Commits on Dec 2, 2024

Merged PR 2337: [BUG] Fix split function
Yilei Yang
committed

Commits on Dec 3, 2024

Merged PR 2326: [BugFix] reimplement reshape function
0xWJ
committed
Merged PR 2338: Drop python 3.8 support

Zhe Liu
authored and
Zhe Liu
committed

Commits on Dec 6, 2024

Merged PR 2340: [Doc] add test and doc for local_consumer_multiref
Yi Zhu
committed

Commits on Dec 10, 2024

Merged PR 2341: [Reorg + Parser] Polish parser logic
0xWJ
committed
Merged PR 2339: [AutoDist] Multi-nodes communication profiling

Zhe Liu
authored and
Zhe Liu
committed
Merged PR 2343: Update README.md (merge github)

Zhe Liu
authored and
Zhe Liu
committed

Commits on Dec 12, 2024

Merged PR 2344: [Parser] never do dce on leaf functions
0xWJ
committed

Commits on Dec 13, 2024

Merged PR 2342: [Reorg + Parser] refine infer_shape
0xWJ
committed

Commits on Dec 19, 2024

Merged PR 2346: [Profiler + Test] add tests for profiler
0xWJ
committed

Commits on Dec 20, 2024

Merged PR 2348: [Model Example] Support diff-attention
Yilei Yang
committed

Commits on Dec 26, 2024

Merged PR 2347: [Parser] Clarify only dict/tuple/list/slice are supported collection types
0xWJ
committed

Commits on Dec 29, 2024

Merged PR 2350: [BugFix] import flash_attn only when available
Yilei Yang
committed

Commits on Jan 3, 2025

Merged PR 2349: [BugFix] Fix profiler's test
Yi Zhu
committed

Commits on Jan 6, 2025

Merged PR 2352: [UT] fix dis test

Ning Shang
authored and
Ning Shang
committed

Commits on Jan 9, 2025

Merged PR 2353: Update version numbers and add missing headers

Zhe Liu
authored and
Zhe Liu
committed

Commits on Jan 16, 2025

Merged PR 2355: [Tracer] support subscriptable type hint

Ning Shang
authored and
Ning Shang
committed
Merged PR 2345: [Tracer + Codegen] support grad mode & autocast

Ning Shang
authored and
Ning Shang
committed

Commits on Jan 20, 2025

Commits on Jan 22, 2025

Merged PR 2351: [Codegen] add async support for pipeline parallelism
0xWJ
committed

Commits on Jan 23, 2025

Merged PR 2357: [Trainer] Add partial parallelized module support.
0xWJ
committed

Commits on Feb 12, 2025

Merged PR 2363: Merge github PR and update version

Zhe Liu
authored and
Zhe Liu
committed
Merged PR 2358: [AutoDist] Support pipeline_nstages option

Zhe Liu
authored and
Zhe Liu
committed
Merged PR 2361: [Example] improve diff attention
Yilei Yang
committed

Commits on Feb 13, 2025

Merged PR 2366: [BugFix] Fix autodist's data parallel test
Yi Zhu
committed
Merged PR 2364: [UT] Fix pipeline_nstages UT

Zhe Liu
authored and
Zhe Liu
committed

Commits on Feb 20, 2025

Merged PR 2368: [AutoDist] Fix when model too small autodist does not use all GPUs

Zhe Liu
authored and
Zhe Liu
committed

Commits on Mar 21, 2025

Merged PR 2362: [Example] longrope

Ning Shang
authored and
Ning Shang
committed

Commits on Mar 24, 2025

Merged PR 2370: [BugFix] Fix autocast in AutoDist
Yi Zhu
committed

Commits on Mar 26, 2025

Commits on Jun 9, 2025

Merged PR 2378: add iter dataset support and stateful dataloader
0xWJ
committed

Commits on Jun 10, 2025

Merged PR 2379: [Trainer]Load auto-merged checkpoint when world size changes
0xWJ
committed

Commits on Jun 18, 2025

Merged PR 2377: [AutoDist] fix bug: calculate split info correctly
Yi Zhu
committed

Commits on Jul 21, 2025

Merged PR 2380: add Fairseq example
0xWJ
committed

Commits on Jul 22, 2025

Merged PR 2381: [AutoDist] Remove cppimport
Yi Zhu
committed

Commits on Jul 23, 2025

Merged PR 2382: [BugFix] TrainerArgs Resolution for command line args
0xWJ
committed

Commits on Jul 24, 2025

Commits on Jul 25, 2025

Commits on Aug 6, 2025

Merged PR 2386: [BugFix] torch.compile support
0xWJ
committed

Commits on Aug 7, 2025

Commits on Aug 12, 2025

Merged PR 2387: Add Copyright and update version of package datasets
Yilei Yang
committed

Commits on Aug 13, 2025

Merged PR 2390: [BugFix] Add functools.cache support
0xWJ
committed

Commits on Aug 14, 2025

Commits on Aug 20, 2025

Merged PR 2394: [BugFix] fix bug in dedup load: handle persistent buffer correctly & use broadcast instead of broadcast_object_list to save time
Yi Zhu
committed

Commits on Aug 21, 2025

Merged PR 2396: [AutoDist] Refine error handling and logging
Yi Zhu
committed

Commits on Aug 27, 2025

Commits on Sep 1, 2025

Merged PR 2399: [Runtime] ParallelModule: move more instance member variable to class variable.
0xWJ
committed

Commits on Sep 8, 2025

Commits on Sep 9, 2025

Merged PR 2401: [Runtime] Add option to load merged state dict with less memory usage.
0xWJ
committed

Commits on Sep 12, 2025

Merged PR 2406: [Runtime] Allow custom parameter bucketing
0xWJ
committed

Commits on Sep 15, 2025

Merged PR 2407: [BwCompat]: add backward compatibiity for cli checkpoint config
0xWJ
committed

Commits on Sep 17, 2025

Merged PR 2408: [Runtime] Bugfix: calculate gnorm correctly when existing different precision params
Yi Zhu
committed

Commits on Sep 19, 2025

Merged PR 2409: [Refine] Parallelize: skip graph.forward call to simplify code
0xWJ
committed

Commits on Oct 13, 2025

Commits on Oct 17, 2025

Merged PR 2414: [Refine] Make parser a real class
0xWJ
committed

Commits on Oct 27, 2025

Merged PR 2417: [Runtime] Refine ring attention related implementation
Yi Zhu
committed

Commits on Oct 29, 2025

Commits on Nov 3, 2025

Merged PR 2420: [Runtime] Add runtime pre-hook and post-hook for node
0xWJ
committed

Commits on Nov 20, 2025

Commits on Dec 9, 2025

Merged PR 2425: [Feat] cli: Add safetensors support
0xWJ
committed

Commits on Dec 15, 2025

Commits on Dec 16, 2025

Merged PR 2431: [BugFix] Handle graph output correctly in dynamic programming solver
Yi Zhu
committed
Merged PR 2432: [Bugfix] return cloned tensor when a parameter is multiref'ed in zero3
0xWJ
committed
Merged PR 2427: [BugFix] auto pack kwargs in dummy input

Ning Shang
authored and
Ning Shang
committed

Commits on Dec 18, 2025

Merged PR 2435: [Bugfix] Fix the generated code when **kwargs are used in module.forward
0xWJ
committed

Commits on Dec 23, 2025

Merged PR 2438: [BwCompat] Fix backward compatiblity when metadata lacks zero version
0xWJ
committed

Commits on Dec 25, 2025

Merged PR 2442: [BugFix] Rebuild param groups after resuming in HybridOptimizer
Yi Zhu
committed

Commits on Dec 29, 2025

Merged PR 2443: [Feat] Add option disable_shared_param_constraint and recompute_ratio to autodist
Yi Zhu
committed

Commits on Dec 30, 2025

Merged PR 2437: [Feat] cli: add customized serialization runner
0xWJ
committed

Commits on Dec 31, 2025

Fix & Update
nnScaler
committed
Bump version to 0.8
lynex
authored

Commits on Jan 4, 2026

Merged PR 2450: [AutoDist] Add dynamic constraint
Yi Zhu
committed
Merged PR 2451: [Parser] Support `torch.eye` and `.T`
Yi Zhu
committed
Merged PR 2441: [Feat] Add option: `reducer_pre_divisor`
Yi Zhu
committed

Commits on Jan 5, 2026

Commits on Jan 8, 2026

Merge pull request #1 from msrasys/devops
lynex
authored

Commits on Jan 14, 2026

save work
0xWJ
committed

Commits on Jan 15, 2026

Commits on Jan 19, 2026

Commits on Jan 20, 2026

Commits on Jan 21, 2026

Commits on Jan 23, 2026

refine code
0xWJ
committed

Commits on Jan 24, 2026

add barrier
0xWJ
committed

Commits on Jan 26, 2026

Merge pull request #6 from msrasys/weijiangxu/mem-fragment-in-resume
lynex
authored

Commits on Jan 28, 2026

[Tracer] Provide better einops tracing by skipping tracing some internal einops functions.
0xWJ
authored

Commits on Jan 30, 2026

[Refine] Normalize device handling in state dicts and more (#9)
0xWJ
authored
Add Doc Autodist Constraints Guide (#5)

yyl9510
and
yileiyang
authored
CI/CD (#11)

yyl9510
and
yileiyang
authored

Commits on Feb 3, 2026

Add nightly test to the repo (#12)

yyl9510
and
yileiyang
authored

Commits on Feb 4, 2026

Commits on Feb 6, 2026

Warmup triggers 150K times, put too much pressure on memory and lead to segmentation fault
lynex
committed