Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2026 commits
Select commit Hold shift + click to select a range
49433e5
Merged PR 2150: fix fullslice
May 29, 2024
8036b82
Merged PR 2153: Add dedup_attrs to reduce model size in checkpointing
Jun 3, 2024
79070fb
Merged PR 2158: parallel module: rename cube related names
0xWJ Jun 4, 2024
7371021
Merged PR 2159: Integrate zigzag attention in nnscaler
Jun 4, 2024
30ec31d
Merged PR 2156: initialize lightning support
0xWJ Jun 4, 2024
16e99c3
Merged PR 2166: rename dynamic_shape to constant_folding
0xWJ Jun 5, 2024
906a574
Merged PR 2173: Refine infer batch dim in autodist
Jun 13, 2024
ebfe230
Merged PR 2175: change use_reentrant to False
Jun 13, 2024
8080e2a
Merged PR 2172: refine IRObject handling
0xWJ Jun 14, 2024
9c57adb
Merged PR 2178: parallelize: rename dummy_input to dummy_forward_args
0xWJ Jun 15, 2024
17877a9
Merged PR 2177: Fix policy for autodist bug
Jun 15, 2024
046e020
Merged PR 2176: add a mini-trainer
0xWJ Jun 17, 2024
8dfec28
Merged PR 2040: add interface for cube integration test and script fo…
Jun 18, 2024
77cabf5
Merged PR 2181: quick fix compile huggingface
Jun 18, 2024
d0b7e5e
Merged PR 2111: refine optimizer state dict merge
0xWJ Jun 19, 2024
4dc1166
Merged PR 2180: lightning: fix gradient sync and gradient averaging
0xWJ Jun 21, 2024
98d57c8
Merged PR 2183: fix cache dir
Jun 24, 2024
312539e
Merged PR 2179: Refine follow logic in autodist
Jun 24, 2024
eeef286
Merged PR 2186: hotfix: non-tensor support for consistence check in …
0xWJ Jun 25, 2024
deb1d84
Merged PR 2187: Fix parity alert: forbidden to follow operators that …
Jun 25, 2024
6bb80de
Merged PR 2185: bugfix to train_mem2in_idx
Jun 25, 2024
b943f8e
Merged PR 2184: parser: never fold getattr node 'self.training'
0xWJ Jun 26, 2024
04c608a
Merged PR 2188: self.training in submodules: hotfix for nightly test
0xWJ Jun 28, 2024
ef2586e
Merged PR 2189: Lightning: refine code/add more tests
0xWJ Jul 2, 2024
a182bcc
Merged PR 2144: Nightly build scripts
Jul 2, 2024
9eebf40
Merged PR 2194: Reset version to v0.1 and update email
Jul 2, 2024
7a2485d
Merged PR 2193: lightning: refine docs about checkpoint
0xWJ Jul 2, 2024
d6f6c09
Merged PR 2196: never fold nnscaler runtime functions
0xWJ Jul 3, 2024
d6c25a3
Merged PR 2171: Refine dp solver in autodist
Jul 4, 2024
d1d123c
Merged PR 2198: Fix bugs in autodist
Jul 8, 2024
42f64b1
Merged PR 2169: support conv1d-2d
Jul 11, 2024
d82882e
Merged PR 2200: TensorBase adaption in torch>=2.3
Jul 12, 2024
8ee468b
Merged PR 2197: update doc for v0.1
lynex Jul 12, 2024
0f9d8da
Merged PR 2191: nanoGPT example
Jul 15, 2024
13ad8ff
Merged PR 2199: fix ifexpr warning
0xWJ Jul 15, 2024
d185206
Merged PR 2205: Fix sum anno bug
Jul 16, 2024
5368dad
Merged PR 2201: Add release pipeline and update dev version to v0.2
Jul 16, 2024
c59f7b1
Merged PR 2202: add scalar tensor support
0xWJ Jul 16, 2024
106bbf2
Merged PR 2204: fix embedding padding index
Jul 17, 2024
992b945
Merged PR 2192: minitrainer: refine config
0xWJ Jul 22, 2024
027fd64
Merged PR 2209: Nanogpt with mini-trainer
0xWJ Jul 24, 2024
167704a
Merged PR 2210: Minitrainer: refine names / precision support
0xWJ Jul 25, 2024
c5b6dfb
Merged PR 2212: add mixed precision f16 optimizer
0xWJ Jul 29, 2024
67f0e81
Merged PR 2214: lightning: add merged checkpoint support
0xWJ Jul 30, 2024
6a069fa
Merged PR 2213: Refine ring flash attn: add llama 3.1's implementation
Jul 30, 2024
965907c
Merged PR 2207: Fix loss related gencode
Jul 30, 2024
65d84d6
Merged PR 2216: fix out of disk error when running pipeline.
0xWJ Jul 30, 2024
9b8bdb7
Merged PR 2217: scalar tensor: use right shape in generated module in…
0xWJ Jul 31, 2024
f6448a5
Merged PR 2218: minitrainer: fix checkpoint bug (last checkpoint may …
0xWJ Aug 2, 2024
7c3a85f
Merged PR 2219: Add example chunk_linear_cross_entropy and refine aut…
Aug 2, 2024
a9942a7
Merged PR 2203: Refine pipeline implementations
Aug 6, 2024
b34d283
Merged PR 2221: minitrainer: add document
0xWJ Aug 7, 2024
cb40f76
Merged PR 2222: parallel module: remove pipeline related config from …
0xWJ Aug 8, 2024
06b94fc
Merged PR 2225: Add missing annotation for cross_entropy
Aug 14, 2024
a999b25
Merged PR 2230: minitrainer: Add model/optimizer/lr_scheduler hook su…
0xWJ Aug 15, 2024
c4e574a
Merged PR 2231: Align grad computation with fairseq
Aug 15, 2024
c07cd2d
Merged PR 2223: refine pytree
Aug 15, 2024
6bb0a9e
Merged PR 2233: Minitrainer logging: log tag support
0xWJ Aug 15, 2024
0c394ff
Merged PR 2211: support batchnorm2d
Aug 15, 2024
fefb87a
Merged PR 2234: minitrainer: remove torchrun requirements for compile
0xWJ Aug 16, 2024
8fec4b8
Merged PR 2227: Log importance ratio for operators to make it easy to…
Aug 16, 2024
e6d19e9
Merged PR 2226: add nested output support
0xWJ Aug 16, 2024
c4335c3
Merged PR 2238: minitrainer: fix bug when running compile with multip…
0xWJ Aug 16, 2024
5d88266
Merged PR 2235: bugfix: grad track in trace
Aug 21, 2024
ec6137a
Merged PR 2239: parallel module: decouple from Program()
0xWJ Aug 21, 2024
07637d8
Merged PR 2243: Fix parallel module when loading from a merged checkp…
Aug 22, 2024
969f3f7
Merged PR 2240: replace _orig_xxx with orig_func
Aug 22, 2024
4b05bbc
Merged PR 2242: Refine requires_grad of input tensors handling
0xWJ Aug 23, 2024
2ac6499
Merged PR 2245: Hotfix for reducer generation.
0xWJ Aug 26, 2024
ce8ff4b
Merged PR 2246: refine tracer wrap logic
Aug 28, 2024
11aa2c7
Merged PR 2250: bugfix: submodule buffer persistent
0xWJ Aug 28, 2024
ae9d0f9
Merged PR 2248: Minitrainer: refine progress bar and load_type
0xWJ Aug 29, 2024
1472c0a
Merged PR 2208: auto op partition testing
Aug 29, 2024
93e35da
Merged PR 2249: Llama3 128K finetuning for v0.3 release
Sep 2, 2024
a0b5775
Merged PR 2254: Code Refine: Input to IRObject conversion
0xWJ Sep 3, 2024
adc72f6
Merged PR 2257: minitrainer: add run method for both train/compile
0xWJ Sep 3, 2024
a375445
Merged PR 2258: remove unpolished examples
0xWJ Sep 6, 2024
d1e1c24
Merged PR 2253: refine tracer utils
Sep 6, 2024
a25f753
Merged PR 2263: minitrainer: empty dataset train_args support
0xWJ Sep 10, 2024
e0143fb
Merged PR 2265: Add help message for cpp module when install from source
0xWJ Sep 18, 2024
3577eb8
Merged PR 2266: Fix np repr break change from latest np
0xWJ Sep 22, 2024
5006d53
Merged PR 2256: update readme
lynex Sep 23, 2024
d6ec7d8
Merged PR 2260: modularize run target logic to strategy
Sep 23, 2024
c3ee6cf
Merged PR 2262: fix module creation during trace
Sep 23, 2024
095645f
Merged PR 2264: fix proxy in tensor metadata
Sep 23, 2024
1527c7e
Merged PR 2244: Llama 3 8B-8K example (recreated PR)
Sep 23, 2024
1dc54fc
Merged PR 2269: UT refine: add model to cuda
Sep 24, 2024
716b9dc
Merged PR 2272: Pin llama 3 demo example's dependency versions
Sep 27, 2024
2c05d39
Merged PR 2271: add copyright
lynex Sep 27, 2024
44725cf
Merged PR 2273: bump version to v0.3
0xWJ Sep 27, 2024
498491d
Merged PR 2275: Fix packaging bug
Sep 27, 2024
2866af1
Merged PR 2267: Refine Async Support
0xWJ Oct 8, 2024
58008fa
Merged PR 2277: Update pyproject meta fields
Oct 9, 2024
4f02e2c
Merged PR 2280: add options to use reduce scatter when zero is on
0xWJ Oct 11, 2024
94e76e7
Merged PR 2283: Support constant pad
Oct 14, 2024
cdf7944
Merged PR 2281: Refine code for MoE
Oct 16, 2024
c314670
Merged PR 2284: Add modeling code and doc for deepseek coder v2 lite
Oct 17, 2024
85560cb
Merged PR 2286: Fix bugs for gnorm computation
Oct 17, 2024
5982766
Merged PR 2282: add option to use reduce-scatter adapter
0xWJ Oct 22, 2024
ce1e31a
Merged PR 2292: rvd: retry without reduce-scatter adapter when any ad…
0xWJ Oct 23, 2024
2e7cbd8
Merged PR 2293: add dis test
0xWJ Oct 24, 2024
16a1635
Merged PR 2291: update llama 128k example args
Oct 24, 2024
f5688b2
Merged PR 2294: wrapnn: integrate to parallelize
0xWJ Oct 25, 2024
4e305a1
Merged PR 2296: [skip ci] bump version to 0.4
0xWJ Oct 25, 2024
4e2c30e
Merged PR 2297: fix unpartitionable identifiers in annotation
Oct 30, 2024
fa98abb
Merged PR 2295: add grad check in trainer
Oct 31, 2024
442e012
Merged PR 2302: [Bugfix] grad check config
Nov 1, 2024
b59b82f
Merged PR 2299: refine tracer hierarchy
Nov 1, 2024
aba5aca
Merged PR 2298: [BugFix]: infer grad correctly
Nov 5, 2024
51dd454
Merged PR 2304: Transformer (vit) example
0xWJ Nov 11, 2024
206e541
Merged PR 2305: add unit test for depthwise conv2d
Nov 12, 2024
edb4510
Merged PR 2307: add async support for non-end2end modules with sync_g…
0xWJ Nov 13, 2024
8d44619
Merged PR 2306: Refine function.To implementation
0xWJ Nov 13, 2024
4126a14
Merged PR 2290: Resume RNG state and add test case
Nov 14, 2024
91fd30a
Merged PR 2288: make all parameters in reducer buffer to be aligned t…
Nov 14, 2024
9fb800d
Merged PR 2308: [BugFix] add detach loss in codegen to deallocate ten…
Nov 15, 2024
cfc5340
Merged PR 2309: [Bugfix] tracer: handle importlib carefully
0xWJ Nov 15, 2024
592ff83
Merged PR 2316: [BugFix] fix progress bar and file expire in trainer
0xWJ Nov 19, 2024
5fc3326
Merged PR 2311: [BugFix] Estimate memory cost correctly in Pipeline S…
Nov 19, 2024
d04f20a
Merged PR 2319: [Tracer] add scoped constant folding option
0xWJ Nov 21, 2024
fb86ffc
Merged PR 2321: [BugFix] Fix cli random seed unit test
0xWJ Nov 21, 2024
1bdedcc
Merged PR 2285: Add max train steps arg to llama 128k example
Nov 22, 2024
1be48b2
Merged PR 2322: trust remote code for load dataset
Nov 22, 2024
7abfca3
Merged PR 2315: [Model Example] Support Llama3 70B 8k by pipeline par…
Nov 22, 2024
926ee1c
Merged PR 2287: add dagan example to nnscaler examples
Nov 22, 2024
a6f82b7
Merged PR 2320: [Tracer] add new jump op name
Nov 25, 2024
c0338f2
Merged PR 2313: [Doc] doc update
lynex Nov 25, 2024
1861c7c
Merged PR 2325: [BugFix] detach correctly when multiple outputs
Nov 26, 2024
0975639
Merged PR 2328: Merge changes from github release
Nov 28, 2024
76acedb
Merged PR 2331: Minor fixes for llama3 demo and misc changes
Nov 28, 2024
b1eb666
Merged PR 2329: [CI/Build] attn_implementation as an option in exampl…
Nov 28, 2024
7800927
Merged PR 2327: [Reorg] Refine DistAlgo logic
0xWJ Nov 29, 2024
ba1ec5b
Merged PR 2317: [BugFix] Fix wrong loss name in generated code
Nov 29, 2024
df23b26
Merged PR 2337: [BUG] Fix split function
Dec 2, 2024
9b0f871
Merged PR 2326: [BugFix] reimplement reshape function
0xWJ Dec 3, 2024
0a80aff
Merged PR 2338: Drop python 3.8 support
Dec 3, 2024
f02bdcc
Merged PR 2340: [Doc] add test and doc for local_consumer_multiref
Dec 6, 2024
1ac5b17
Merged PR 2341: [Reorg + Parser] Polish parser logic
0xWJ Dec 10, 2024
93268ef
Merged PR 2339: [AutoDist] Multi-nodes communication profiling
Dec 10, 2024
7433277
Merged PR 2343: Update README.md (merge github)
Dec 10, 2024
4d6cd75
Merged PR 2344: [Parser] never do dce on leaf functions
0xWJ Dec 12, 2024
a7a5089
Merged PR 2342: [Reorg + Parser] refine infer_shape
0xWJ Dec 13, 2024
a9bdf58
Merged PR 2346: [Profiler + Test] add tests for profiler
0xWJ Dec 19, 2024
e8d7632
Merged PR 2348: [Model Example] Support diff-attention
Dec 20, 2024
0c23d53
Merged PR 2347: [Parser] Clarify only dict/tuple/list/slice are suppo…
0xWJ Dec 26, 2024
7899103
Merged PR 2350: [BugFix] import flash_attn only when available
Dec 29, 2024
4c75567
Merged PR 2349: [BugFix] Fix profiler's test
Jan 3, 2025
5795f71
Merged PR 2352: [UT] fix dis test
Jan 6, 2025
9caab23
Merged PR 2353: Update version numbers and add missing headers
Jan 9, 2025
092061e
Merged PR 2355: [Tracer] support subscriptable type hint
Jan 16, 2025
ac45426
Merged PR 2345: [Tracer + Codegen] support grad mode & autocast
Jan 16, 2025
fc91977
Merged PR 2356: [Schedule] add interleaved 1f1b
Jan 20, 2025
2952d2b
Merged PR 2359: [BugFix] Insert multiref for single stage correctly
Jan 20, 2025
5a3f178
Merged PR 2351: [Codegen] add async support for pipeline parallelism
0xWJ Jan 22, 2025
30b4286
Merged PR 2357: [Trainer] Add partial parallelized module support.
0xWJ Jan 23, 2025
1b91d7e
Merged PR 2363: Merge github PR and update version
Feb 12, 2025
8c57978
Merged PR 2358: [AutoDist] Support pipeline_nstages option
Feb 12, 2025
078d187
Merged PR 2361: [Example] improve diff attention
Feb 12, 2025
d5e8b30
Merged PR 2366: [BugFix] Fix autodist's data parallel test
Feb 13, 2025
2a50991
Merged PR 2364: [UT] Fix pipeline_nstages UT
Feb 13, 2025
075979f
Merged PR 2368: [AutoDist] Fix when model too small autodist does not…
Feb 20, 2025
e81bb5d
Merged PR 2362: [Example] longrope
Mar 21, 2025
ffa8831
Merged PR 2370: [BugFix] Fix autocast in AutoDist
Mar 24, 2025
4390c16
Merged PR 2371: [BugFix] Fix AutoDist's implementation
Mar 26, 2025
eded791
Merged PR 2372: [Parser] Support new functions & Refine apex register
Mar 26, 2025
a1368ee
Merged PR 2373: [BugFix] torch load fix for pytorch 2.6
0xWJ Mar 26, 2025
e96c0f4
Merged PR 2378: add iter dataset support and stateful dataloader
0xWJ Jun 9, 2025
3f49650
Merged PR 2379: [Trainer]Load auto-merged checkpoint when world size …
0xWJ Jun 10, 2025
ac9bb54
Merged PR 2377: [AutoDist] fix bug: calculate split info correctly
Jun 18, 2025
344083c
Merged PR 2380: add Fairseq example
0xWJ Jul 21, 2025
7ad94d1
Merged PR 2381: [AutoDist] Remove cppimport
Jul 22, 2025
ec51935
Merged PR 2382: [BugFix] TrainerArgs Resolution for command line args
0xWJ Jul 23, 2025
3537d43
Merged PR 2374: [Example] Refine ring-attention
Jul 24, 2025
c929f27
Merged PR 2383: [BugFix] Refine broadcasting merge state dict to avoi…
0xWJ Jul 24, 2025
df303c1
Merged PR 2384: [BugFix] dij implementation and deepseek coder v2 lit…
Jul 25, 2025
f01591e
Merged PR 2385: [BugFix] fix arange op
0xWJ Jul 25, 2025
7728b2a
Merged PR 2386: [BugFix] torch.compile support
0xWJ Aug 6, 2025
aa5ac3b
Merged PR 2389: [Feature][Breaking Change] Refine hooks
0xWJ Aug 7, 2025
65dfa75
Merged PR 2388: [AutoDist] Refine recompute modules implementation
Aug 7, 2025
fde55a2
Merged PR 2387: Add Copyright and update version of package datasets
Aug 12, 2025
385e7aa
Merged PR 2390: [BugFix] Add functools.cache support
0xWJ Aug 13, 2025
cbf14fc
Merged PR 2393: [Tracer] add torch compile check
0xWJ Aug 14, 2025
f0576f2
Merged PR 2391: [Runtime] Refine dedup ckpt save and load
Aug 14, 2025
ac1c72f
Merged PR 2394: [BugFix] fix bug in dedup load: handle persistent buf…
Aug 20, 2025
1474acd
Merged PR 2396: [AutoDist] Refine error handling and logging
Aug 21, 2025
401bea6
Merged PR 2397: [Runtime] Support offload params for parallel module
Aug 27, 2025
47ea468
Merged PR 2398: [Feature] Add multiple-optimizer/param groups support
0xWJ Aug 27, 2025
bf83275
Merged PR 2399: [Runtime] ParallelModule: move more instance member v…
0xWJ Sep 1, 2025
5039b80
Merged PR 2404: [Parser] Fix Stack and Add Dot
Sep 8, 2025
0ce3b73
Merged PR 2403: [Trainer] Refine logging related to dedup and dataloader
Sep 8, 2025
122fdf6
Merged PR 2400: [Runtime] Refine f16 optimizer loading logic
Sep 8, 2025
12c3f0e
Merged PR 2401: [Runtime] Add option to load merged state dict with l…
0xWJ Sep 9, 2025
e15b7f1
Merged PR 2406: [Runtime] Allow custom parameter bucketing
0xWJ Sep 12, 2025
e71c869
Merged PR 2407: [BwCompat]: add backward compatibiity for cli checkpo…
0xWJ Sep 15, 2025
fa5ec8b
Merged PR 2408: [Runtime] Bugfix: calculate gnorm correctly when exis…
Sep 17, 2025
eeae485
Merged PR 2409: [Refine] Parallelize: skip graph.forward call to simp…
0xWJ Sep 19, 2025
9c9b58f
Merged PR 2411: [Feature] add value tracker
0xWJ Oct 13, 2025
5d0db5a
Merged PR 2413: [BwCompat] add send/recv_object_list for pytorch < 2.4
0xWJ Oct 13, 2025
e6f2587
Merged PR 2414: [Refine] Make parser a real class
0xWJ Oct 17, 2025
9d4b8ba
Merged PR 2417: [Runtime] Refine ring attention related implementation
Oct 27, 2025
b9c9dbe
Merged PR 2418: [Feat] Add new-style policy API
0xWJ Oct 29, 2025
f6e5d60
Merged PR 2415: [Refine] Refine dynamic shape support
0xWJ Oct 29, 2025
9d303b2
Merged PR 2420: [Runtime] Add runtime pre-hook and post-hook for node
0xWJ Nov 3, 2025
1826f12
Merged PR 2419: [Test] Improve attn test coverage
Nov 20, 2025
b4ca05c
Merged PR 2423: [BugFix] Handle trim correctly when multi-node
Nov 20, 2025
85bd572
Merged PR 2425: [Feat] cli: Add safetensors support
0xWJ Dec 9, 2025
0b6735f
Merged PR 2421: [Feat] Add zero3 support
0xWJ Dec 15, 2025
bb6a819
Merged PR 2424: [Misc] Refine trainer interfaces
Dec 15, 2025
bb67d75
Merged PR 2428: [BugFix] don't set requires_grad on end2end model
0xWJ Dec 15, 2025
244214c
Merged PR 2431: [BugFix] Handle graph output correctly in dynamic pro…
Dec 16, 2025
137e28b
Merged PR 2432: [Bugfix] return cloned tensor when a parameter is mul…
0xWJ Dec 16, 2025
26744c0
Merged PR 2427: [BugFix] auto pack kwargs in dummy input
Dec 16, 2025
1c1c881
Merged PR 2435: [Bugfix] Fix the generated code when **kwargs are use…
0xWJ Dec 18, 2025
c6dd773
Merged PR 2438: [BwCompat] Fix backward compatiblity when metadata la…
0xWJ Dec 23, 2025
1d88d04
Merged PR 2442: [BugFix] Rebuild param groups after resuming in Hybri…
Dec 25, 2025
dbcc36d
Merged PR 2443: [Feat] Add option disable_shared_param_constraint and…
Dec 29, 2025
d1d8901
Merged PR 2437: [Feat] cli: add customized serialization runner
0xWJ Dec 30, 2025
91ccb4a
Fix & Update
Dec 31, 2025
3eb1bc2
Bump version to 0.8
lynex Dec 31, 2025
91ef006
Merged PR 2450: [AutoDist] Add dynamic constraint
Jan 4, 2026
9834d5e
Merged PR 2451: [Parser] Support `torch.eye` and `.T`
Jan 4, 2026
ccb7fa1
Merged PR 2441: [Feat] Add option: `reducer_pre_divisor`
Jan 4, 2026
4a78f4a
Merged PR 2447: [Refine] move dummy_input from trainer to trainer arg…
0xWJ Jan 5, 2026
442b243
Merged PR 2444: [Tracer] Fix for torch 2.8
Jan 5, 2026
5a902bc
Merged PR 2452: [HotFix] Merge z3 model states correctly
Jan 5, 2026
5b5cfc6
Merge remote-tracking branch 'nnscaler_devops/main' into devops
Jan 5, 2026
527c5b7
nit fix
Jan 5, 2026
5b259c3
Merge pull request #1 from msrasys/devops
lynex Jan 8, 2026
8ba2509
save work
0xWJ Jan 14, 2026
7c40eed
Merge pull request #3 from msrasys/weijiangxu/hotfix-zero0
0xWJ Jan 15, 2026
8f5366f
[Refine] Improve generated file broadcast with multithread writing an…
0xWJ Jan 15, 2026
ee5eb8f
[Feat] Add support for gathering full model state from all ranks (#4)
0xWJ Jan 19, 2026
fd8f704
[Refine] Reduce memory fragment when resuming
0xWJ Jan 19, 2026
e584751
refine comment
0xWJ Jan 19, 2026
8024fad
code refine
0xWJ Jan 20, 2026
2ff7e54
refine code
0xWJ Jan 20, 2026
0014141
refine comments
0xWJ Jan 20, 2026
9d5b02e
add more debug info
0xWJ Jan 21, 2026
c799251
refine comment
0xWJ Jan 21, 2026
943b154
refine code
0xWJ Jan 23, 2026
52d9322
add barrier
0xWJ Jan 24, 2026
a77b282
Merge pull request #6 from msrasys/weijiangxu/mem-fragment-in-resume
lynex Jan 26, 2026
68c9cbe
[Tracer] Provide better einops tracing by skipping tracing some inter…
0xWJ Jan 28, 2026
e39d68b
[Refine] Normalize device handling in state dicts and more (#9)
0xWJ Jan 30, 2026
cc97940
Add Doc Autodist Constraints Guide (#5)
yyl9510 Jan 30, 2026
cbbe521
CI/CD (#11)
yyl9510 Jan 30, 2026
8ce8b09
Add nightly test to the repo (#12)
yyl9510 Feb 3, 2026
b9e7993
[BugFix] Refine HybridOptimizer to support mixed precision optimizer …
0xWJ Feb 4, 2026
166635a
[Feat] Add Muon Support (dp without zero)
0xWJ Feb 4, 2026
1a9198c
Warmup triggers 150K times, put too much pressure on memory and lead …
lynex Feb 6, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
32 changes: 32 additions & 0 deletions .github/workflows/ci.yml
Original file line number Diff line number Diff line change
@@ -0,0 +1,32 @@
name: CI Pipeline

on:
push:
branches: [ main ]
pull_request:
branches: [ main ]

jobs:
test:
name: Run Tests
runs-on: ubuntu-latest

steps:
- name: Checkout code
uses: actions/checkout@v4

- name: Install system build tools
run: |
sudo apt-get update
sudo apt-get install -y build-essential python3-dev

- name: Set up uv
uses: astral-sh/setup-uv@v5
with:
enable-cache: true

- name: Install tox
run: uv tool install tox --with tox-uv

- name: Run unit tests
run: tox
21 changes: 21 additions & 0 deletions azure-pipelines.yml
Original file line number Diff line number Diff line change
@@ -0,0 +1,21 @@
# Starter pipeline
# Start with a minimal pipeline that you can customize to build and deploy your code.
# Add steps that build, run tests, deploy, and more:
# https://aka.ms/yaml

trigger:
- main

pool:
vmImage: ubuntu-latest

steps:
- script: |
conda tos accept --override-channels --channel https://repo.anaconda.com/pkgs/main;
conda tos accept --override-channels --channel https://repo.anaconda.com/pkgs/r;
pip install tox
pip install tox-conda
displayName: 'Install tox'
- script: |
tox
displayName: 'Run unit tests'
Loading