Commits

Branch selector

 master

User selector

fegin

Datepicker

All time

Commit History

Commits on Aug 30, 2025

• [SymmMEM] Move AsyncTP tests to a seperate test class (#161820)

  Show description for c1e504e
  

  fegin

  authored and

  pytorchmergebot

  committed
  c1e504e

  Copy full SHA for c1e504e

Commits on Aug 29, 2025

• [SymmMEM] Fix test_empty_strided_p2p_persistent (#161677)

  Show description for cd6d63f
  

  fegin

  authored and

  pytorchmergebot

  committed
  cd6d63f

  Copy full SHA for cd6d63f

• [BE][SymmMEM] Change Optional to the shorthand expression for symmetric memory modules (#161676)

  Show description for d8a0bdb
  

  fegin

  authored and

  pytorchmergebot

  committed
  d8a0bdb

  Copy full SHA for d8a0bdb

Commits on May 16, 2025

• [DSD] Don't pop tensors if they are on Meta device (#153185)

  Show description for 1503b3f
  

  fegin

  authored and

  pytorchmergebot

  committed
  1503b3f

  Copy full SHA for 1503b3f

Commits on May 15, 2025

• Fix AsyncMM not compiled with SM90a issue (#153519)

  Show description for b992a66
  

  fegin

  authored and

  pytorchmergebot

  committed
  b992a66

  Copy full SHA for b992a66

Commits on May 12, 2025

• Fix test_fused_scaled_matmul_reduce_scatter when scatter_dim is 0 (#153286)

  Show description for 498f364
  

  fegin

  authored and

  pytorchmergebot

  committed
  498f364

  Copy full SHA for 498f364

• [DSD] Don't pop tensors if they are on Meta device (#153185)

  Show description for 7243c69
  

  fegin

  authored and

  pytorchmergebot

  committed
  7243c69

  Copy full SHA for 7243c69

Commits on May 2, 2025

• [CP] Fix the offsets to KV in backward (#152625)

  Show description for 36e5ff6
  

  fegin

  authored and

  pytorchmergebot

  committed
  36e5ff6

  Copy full SHA for 36e5ff6

Commits on Apr 25, 2025

• [CP] Use TorchFunctionMode to dispatch SDPA for CP (#147902)

  Show description for 6aa9280
  

  fegin

  authored and

  pytorchmergebot

  committed
  6aa9280

  Copy full SHA for 6aa9280

Commits on Apr 15, 2025

• Fix score_mod.py dynamic max autotune for backward (#151270)

  Show description for ccfce9a
  

  fegin

  authored and

  pytorchmergebot

  committed
  ccfce9a

  Copy full SHA for ccfce9a

Commits on Mar 11, 2025

• [DSD] Update the document to mention the limitation of set_optimizer_state_dict (#148918)

  Show description for 52acc1f
  

  fegin

  authored and

  pytorchmergebot

  committed
  52acc1f

  Copy full SHA for 52acc1f

Commits on Mar 10, 2025

• [DSD] Fix the shared parameter mismatch for optimizer state_dict when flattening FQNs are used (#148825)

  Show description for ed969d1
  

  fegin

  authored and

  pytorchmergebot

  committed
  ed969d1

  Copy full SHA for ed969d1

Commits on Feb 25, 2025

• [DSD] Fixes issue when there is a PG without parameters (#147730)

  Show description for d73b927
  

  fegin

  authored and

  pytorchmergebot

  committed
  d73b927

  Copy full SHA for d73b927

Commits on Feb 12, 2025

• Let _create_cpu_state_dict and _copy_state_dict support DTensor (#146852)

  Show description for 0de27ee
  

  fegin

  authored and

  pytorchmergebot

  committed
  0de27ee

  Copy full SHA for 0de27ee

Commits on Jan 17, 2025

• [DSD][BE] Rewrite some tests to remove `with_comms` (#143241)

  Show description for 6077102
  

  fegin

  authored and

  pytorchmergebot

  committed
  6077102

  Copy full SHA for 6077102

Commits on Jan 16, 2025

• [BE][CP] Use run_subtests instead of parametrize (#143240)

  Show description for 573fc42
  

  fegin

  authored and

  pytorchmergebot

  committed
  573fc42

  Copy full SHA for 573fc42

Commits on Dec 12, 2024

• [DSD] Fix strict=False case for DDP (#143038)

  Show description for 602c86a
  

  fegin

  authored and

  pytorchmergebot

  committed
  602c86a

  Copy full SHA for 602c86a

Commits on Dec 10, 2024

• [DeviceMesh] Call no_dispatch before doing tensor slicing in DeviceMesh (#142287)

  Show description for daa27fe
  

  fegin

  authored and

  pytorchmergebot

  committed
  daa27fe

  Copy full SHA for daa27fe

Commits on Dec 3, 2024

• Initialize lr as a tensor if it is originally a tensor (#141620)

  Show description for 8b5c262
  

  fegin

  authored and

  pytorchmergebot

  committed
  8b5c262

  Copy full SHA for 8b5c262

Commits on Nov 15, 2024

• [CP] Enable CP + compiler tests when there are more than 2 GPUs (#133736)

  Show description for 924c1fe
  

  fegin

  authored and

  pytorchmergebot

  committed
  924c1fe

  Copy full SHA for 924c1fe

Commits on Oct 22, 2024

• [CP] Implement AllGather based context parallelism (#132820)

  Show description for 0b4a071
  

  fegin

  authored and

  pytorchmergebot

  committed
  0b4a071

  Copy full SHA for 0b4a071

Commits on Sep 10, 2024

• [DCP] Fixes the stateless optimizer issue of distributed state_dict (#135535)

  Show description for 1d9feff
  

  fegin

  authored and

  pytorchmergebot

  committed
  1d9feff

  Copy full SHA for 1d9feff

Commits on Sep 9, 2024

• [CP] Extend CP to support load-balancing shards (#132442)

  Show description for 21241bf
  

  fegin

  authored and

  pytorchmergebot

  committed
  21241bf

  Copy full SHA for 21241bf

Commits on Aug 28, 2024

• [DCP] Fixes the BC issue where the traversal doesn't support versions before 2.4 (#134158)

  Show description for c7338f4
  

  fegin

  authored and

  pytorchmergebot

  committed
  c7338f4

  Copy full SHA for c7338f4

Commits on Aug 15, 2024

• [CP] Rewrite ring attention backward algorithm and enablement APIs (#131351)

  Show description for 3434a54
  

  fegin

  authored and

  pytorchmergebot

  committed
  3434a54

  Copy full SHA for 3434a54

Commits on Aug 14, 2024

• [FSDP2] Enable HSDP + TP (#133335)

  Show description for d114fd7
  

  fegin

  authored and

  pytorchmergebot

  committed
  d114fd7

  Copy full SHA for d114fd7

Commits on Aug 12, 2024

• [BE] Raise when the target model has scalar parameters (#132934)

  Show description for d53dfa4
  

  fegin

  authored and

  pytorchmergebot

  committed
  d53dfa4

  Copy full SHA for d53dfa4

Commits on Aug 9, 2024

• [BE] Raise when the target model has scalar parameters (#132934)

  Show description for ea00036
  

  fegin

  authored and

  pytorchmergebot

  committed
  ea00036

  Copy full SHA for ea00036

Commits on Aug 8, 2024

• [BE] Test foreach optimizer for FSDP1 optimizer state_dict (#132933)

  Show description for 9e584d0
  

  fegin

  authored and

  pytorchmergebot

  committed
  9e584d0

  Copy full SHA for 9e584d0

• [BE] Using with_temp_dir for test_distributed_checkpoint (#132908)

  Show description for 745665d
  

  fegin

  authored and

  pytorchmergebot

  committed
  745665d

  Copy full SHA for 745665d

Commits on Aug 7, 2024

• Allow distributed breakpoint to skip the first few calls (#129511)

  Show description for 21d4c48
  

  fegin

  authored and

  pytorchmergebot

  committed
  21d4c48

  Copy full SHA for 21d4c48

Commits on Aug 2, 2024

• Only make wait_tensor as a side_effect op (#132341)

  Show description for bc51091
  

  fegin

  authored and

  pytorchmergebot

  committed
  bc51091

  Copy full SHA for bc51091

Commits on Jul 12, 2024

• [CP] Make CP loss curve on par with TP (#129515)

  Show description for dafef3f
  

  fegin

  authored and

  pytorchmergebot

  committed
  dafef3f

  Copy full SHA for dafef3f

• [CP] Fix the incorrect ring schedule in the fwd and bwd (#129514)

  Show description for e568c91
  

  fegin

  authored and

  pytorchmergebot

  committed
  e568c91

  Copy full SHA for e568c91

• [dtensor] Add dtensor to TORCH_LOGS (#129512)

  Show description for 0d8dedb
  

  fegin

  authored and

  pytorchmergebot

  committed
  0d8dedb

  Copy full SHA for 0d8dedb

Pagination

Previous

Next