Skip to content

Actions: NVIDIA/TransformerEngine

Actions

Blossom-CI

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
blossom-ci.yml
2,500+ workflow runs
2,500+ 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
Add the getter and setter of skip_fp8_weight_update_tensor Blossom-CI #10003: Issue comment #3015 (comment) created by ptrendx
1s
1s
fix(grouped_linear): handle all-zero-token forward and backward Blossom-CI #10002: Issue comment #3019 (comment) created by jubick1337
1s
1s
Add MXFP8 quantized_model_init memory profiler for FSDP2 qinit analysis Blossom-CI #10001: Issue comment #3008 (comment) created by nvMelissa
1s
1s
[Common] Fix fused MoE aux loss for sequence aux loss Blossom-CI #10000: Issue comment #3018 (comment) created by ptrendx
2s
2s
Optimize function that loads pointers on GPU Blossom-CI #9999: Issue comment #3001 (comment) created by timmoon10
2s
2s
[Common/PyTorch/JAX] make offset of ClampedSwiGLU configurable Blossom-CI #9998: Issue comment #2938 (comment) created by timmoon10
2s
2s
[PyTorch Debug] Add scale_inv_std stat and skip NVFP4 layers in LogFp8TensorStats Blossom-CI #9997: Issue comment #3044 (comment) created by greptile-apps Bot
2s
2s
[PyTorch Debug] Add scale_inv_std stat and skip NVFP4 layers in LogFp8TensorStats Blossom-CI #9996: Issue comment #3044 (comment) created by pggPL
6s
6s
[PyTorch] Make modules.GroupedLinear graph-safe Blossom-CI #9995: Issue comment #3038 (comment) created by yaox12
1s
1s
docs: expand comm gemm overlap guidance Blossom-CI #9994: Issue comment #3043 (comment) created by omribz156
1s
1s
docs: expand comm gemm overlap guidance Blossom-CI #9993: Issue comment #3043 (comment) created by greptile-apps Bot
1s
1s
adding NVIDIA_TF32_OVERRIDE=0 to test_numerics.py Blossom-CI #9992: Issue comment #3014 (comment) created by francesco-bertolotti
2s
2s
[common] Grouped gemm update - nvfp4 for blackwell and fp8 blockwise hopper Blossom-CI #9991: Issue comment #2971 (comment) created by pggPL
2s
2s
[PyTorch Debug] Fix scale_inv_min returning 0 for MXFP8/NVFP4 Blossom-CI #9990: Issue comment #3041 (comment) created by pggPL
1s
1s
[PyTorch Debug] Fix scale_inv_min returning 0 for MXFP8/NVFP4 Blossom-CI #9989: Issue comment #3041 (comment) created by greptile-apps Bot
2s
2s
[PyTorch] Make modules.GroupedLinear graph-safe Blossom-CI #9988: Issue comment #3038 (comment) created by yaox12
1s
1s
[PyTorch debug] FakeQuant: support Float8BlockScaling and fix MoE / w… Blossom-CI #9987: Issue comment #3040 (comment) created by greptile-apps Bot
1s
1s
[PyTorch] Enable head dim 256 for FA4 Blossom-CI #9986: Issue comment #2932 (comment) created by yaox12
1s
1s
[PyTorch] Add FA4 Support Blossom-CI #9985: Issue comment #2432 (comment) created by aflah02
1s
1s
[PyTorch] Add pad_between_seqs support for non-CP and CP (A2A and P2P) with FA3 + THD (varlen) Blossom-CI #9984: Issue comment #2596 (comment) created by sudhakarsingh27
1s
1s
[JAX] Add an MoE Block (Layer) that compound router, permutation, groupedGEMM and communication Blossom-CI #9983: Issue comment #2912 (comment) created by jberchtold-nvidia
1s
1s
[Common/PyTorch/JAX] make offset of ClampedSwiGLU configurable Blossom-CI #9982: Issue comment #2938 (comment) created by timmoon10
1s
1s
Implement 4over6 NVFP4 recipe Blossom-CI #9981: Issue comment #2972 (comment) created by timmoon10
1s
1s
[JAX] Expert Parallelism: JAX primitives + VJPs Blossom-CI #9980: Issue comment #3036 (comment) created by phu0ngng
2s
2s
[Common] NVTETensor peer-handle annotation + nccl_comm backend Blossom-CI #9979: Issue comment #3017 (comment) created by phu0ngng
2s
2s