Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
23 commits
Select commit Hold shift + click to select a range
d04b771
Implement Mamba2Block Scan for multi-token sequence support
gramalingam Mar 31, 2026
b47460a
Fix HF dt_bias random initialization bug in NemotronH example
gramalingam Apr 2, 2026
208f468
Add mamba_scan flag to toggle Scan vs single-token Mamba2
gramalingam Apr 2, 2026
204a5b9
Add chunked SSD (multi-token) forward for Mamba2Block
gramalingam Apr 3, 2026
a0b0a17
Generalize mamba_scan flag to 3-way: chunked_ssd, scan, single
gramalingam Apr 3, 2026
fa2b7d3
fix: reload out_proj.weight from checkpoint in HF patch
gramalingam Apr 3, 2026
12005cb
chore: change mamba_scan default from chunked_ssd to single
gramalingam Apr 3, 2026
4f9a993
feat: add interactive mode to nemotron example script
gramalingam Apr 3, 2026
15c3d1c
feat: stream tokens to stdout during ONNX generation
gramalingam Apr 3, 2026
15532b7
feat: add repetition penalty to nemotron example script
gramalingam Apr 3, 2026
027d9e1
chore: suppress harmless HF model_type mismatch warning
gramalingam Apr 3, 2026
cd81309
fix(example): use rescale_prenorm_residual=False for HF weight loading
gramalingam Apr 7, 2026
704a06c
refactor(mamba2): split Mamba2Block into base + subclasses (Option B+D)
gramalingam Apr 7, 2026
5fdf3dd
fix: address PR #110 review feedback
gramalingam Apr 8, 2026
78a3ff0
fix: resolve all lintrunner warnings
gramalingam Apr 8, 2026
d18e99b
chore: add .copilot to .gitignore and remove from tracking
gramalingam Apr 9, 2026
108bd1d
fix: Pad type mismatch when building with f16/bf16 dtype
gramalingam Apr 9, 2026
f6bdd26
fix: revert mamba_scan default to 'single'
gramalingam Apr 9, 2026
7a93dee
fix: use bool padding mask in NemotronH for bf16 compatibility
gramalingam Apr 9, 2026
ff9520f
fix: remove spurious RoPE from NemotronH attention layers
gramalingam Apr 10, 2026
f59a5f1
fix: pass execution_provider to build() in NemotronH example script
gramalingam Apr 10, 2026
84679c9
Merge remote-tracking branch 'origin/main' into rama/chunkscan
gramalingam Apr 10, 2026
b43f0a1
Fix mamba2 L1 smoke test: shape mismatch in test config
gramalingam Apr 10, 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
1 change: 1 addition & 0 deletions .gitignore
Original file line number Diff line number Diff line change
Expand Up @@ -223,3 +223,4 @@ cache_dir/**
.worktrees/
dashboard_preview.html
docs/feature-flags.md
.copilot/
Loading
Loading