Skip to content

Pull requests: google/qwix

New pull request New
30 Open 338 Closed
30 Open 338 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Add Gemma 4 E2B MLP quantization and multi-pass sweep configuration.
#375 opened Sep 15, 2026 by copybara-service Bot Loading…
Support hierarchical scaling with per-axis BF16 and subchannel microscaling in QWIX.
#374 opened Sep 15, 2026 by copybara-service Bot Loading…
Support approximate multi-pass INT8 matrix multiplication (2-pass asymmetric and 3-pass triangular) in QWIX
#373 opened Sep 15, 2026 by copybara-service Bot Loading…
Support hybrid multi-pass quantization (mxfp8_16 Pass 1 + mxfp4 Pass 2) in QWIX
#372 opened Sep 15, 2026 by copybara-service Bot Loading…
Support bit-exact INT8 matrix multiplication via INT4 building blocks in QWIX
#371 opened Sep 15, 2026 by copybara-service Bot Loading…
Use jax.lax.scaled_dot for MXFP dot_general in Qwix instead of jax.nn.scaled_matmul that binds to cuDNN
#369 opened Sep 14, 2026 by copybara-service Bot Loading…
Support multi-pass residual quantization emulation for mxfp8_16 in QWIX
#367 opened Sep 11, 2026 by copybara-service Bot Loading…
Support configurable power-of-2 scaling ('oas' vs 'ceil') for floating-point microscaling in QWIX
#366 opened Sep 11, 2026 by copybara-service Bot Loading…
Fix _is_same_quantization_schema to correctly handle broadcast-able scales.
#359 opened Sep 4, 2026 by copybara-service Bot Loading…
feat(int4): add nibble_pack/unpack for true 0.5 B/elem int4 storage
#329 opened Jul 8, 2026 by lokic233 Loading…
1
docs: document MXFP numerics (mxfp8/mxfp4/nvfp4) and hardware dispatch
#324 opened Jul 6, 2026 by lokic233 Loading…
2
[pallas:mosaic] Migrated all uses of pltpu.emit_pipeline to explicit indices
#322 opened Jul 3, 2026 by copybara-service Bot Loading…
Force int64 bias for w8a16
#320 opened Jun 29, 2026 by copybara-service Bot Loading…
Adding test
#319 opened Jun 24, 2026 by copybara-service Bot Loading…
Suppress new Pyrefly findings before upgrading to V1.1.1
#310 opened Jun 22, 2026 by copybara-service Bot Loading…
Introduce HiQArray and its HiJax integration.
#293 opened Jun 2, 2026 by copybara-service Bot Loading…
Internal
#290 opened Jun 1, 2026 by copybara-service Bot Loading…
Rename save/load_pytree to save/load. Eliminate most user-facing "pytree" terminology in favor of "state" as a more specific term. Add deprecations.py for handling deprecated public functions.
#280 opened May 18, 2026 by copybara-service Bot Loading…
Add Sparsity NNX module
#236 opened Mar 27, 2026 by BirdsOfAFthr Contributor Loading…
2 tasks done
Internal
#232 opened Mar 26, 2026 by copybara-service Bot Loading…
Just updated the README.md
#209 opened Feb 5, 2026 by lshariprasad Loading…
1 of 2 tasks
[Tunix] Add _put_prompts_to_queue to handle dataset iteration and partial batches.
#202 opened Jan 29, 2026 by copybara-service Bot Loading…
Remove QuantizedArray class.
#180 opened Nov 24, 2025 by copybara-service Bot Loading…
Remove support for QuantizedArray inputs to attention.
#179 opened Nov 24, 2025 by copybara-service Bot Loading…
Remove the ptq change, move to another cl or use model surgery
#156 opened Nov 7, 2025 by copybara-service Bot Loading…
1
ProTip! Exclude everything labeled bug with -label:bug.