8000 Commits · pytorch/ao · GitHub

More Web Proxy on the site http://driver.im/

Commits

Commits on Jun 21, 2025

Build mxfp4 kernel for sm120a (#2285)
gau-nernst
authored

Commits on Jun 4, 2025

[optim] Fix bug when default dtype is BF16 (#2286)
gau-nernst
authored

Commits on May 10, 2025

[optim] Fix low-bit optim when used with FSDP2+CPUOffload (#2195)
gau-nernst
authored

Commits on Feb 5, 2025

Commits on Feb 2, 2025

Fix LR scheduler issue with CPU offload optimizer (#1649)
gau-nernst
authored

Commits on Nov 26, 2024

Commits on Nov 9, 2024

[Low-bit optim] Support for `dcp.save()` and `dcp.load()` (#1217)

gau-nernst
and
msaroufim
authored

Commits on Nov 7, 2024

Add module-swap UX for INT8 mixed-precision training (#1179)
gau-nernst
authored

Commits on Nov 4, 2024

[NF4] Add `quantize_()` API support for NF4 (#1216)
gau-nernst
authored

Commits on Nov 2, 2024

[CPU offload optim] Fix when there are non-trainable params (#1210)
gau-nernst
authored

Commits on Oct 23, 2024

Add BF16 stochastic rounding option for optimizers (#1124)
gau-nernst
authored

Commits on Oct 21, 2024

Fix out-of-bounds memory access in Galore dequant kernel (#1125)
gau-nernst
authored

Commits on Oct 14, 2024

[low-bit optim] Upcast everything to FP32 for internal calculations (#1068)

gau-nernst
and
msaroufim
authored

Commits on Oct 6, 2024

[low-bit optim] Fix load state dict when device is different (#1021)
gau-nernst
authored

Commits on Oct 1, 2024

BitNet b1.58 training (#930)
gau-nernst
authored

Commits on Sep 29, 2024

[low-bit optim] Update docs on supported platforms and caveats (#971)
gau-nernst
authored

Commits on Sep 26, 2024

Commits on Sep 24, 2024

[low-bit optim] Change 8-bit and FP8 optim block size from 2048 to 256 to match new bnb v0.44 (#927)
gau-nernst
authored

Commits on Sep 9, 2024

Add INT8 mixed-precision training (#748)
gau-nernst
authored

Commits on Sep 5, 2024

[Low-bit optim] Improve compile time + Fix PyTorch 2.3 support for 4-bit optim (#812)
gau-nernst
authored

Commits on Sep 2, 2024

Commits on Aug 24, 2024

Add AdamW to `CPUOffloadOptimizer` default (#742)
gau-nernst
authored

Commits on Aug 23, 2024

Fix LR schedule handling for low-bit optimizers (#736)
gau-nernst
authored

Commits on Aug 22, 2024

[Llama] Make torchao's Llama trainable (#728)
gau-nernst
authored

Commits on Aug 19, 2024

Add option to move param to `device` before quantization (#699)
gau-nernst
authored

Commits on Aug 16, 2024

Add experimental INT8 quantized training (#644)
gau-nernst
authored

Commits on Aug 8, 2024

Add CUDA compute capability compile guard (#636)
gau-nernst
authored

Commits on Aug 6, 2024

Optimizer CPU offload for single GPU training (#584)
gau-nernst
authored

Commits on Aug 5, 2024

Fix FP6-LLM API and add `.to(device)` op (#595)
gau-nernst
authored

Commits on Jul 26, 2024

Improve FSDP support for low-bit optimizers (#538)
gau-nernst
authored

Commits on Jul 18, 2024

Parallel test with pytest-xdist (#518)
gau-nernst
authored

0