huggingface/accelerate

Pull Requests Commits

fix triton version check (#3345)

faaany committed 1 year ago

Verified 02d25612

dataloader: check that in_order is in kwargs before trying to drop it (#3346)

dvrogozh committed 1 year ago

Verified fbfa53bc

[docs] fix typo, change "backoff_filter" to "backoff_factor" (#3296)

suchot committed 1 year ago

Verified d09040df

add torchdata version check to avoid "in_order" error (#3344)

faaany committed 1 year ago

Verified 828aae4e

Fix for offloading when using TorchAO >= 0.7.0 (#3332)

a-r-r-o-w committed 1 year ago

Verified f0b03055

latest bnb no longer has optim_args attribute on optimizer (#3311)

winglian committed 1 year ago

Verified 80973430

[tests] make cuda-only test case device-agnostic (#3340)

faaany committed 1 year ago

Verified c67d47ae

Fix offload generate tests (#3334)

SunMarc committed 1 year ago

Verified 8c423cff

feat(tpu): remove nprocs from xla.spawn (#3324)

tengomucho committed 1 year ago

Verified 95f34d62

Fixup docker build err (#3333)

muellerzr committed 1 year ago

Verified ba90f856

Bye bye torch <2 (#3331)

muellerzr committed 1 year ago

Verified b13aadcb

Ensure that tied parameter is children of module (#3327)

pablomlago committed 1 year ago

Verified 58f14364

Adding `keep_torch_compile` argument to `unwrap_model` and `extract_model_from_parallel`. (#3282)

ggoggam committed 1 year ago

Verified 54370d45

Use torch.xpu.mem_get_info for XPU (#3275)

dvrogozh committed 1 year ago

Verified d6d3e03c

Give example on how to handle gradient accumulation with cross-entropy (#3193)

ylacombe committed 1 year ago

Verified acfbf72a

fix: add max_memory to _init_infer_auto_device_map's return statement (#3279)

Nech-C committed 1 year ago

Verified 200c9eb7

Fix test_nested_hook (#3289)

SunMarc committed 1 year ago

Verified 7b2edc0b

fix load_state_dict for npu (#3211)

ji-huazhong committed 1 year ago

Verified b92fb477

[docs] no hard-coding cuda (#3270)

faaany committed 1 year ago

Verified 3e62fbb0

Fixed typos for Tutorials and Guides docs (#3274)

henryhmko committed 1 year ago

Verified cb8b7c63

[docs] use real path for `checkpoint` (#3220)

faaany committed 1 year ago

Verified aa16d695

fix typo (#3221)

faaany committed 1 year ago

Verified f9a2e790

[docs] update set-seed (#3228)

faaany committed 1 year ago

Verified 51fd482d

Fix: Resolve #3060, `preload_module_classes` is lost for nested modules (#3248)

wejoncy committed 1 year ago

Verified 60461ff7

Revert default behavior of `get_state_dict_from_offload` (#3253)

kylesayrs committed 1 year ago

Verified f8c77f05

Select the DeepSpeedCPUOptimizer based on the original optimizer class. (#3255)

eljandoubi committed 1 year ago

Verified b626ef5f

Update troubleshooting.md (#3259)

relh committed 1 year ago

Verified dd68af88

Fix: Resolve #3257 (#3261)

sunyi0505 committed 1 year ago

Verified 11818e65

Update deferring_execution.md (#3262)

max-yue committed 1 year ago

Verified 1f508a6d

support for wrapped schedulefree optimizer when using deepspeed (#3266)

winglian committed 1 year ago

Verified 4a100eef

Newer Older