pytorch ao issues - Githubissues

pytorch / ao

PyTorch native quantization and sparsity for training and inference

BSD 3-Clause "New" or "Revised" License

1.6k stars 179 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

[WIP] cli.py on modal

#1349 cpuhrsch opened 53 minutes ago
1
Fixed invalid url in citation section

#1348 galqiwi closed 48 minutes ago
1
Remove lm_eval warning

#1347 msaroufim closed 48 minutes ago
1
Remove LM eval warning in GPTQ

#1346 msaroufim closed 2 hours ago
1
msaroufim/removewarning

#1345 msaroufim closed 2 hours ago
1
Add support for quantize_() with Float8Linear module

#1344 jainapurva opened 3 hours ago
1
Fix Failing CI - Update bitsandbytes import

#1343 jainapurva closed 6 hours ago
1
Self-Compression QAT and Linear

#1342 lliangthomas opened 9 hours ago
3
float8 training: move module attribute setting to sync function

#1341 vkuzo closed 5 hours ago
1
Update build_wheels_linux.yml - Pass test-infra input params

#1340 atalman closed 11 hours ago
1
check `scale.ndim` before applying `t`/`transpose`

#1339 crcrpar opened 13 hours ago
1
[CI] CUDA nightly regression test is failing due to bnb + `triton.ops`

#1338 gau-nernst closed 3 hours ago
0
Update SAM AMG README with more descriptive install instructions

#1337 philipbutler opened 1 day ago
1
SAM2 AMG cli and other QoL improvements

#1336 cpuhrsch closed 2 days ago
1
Very large discrepancy in the quantized model's output compared to the original model when quantizing on CPU

#1335 JohnnyRacer opened 3 days ago
4
fix torchtitan + float8 + delayed + compile

#1334 vkuzo closed 11 hours ago
1
Update nightly job to use 12.4 since 12.1 is deprecated

#1333 jerryzh168 closed 2 days ago
1
Update SAM2 configs with ao targets

#1332 philipbutler closed 2 days ago
1
[wip] unbreak float8 + delayed + compile + AC

#1331 vkuzo opened 3 days ago
1
Update nightly job to use 12.4 since 12.1 is deprecated

#1330 jerryzh168 closed 2 days ago
2
float8 training: fix bug with AC + compile

#1329 vkuzo closed 3 days ago
1
Some improvement to make autoquant v2 work with Mixtral-8x7B-v0.1

#1328 jerryzh168 closed 3 days ago
2
fix bug with float8 training + FSDP2 + TP

#1327 vkuzo closed 3 days ago
1
[float8] Allow specifying arbitrary dtype for each tensor

#1326 lw opened 3 days ago
1
[float8] Re-enable slow-accum in the bwd of axis-wise scaling schemes

#1325 lw opened 3 days ago
1
Enable CPU Offload for Intel GPU

#1324 dbyoung18 closed 45 minutes ago
6
Skipping nvcc compilation for sm equal or lower than 80

#1323 jerryzh168 closed 4 days ago
1
metal lowbit kernels: executorch ops

#1322 manuelcandales opened 4 days ago
3
Use unittest instead of pytest everywhere

#1321 jerryzh168 opened 4 days ago
0
[Feature Request] Support of `int8_dynamic_activation_int8_weight` with asymmetrically quantized weights

#1320 sanchitintel opened 5 days ago
2
Update README.md

#1319 drisspg closed 5 days ago
1
fix typo of README.md

#1318 22dimensions closed 5 days ago
3
`int8_dynamic_activation_int8_weight` uses zero-points for weight when activation is asymmetrically quantized

#1317 sanchitintel opened 6 days ago
2
Fix pickle.dump missing file argument typo in README

#1316 readleyj closed 6 days ago
3
How to trigger torchao unit tests?

#1315 goldhuang opened 6 days ago
3
Add hardware check to fp8 quant

#1314 jainapurva closed 3 hours ago
1
`attempting to run aten.abs.default, this is not supported` with latest torchtitan + torchao

#1313 lchu-ibm closed 3 days ago
3
[NF4] `.to()` fixes

#1312 gau-nernst opened 6 days ago
2
torch.compile(sync_float8_amax_and_scale_history) not working with triton latest main

#1311 goldhuang opened 6 days ago
2
[NF4] Various bugs in how NF4 handles `.to()` to move to a different device

#1310 gau-nernst opened 6 days ago
0
[AQT] Failed to move compiled module with AQT to a different device

#1309 gau-nernst opened 1 week ago
2
[float8] DDP GPT1.5B Torch.compile dynamo error

#1308 OrenLeung opened 1 week ago
1
Added cpu support for llama generate.py/eval.py

#1307 jainapurva closed 5 days ago
2
float8 with delayed scaling: fix autocast handling

#1306 vkuzo closed 6 days ago
1
Lint test dtypes

#1305 jainapurva closed 1 week ago
1
cmake torchao_ops_mps_linear_fp_act_xbit_weight

#1304 manuelcandales closed 4 days ago
12
Lint fixes for test/float8

#1303 jainapurva closed 1 week ago
1
migrate to linux_job_v2 and manylinux 2_28

#1302 HDCharles closed 1 week ago
1
DO NOT LAND TEST BUTTERFLY BOT

#1301 jcaip opened 1 week ago
2
pip install torchao cannot get latest versions (only 0.1 and 2 other version in the same level)

#1300 moreAImore opened 1 week ago
8