Search results

export mobile model with 4bit failed, but 8bit was ok.

ExecuTorchIssues related to ExecuTorch installation, export, or build. Mobile uses separate tagsIssues related to ExecuTorch installation, export, or build. Mobile uses separate tags

QuantizationIssues related to Quantization or torchaoIssues related to Quantization or torchao

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Closed (completed).

#1506 In pytorch/torchchat;

· by TheBetterSolution was closed on Feb 28, 2025

Bump torchao pin closer to nightly: 2/26/2025

CLA SignedThis label is managed by the Meta Open Source bot.This label is managed by the Meta Open Source bot.

QuantizationIssues related to Quantization or torchaoIssues related to Quantization or torchao

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Merged (completed).

pytorch/torchchatnumber 1505

#1505 In pytorch/torchchat;

· by Jack-Khuu was closed on Feb 26, 2025

Update ARM CPU experimental kernels from AO to leverage pip install

CLA SignedThis label is managed by the Meta Open Source bot.This label is managed by the Meta Open Source bot.

QuantizationIssues related to Quantization or torchaoIssues related to Quantization or torchao

Status: Merged (completed).

pytorch/torchchatnumber 1458

#1458 In pytorch/torchchat;

· by metascroy was closed on Mar 11, 2025

If scripts need <code>bash</code>, don't say to use <code>sh</code>

actionableItems in the backlog waiting for an appropriate impl/fixItems in the backlog waiting for an appropriate impl/fix

bugSomething isn't workingSomething isn't working

documentationImprovements or additions to documentationImprovements or additions to documentation

QuantizationIssues related to Quantization or torchaoIssues related to Quantization or torchao

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Closed (completed).

#1436 In pytorch/torchchat;

· by swolchok was closed on Dec 23, 2024

update torchao pin: optimized mps lowbit shaders

CLA SignedThis label is managed by the Meta Open Source bot.This label is managed by the Meta Open Source bot.

QuantizationIssues related to Quantization or torchaoIssues related to Quantization or torchao

Status: Merged (completed).

pytorch/torchchatnumber 1428

#1428 In pytorch/torchchat;

· by manuelcandales was closed on Dec 18, 2024

INT8 has a poor performance with groupsize &gt; 0 in Torchchat, compared with BF16 and INT8 groupsize == 0

QuantizationIssues related to Quantization or torchaoIssues related to Quantization or torchao

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Closed (completed).

#1427 In pytorch/torchchat;

· by yanbing-j was closed on Dec 19, 2024

[KNOWN BUG] gguf + GPU AOTI Inference bug due to PT version, fix in progress

bugSomething isn't workingSomething isn't working

CI InfraIssues related to CI infrastructure and setupIssues related to CI infrastructure and setup

Compile / AOTIIssues related to AOT Inductor and torch compileIssues related to AOT Inductor and torch compile

Known GapsThese are known Gaps/Issues/Bug items in torchchatThese are known Gaps/Issues/Bug items in torchchat

QuantizationIssues related to Quantization or torchaoIssues related to Quantization or torchao

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Closed (completed).

#1423 In pytorch/torchchat;

· by Jack-Khuu was closed on Dec 19, 2024

Update int4pack related in torchchat gguf

CLA SignedThis label is managed by the Meta Open Source bot.This label is managed by the Meta Open Source bot.

QuantizationIssues related to Quantization or torchaoIssues related to Quantization or torchao

Status: Merged (completed).

pytorch/torchchatnumber 1404

#1404 In pytorch/torchchat;

· by yanbing-j was closed on Dec 17, 2024

Update dead link in https://github.com/pytorch/torchchat/blob/main/docs/quantization.md

documentationImprovements or additions to documentationImprovements or additions to documentation

QuantizationIssues related to Quantization or torchaoIssues related to Quantization or torchao

Status: Closed (completed).

#1385 In pytorch/torchchat;

· by yanbing-j was closed on Dec 9, 2024

Replace WeightOnlyInt8Linear with TorchAO int8_weight_only quantization

CLA SignedThis label is managed by the Meta Open Source bot.This label is managed by the Meta Open Source bot.

QuantizationIssues related to Quantization or torchaoIssues related to Quantization or torchao

Status: Closed (abandoned).

pytorch/torchchatnumber 1328

#1328 In pytorch/torchchat;

· by vmpuri was closed on Feb 13, 2025

Implement the AO API in torchchat quantization handlers and unify logic.

CLA SignedThis label is managed by the Meta Open Source bot.This label is managed by the Meta Open Source bot.

QuantizationIssues related to Quantization or torchaoIssues related to Quantization or torchao

Status: Merged (completed).

pytorch/torchchatnumber 1291

#1291 In pytorch/torchchat;

· by mikekgfb was closed on Oct 23, 2024

int4_weight_only in Cuda compile := RuntimeError: _apply(): Couldn't swap Linear.weight

bugSomething isn't workingSomething isn't working

Compile / AOTIIssues related to AOT Inductor and torch compileIssues related to AOT Inductor and torch compile

QuantizationIssues related to Quantization or torchaoIssues related to Quantization or torchao

Status: Closed (completed).

#1125 In pytorch/torchchat;

· by Jack-Khuu was closed on Jan 6, 2025