Lower quant/dequant torch op to StableHLO#5763

Merged

lsy323 merged 23 commits intomasterfrom lsiyuan/quant-dequant-dispatch

Nov 28, 2023

+826-16

Commits on Nov 18, 2023

Add q/dq custom op in torch_xla
Siyuan Liu
committed
Add stablehlo converter change to support quantized tensor. Update
Siyuan Liu
committed
Add per-channel quant, add resnet test
Siyuan Liu
committed
add stablehlo patch for serializing qdq into bytecode

lsy323
authored and
Siyuan Liu
committed
add support of exporting qdq to tf.saved_model

lsy323
authored and
Siyuan Liu
committed
make xla quant/dequant op compatible with decomp table

lsy323
authored and
Siyuan Liu
committed
format

lsy323
authored and
Siyuan Liu
committed
untrack some tests

lsy323
authored and
Siyuan Liu
committed
move resnet test under stablehlo test folder

lsy323
authored and
Siyuan Liu
committed
add patch for stablehlo.uniform_quantize/dequantize bytecode serialization

lsy323
authored and
Siyuan Liu
committed
rename files

lsy323
authored and
Siyuan Liu
committed
use python dispatcher for qdq

lsy323
authored and
Siyuan Liu
committed
fix compile

lsy323
authored and
Siyuan Liu
committed
fix linter

lsy323
authored and
Siyuan Liu
committed
add per tensor/channel tests

lsy323
authored and
Siyuan Liu
committed
format

lsy323
authored and
Siyuan Liu
committed
have all scalers have positive values
Siyuan Liu
committed

Commits on Nov 20, 2023

add comment for STABLEHLO_BYTECODE_FROM_PRETTYPRINT
Siyuan Liu
committed

Commits on Nov 28, 2023

do not udpate save_torch_module_as_tf_saved_model signature to take graph module
Siyuan Liu
committed
add comment for QuantParams struct
Siyuan Liu
committed
add new lines to the end of header file
Siyuan Liu
committed
update test
Siyuan Liu
committed
add assertions, use int for zp
Siyuan Liu
committed