From bc97d48fc7ec93b23dea54e1d0a635724e1711a4 Mon Sep 17 00:00:00 2001 From: yiliu30 Date: Thu, 20 Nov 2025 19:04:09 -0800 Subject: [PATCH 1/5] update autoround version Signed-off-by: yiliu30 --- setup.py | 3 +-- src/llmcompressor/modifiers/autoround/base.py | 2 +- 2 files changed, 2 insertions(+), 3 deletions(-) diff --git a/setup.py b/setup.py index 05e1ad0c1..c8065de2c 100644 --- a/setup.py +++ b/setup.py @@ -144,8 +144,7 @@ def localversion_func(version: ScmVersion) -> str: if BUILD_TYPE == "release" else "compressed-tensors>=0.12.3a2" ), - # TODO: replace it with the release version - ("auto_round @ git+https://github.com/intel/auto-round.git@llmc"), + ("auto-round==0.9.0"), ], extras_require={ "dev": [ diff --git a/src/llmcompressor/modifiers/autoround/base.py b/src/llmcompressor/modifiers/autoround/base.py index 2480751a9..fd4b14bef 100644 --- a/src/llmcompressor/modifiers/autoround/base.py +++ b/src/llmcompressor/modifiers/autoround/base.py @@ -236,7 +236,7 @@ def apply_autoround(self, state, subgraph): block=decoding_layer, inputs=cur_inputs, q_input=self._q_input, - device=device, + device=str(device), # Leave offload for LLMC auto_offload=False, ) From 9ba113c107c8fcf6f1e73544e51201df39efe96e Mon Sep 17 00:00:00 2001 From: yiliu30 Date: Mon, 24 Nov 2025 04:50:32 -0800 Subject: [PATCH 2/5] expose bs Signed-off-by: yiliu30 --- src/llmcompressor/modifiers/autoround/base.py | 2 ++ 1 file changed, 2 insertions(+) diff --git a/src/llmcompressor/modifiers/autoround/base.py b/src/llmcompressor/modifiers/autoround/base.py index fd4b14bef..de2fa31e5 100644 --- a/src/llmcompressor/modifiers/autoround/base.py +++ b/src/llmcompressor/modifiers/autoround/base.py @@ -107,6 +107,7 @@ class AutoRoundModifier(Modifier, QuantizationMixin): # AutoRound modifier arguments iters: int = 200 enable_torch_compile: bool = True + batch_size: int = 8 # private variables _module_names: Dict[torch.nn.Module, str] = PrivateAttr(default_factory=dict) @@ -223,6 +224,7 @@ def apply_autoround(self, state, subgraph): scheme=ar_quant_scheme, iters=self.iters, enable_torch_compile=self.enable_torch_compile, + batch_size=self.batch_size, ) # TODO: configure layer-wise config based on self.resolved_config ar.configure_layer_config(enable_gguf_official_mixed=False) From 105033563ae4b3e118e7e7731be7ff545132e21e Mon Sep 17 00:00:00 2001 From: yiliu30 Date: Tue, 25 Nov 2025 22:22:30 -0800 Subject: [PATCH 3/5] use 0.9.1 Signed-off-by: yiliu30 --- setup.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/setup.py b/setup.py index c8065de2c..79325409c 100644 --- a/setup.py +++ b/setup.py @@ -144,7 +144,7 @@ def localversion_func(version: ScmVersion) -> str: if BUILD_TYPE == "release" else "compressed-tensors>=0.12.3a2" ), - ("auto-round==0.9.0"), + ("auto-round==0.9.1"), ], extras_require={ "dev": [ From 50e6682b2e9314717e58a93d539ea023a36f8a9d Mon Sep 17 00:00:00 2001 From: yiliu30 Date: Wed, 26 Nov 2025 19:47:56 -0800 Subject: [PATCH 4/5] fix Signed-off-by: yiliu30 --- setup.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/setup.py b/setup.py index 79325409c..d1363fe07 100644 --- a/setup.py +++ b/setup.py @@ -144,7 +144,7 @@ def localversion_func(version: ScmVersion) -> str: if BUILD_TYPE == "release" else "compressed-tensors>=0.12.3a2" ), - ("auto-round==0.9.1"), + "auto-round==0.9.1", ], extras_require={ "dev": [ From d139071b522d0a9b32a4a9a8e5b3365ddd1c20f3 Mon Sep 17 00:00:00 2001 From: yiliu30 Date: Wed, 26 Nov 2025 19:48:06 -0800 Subject: [PATCH 5/5] update Signed-off-by: yiliu30 --- setup.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/setup.py b/setup.py index d1363fe07..79325409c 100644 --- a/setup.py +++ b/setup.py @@ -144,7 +144,7 @@ def localversion_func(version: ScmVersion) -> str: if BUILD_TYPE == "release" else "compressed-tensors>=0.12.3a2" ), - "auto-round==0.9.1", + ("auto-round==0.9.1"), ], extras_require={ "dev": [