tico 0.1.0.dev251013__py3-none-any.whl → 0.1.0.dev251014__py3-none-any.whl

tico/__init__.py

@@ -29,7 +29,7 @@ __all__ = [
 ]
 
 # THIS LINE IS AUTOMATICALLY GENERATED BY setup.py
-__version__ = "0.1.0.dev251013"
+__version__ = "0.1.0.dev251014"
 
 MINIMUM_SUPPORTED_VERSION = "2.5.0"
 SECURE_TORCH_VERSION = "2.6.0"

tico/experimental/quantization/{ptq/quant_config.py → config/ptq.py}

@@ -15,6 +15,7 @@
 from dataclasses import dataclass, field
 from typing import Any, Dict, Mapping, Type
 
+from tico.experimental.quantization.config.base import BaseConfig
 from tico.experimental.quantization.ptq.dtypes import DType
 from tico.experimental.quantization.ptq.observers.base import ObserverBase
 from tico.experimental.quantization.ptq.observers.minmax import MinMaxObserver
@@ -22,7 +23,7 @@ from tico.experimental.quantization.ptq.qscheme import QScheme
 
 
 @dataclass
-class QuantConfig:
+class PTQConfig(BaseConfig):
     """
     One object describes the quantization preferences for a single wrapper
     and its descendants.
@@ -56,7 +57,7 @@ class QuantConfig:
     ```python
     from ptq.observers import PercentileObserver
 
-    cfg = QuantConfig(
+    cfg = PTQConfig(
         default_dtype   = DType.uint(8),
         default_qscheme  = QScheme.PER_TENSOR_SYMM,        # <- global scheme
         default_observer = PercentileObserver,             # <- global algorithm
@@ -75,6 +76,10 @@ class QuantConfig:
     default_qscheme: QScheme = QScheme.PER_TENSOR_ASYMM
     overrides: Mapping[str, Mapping[str, Any]] = field(default_factory=dict)
 
+    @property
+    def name(self) -> str:
+        return "ptq"
+
     def get_kwargs(self, obs_name: str) -> Dict[str, Any]:
         """
         Return user-specified kwargs for *obs_name* inside **this** wrapper.
@@ -87,7 +92,7 @@ class QuantConfig:
         """
         return dict(self.overrides.get(obs_name, {}))
 
-    def child(self, scope: str) -> "QuantConfig":
+    def child(self, scope: str) -> "PTQConfig":
         """
         Produce a *view* for a child wrapper.
 
@@ -100,7 +105,7 @@ class QuantConfig:
         Other scopes remain invisible to the child.
         """
         sub_overrides = self.overrides.get(scope, {})
-        return QuantConfig(
+        return PTQConfig(
             self.default_dtype,
             self.default_observer,
             default_qscheme=self.default_qscheme,
@@ -108,4 +113,4 @@ class QuantConfig:
         )
 
     def __repr__(self):
-        return f"QuantConfig(default_dtype={self.default_dtype}, default_observer={self.default_observer}, default_qscheme={self.default_qscheme}, overrides={dict(self.overrides)})"
+        return f"PTQConfig(default_dtype={self.default_dtype}, default_observer={self.default_observer}, default_qscheme={self.default_qscheme}, overrides={dict(self.overrides)})"

tico/experimental/quantization/evaluation/utils.py

@@ -44,7 +44,7 @@ def quantize(
         data = np.array(data)
     # Perfrom quantization
     if not scale:
-        logger.warn("WARNING: scale value is 0. 1e-7 will be used instead.")
+        logger.warning("WARNING: scale value is 0. 1e-7 will be used instead.")
         scale = 1e-7
     rescaled = np.round(data / scale) + zero_point
     # Clamp the values

tico/experimental/quantization/ptq/examples/compare_ppl.py

@@ -29,7 +29,7 @@ import tqdm
 from datasets import load_dataset
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
-from tico.experimental.quantization.ptq.quant_config import QuantConfig
+from tico.experimental.quantization.config.ptq import PTQConfig
 from tico.experimental.quantization.ptq.utils.metrics import perplexity
 from tico.experimental.quantization.ptq.wrappers.ptq_wrapper import PTQWrapper
 
@@ -165,7 +165,7 @@ def main():
         # ---------------------------------------------------------------------
         # 2. Wrap every Transformer layer with PTQWrapper
         # ---------------------------------------------------------------------
-        qcfg = QuantConfig()  # all-uint8 defaults
+        qcfg = PTQConfig()  # all-uint8 defaults
 
         wrapped_layers = torch.nn.ModuleList()
         for idx, layer in enumerate(uint8_model.model.layers):

tico/experimental/quantization/ptq/examples/debug_quant_outputs.py

@@ -38,7 +38,7 @@ import tqdm
 from datasets import load_dataset
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
-from tico.experimental.quantization.ptq.quant_config import QuantConfig
+from tico.experimental.quantization.config.ptq import PTQConfig
 from tico.experimental.quantization.ptq.utils.introspection import (
     build_fqn_map,
     compare_layer_outputs,
@@ -176,7 +176,7 @@ def main():
     # 2. Wrap every layer with PTQWrapper (UINT-8 activations)
     # -------------------------------------------------------------------------
     print("Wrapping layers with PTQWrapper …")
-    qcfg = QuantConfig()  # default: per-tensor UINT8
+    qcfg = PTQConfig()  # default: per-tensor UINT8
 
     new_layers = torch.nn.ModuleList()
     for idx, fp_layer in enumerate(model.model.layers):

tico/experimental/quantization/ptq/examples/quantize_llama_mlp.py

@@ -18,12 +18,12 @@ import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
 import tico
+from tico.experimental.quantization.config.ptq import PTQConfig
 from tico.experimental.quantization.evaluation.metric import compute_peir
 from tico.experimental.quantization.evaluation.utils import plot_two_outputs
 from tico.experimental.quantization.ptq.dtypes import INT16
 from tico.experimental.quantization.ptq.mode import Mode
 from tico.experimental.quantization.ptq.qscheme import QScheme
-from tico.experimental.quantization.ptq.quant_config import QuantConfig
 from tico.experimental.quantization.ptq.wrappers.llama.quant_mlp import QuantLlamaMLP
 from tico.utils.utils import SuppressWarning
 
@@ -38,7 +38,7 @@ model.eval()
 fp32_mlp = model.model.layers[0].mlp
 model.model.layers[0].mlp = QuantLlamaMLP(
     fp32_mlp,
-    qcfg=QuantConfig(default_dtype=INT16, default_qscheme=QScheme.PER_TENSOR_SYMM),
+    qcfg=PTQConfig(default_dtype=INT16, default_qscheme=QScheme.PER_TENSOR_SYMM),
 )  # PTQWrapper(fp32_mlp) is also fine
 model.eval()
 

tico/experimental/quantization/ptq/examples/quantize_with_gptq.py

@@ -35,8 +35,8 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
 
 from tico.experimental.quantization import convert, prepare
 from tico.experimental.quantization.config.gptq import GPTQConfig
+from tico.experimental.quantization.config.ptq import PTQConfig
 from tico.experimental.quantization.ptq.observers.affine_base import AffineObserverBase
-from tico.experimental.quantization.ptq.quant_config import QuantConfig
 from tico.experimental.quantization.ptq.utils.introspection import build_fqn_map
 from tico.experimental.quantization.ptq.utils.metrics import perplexity
 from tico.experimental.quantization.ptq.wrappers.ptq_wrapper import PTQWrapper
@@ -219,7 +219,7 @@ def main():
     if not isinstance(layers, (list, torch.nn.ModuleList)):
         raise TypeError(f"'model.layers' must be list/ModuleList, got {type(layers)}")
 
-    qcfg = QuantConfig()  # default: per-tensor UINT8
+    qcfg = PTQConfig()  # default: per-tensor UINT8
     wrapped = torch.nn.ModuleList()
     for idx, fp_layer in enumerate(layers):
         layer_cfg = qcfg.child(f"layer{idx}")

tico/experimental/quantization/ptq/wrappers/fairseq/quant_decoder.py

@@ -25,7 +25,7 @@ import torch
 import torch.nn.functional as F
 from torch import nn, Tensor
 
-from tico.experimental.quantization.ptq.quant_config import QuantConfig
+from tico.experimental.quantization.config.ptq import PTQConfig
 from tico.experimental.quantization.ptq.wrappers.ptq_wrapper import PTQWrapper
 from tico.experimental.quantization.ptq.wrappers.quant_module_base import (
     QuantModuleBase,
@@ -53,7 +53,7 @@ class QuantFairseqDecoder(QuantModuleBase):
         self,
         fp_decoder: nn.Module,
         *,
-        qcfg: Optional[QuantConfig] = None,
+        qcfg: Optional[PTQConfig] = None,
         fp_name: Optional[str] = None,
     ):
         super().__init__(qcfg, fp_name=fp_name)
@@ -116,7 +116,7 @@ class QuantFairseqDecoder(QuantModuleBase):
 
         prefix = _safe_prefix(fp_name)
 
-        # Prepare child QuantConfig namespaces: layers/<idx>
+        # Prepare child PTQConfig namespaces: layers/<idx>
         layers_qcfg = qcfg.child("layers") if qcfg else None
         for i, layer in enumerate(fp_layers):
             child_cfg = layers_qcfg.child(str(i)) if layers_qcfg else None

tico/experimental/quantization/ptq/wrappers/fairseq/quant_decoder_layer.py

@@ -23,7 +23,7 @@ from typing import Dict, Iterable, List, Optional, Tuple
 import torch
 from torch import nn, Tensor
 
-from tico.experimental.quantization.ptq.quant_config import QuantConfig
+from tico.experimental.quantization.config.ptq import PTQConfig
 from tico.experimental.quantization.ptq.wrappers.fairseq.quant_mha import (
     QuantFairseqMultiheadAttention,
 )
@@ -55,7 +55,7 @@ class QuantFairseqDecoderLayer(QuantModuleBase):
         self,
         fp_layer: nn.Module,
         *,
-        qcfg: Optional[QuantConfig] = None,
+        qcfg: Optional[PTQConfig] = None,
         fp_name: Optional[str] = None,
     ):
         super().__init__(qcfg, fp_name=fp_name)

tico/experimental/quantization/ptq/wrappers/fairseq/quant_encoder.py

@@ -25,7 +25,7 @@ import torch
 import torch.nn as nn
 from torch import Tensor
 
-from tico.experimental.quantization.ptq.quant_config import QuantConfig
+from tico.experimental.quantization.config.ptq import PTQConfig
 from tico.experimental.quantization.ptq.wrappers.ptq_wrapper import PTQWrapper
 from tico.experimental.quantization.ptq.wrappers.quant_module_base import (
     QuantModuleBase,
@@ -56,7 +56,7 @@ class QuantFairseqEncoder(QuantModuleBase):
         self,
         fp_encoder: nn.Module,
         *,
-        qcfg: Optional[QuantConfig] = None,
+        qcfg: Optional[PTQConfig] = None,
         fp_name: Optional[str] = None,
         use_external_inputs: bool = False,  # export-mode flag
         return_type: Literal["tensor", "dict"] = "dict",
@@ -100,7 +100,7 @@ class QuantFairseqEncoder(QuantModuleBase):
         fp_layers = list(fp_encoder.layers)  # type: ignore[arg-type]
         self.layers = nn.ModuleList()
 
-        # Prepare child QuantConfig namespaces: layers/<idx>
+        # Prepare child PTQConfig namespaces: layers/<idx>
         layers_qcfg = qcfg.child("layers") if qcfg else None
         for i, layer in enumerate(fp_layers):
             child_cfg = layers_qcfg.child(str(i)) if layers_qcfg else None

tico/experimental/quantization/ptq/wrappers/fairseq/quant_encoder_layer.py

@@ -23,7 +23,7 @@ from typing import Optional
 import torch.nn as nn
 from torch import Tensor
 
-from tico.experimental.quantization.ptq.quant_config import QuantConfig
+from tico.experimental.quantization.config.ptq import PTQConfig
 from tico.experimental.quantization.ptq.wrappers.fairseq.quant_mha import (
     QuantFairseqMultiheadAttention,
 )
@@ -49,7 +49,7 @@ class QuantFairseqEncoderLayer(QuantModuleBase):
         self,
         fp_layer: nn.Module,
         *,
-        qcfg: Optional[QuantConfig] = None,
+        qcfg: Optional[PTQConfig] = None,
         fp_name: Optional[str] = None,
     ):
         super().__init__(qcfg, fp_name=fp_name)

tico/experimental/quantization/ptq/wrappers/fairseq/quant_mha.py

@@ -24,7 +24,7 @@ import torch
 import torch.nn as nn
 import torch.nn.functional as F
 
-from tico.experimental.quantization.ptq.quant_config import QuantConfig
+from tico.experimental.quantization.config.ptq import PTQConfig
 from tico.experimental.quantization.ptq.wrappers.ptq_wrapper import PTQWrapper
 from tico.experimental.quantization.ptq.wrappers.quant_module_base import (
     QuantModuleBase,
@@ -59,7 +59,7 @@ class QuantFairseqMultiheadAttention(QuantModuleBase):
         self,
         fp_attn: nn.Module,
         *,
-        qcfg: Optional[QuantConfig] = None,
+        qcfg: Optional[PTQConfig] = None,
         fp_name: Optional[str] = None,
         max_seq: int = 4096,
         use_static_causal: bool = False,

tico/experimental/quantization/ptq/wrappers/llama/quant_attn.py

@@ -17,7 +17,7 @@ from typing import Optional, Tuple
 import torch
 import torch.nn as nn
 
-from tico.experimental.quantization.ptq.quant_config import QuantConfig
+from tico.experimental.quantization.config.ptq import PTQConfig
 from tico.experimental.quantization.ptq.wrappers.ptq_wrapper import PTQWrapper
 from tico.experimental.quantization.ptq.wrappers.quant_module_base import (
     QuantModuleBase,
@@ -34,7 +34,7 @@ class QuantLlamaAttention(QuantModuleBase):
         self,
         fp_attn: nn.Module,
         *,
-        qcfg: Optional[QuantConfig] = None,
+        qcfg: Optional[PTQConfig] = None,
         fp_name: Optional[str] = None,
     ):
         super().__init__(qcfg, fp_name=fp_name)

tico/experimental/quantization/ptq/wrappers/llama/quant_decoder_layer.py

@@ -17,7 +17,7 @@ from typing import Optional, Tuple
 import torch
 import torch.nn as nn
 
-from tico.experimental.quantization.ptq.quant_config import QuantConfig
+from tico.experimental.quantization.config.ptq import PTQConfig
 from tico.experimental.quantization.ptq.wrappers.llama.quant_attn import (
     QuantLlamaAttention,
 )
@@ -56,7 +56,7 @@ class QuantLlamaDecoderLayer(QuantModuleBase):
         self,
         fp_layer: nn.Module,
         *,
-        qcfg: Optional[QuantConfig] = None,
+        qcfg: Optional[PTQConfig] = None,
         fp_name: Optional[str] = None,
         return_type: Optional[str] = None,
     ):
@@ -165,7 +165,7 @@ class QuantLlamaDecoderLayer(QuantModuleBase):
         # - If use_cache: always return (hidden_states, present_key_value)
         # - Else: return as configured (tuple/tensor) for HF compatibility
         if use_cache:
-            return hidden_states, present_key_value
+            return hidden_states, present_key_value  # type: ignore[return-value]
 
         if self.return_type == "tuple":
             return (hidden_states,)

tico/experimental/quantization/ptq/wrappers/llama/quant_mlp.py

@@ -17,7 +17,7 @@ from typing import Optional
 import torch
 import torch.nn as nn
 
-from tico.experimental.quantization.ptq.quant_config import QuantConfig
+from tico.experimental.quantization.config.ptq import PTQConfig
 from tico.experimental.quantization.ptq.wrappers.ptq_wrapper import PTQWrapper
 from tico.experimental.quantization.ptq.wrappers.quant_module_base import (
     QuantModuleBase,
@@ -31,7 +31,7 @@ class QuantLlamaMLP(QuantModuleBase):
         self,
         mlp_fp: nn.Module,
         *,
-        qcfg: Optional[QuantConfig] = None,
+        qcfg: Optional[PTQConfig] = None,
         fp_name: Optional[str] = None,
     ):
         super().__init__(qcfg, fp_name=fp_name)

tico/experimental/quantization/ptq/wrappers/nn/quant_layernorm.py

@@ -17,8 +17,9 @@ from typing import Iterable, Optional, Tuple
 import torch
 import torch.nn as nn
 
+from tico.experimental.quantization.config.ptq import PTQConfig
+
 from tico.experimental.quantization.ptq.mode import Mode
-from tico.experimental.quantization.ptq.quant_config import QuantConfig
 from tico.experimental.quantization.ptq.wrappers.quant_module_base import (
     QuantModuleBase,
 )
@@ -46,7 +47,7 @@ class QuantLayerNorm(QuantModuleBase):
         self,
         fp: nn.LayerNorm,
         *,
-        qcfg: Optional[QuantConfig] = None,
+        qcfg: Optional[PTQConfig] = None,
         fp_name: Optional[str] = None
     ):
         super().__init__(qcfg, fp_name=fp_name)

tico/experimental/quantization/ptq/wrappers/nn/quant_linear.py

@@ -17,9 +17,10 @@ from typing import Optional
 import torch.nn as nn
 import torch.nn.functional as F
 
+from tico.experimental.quantization.config.ptq import PTQConfig
+
 from tico.experimental.quantization.ptq.mode import Mode
 from tico.experimental.quantization.ptq.qscheme import QScheme
-from tico.experimental.quantization.ptq.quant_config import QuantConfig
 from tico.experimental.quantization.ptq.wrappers.quant_module_base import (
     QuantModuleBase,
 )
@@ -34,7 +35,7 @@ class QuantLinear(QuantModuleBase):
         self,
         fp: nn.Linear,
         *,
-        qcfg: Optional[QuantConfig] = None,
+        qcfg: Optional[PTQConfig] = None,
         fp_name: Optional[str] = None
     ):
         super().__init__(qcfg, fp_name=fp_name)

tico/experimental/quantization/ptq/wrappers/nn/quant_silu.py

@@ -17,7 +17,7 @@ from typing import Optional
 import torch
 import torch.nn as nn
 
-from tico.experimental.quantization.ptq.quant_config import QuantConfig
+from tico.experimental.quantization.config.ptq import PTQConfig
 from tico.experimental.quantization.ptq.wrappers.quant_module_base import (
     QuantModuleBase,
 )
@@ -37,7 +37,7 @@ class QuantSiLU(QuantModuleBase):
         self,
         fp: nn.SiLU,
         *,
-        qcfg: Optional[QuantConfig] = None,
+        qcfg: Optional[PTQConfig] = None,
         fp_name: Optional[str] = None
     ):
         super().__init__(qcfg, fp_name=fp_name)

tico/experimental/quantization/ptq/wrappers/ptq_wrapper.py

@@ -16,7 +16,7 @@ from typing import Optional
 
 import torch
 
-from tico.experimental.quantization.ptq.quant_config import QuantConfig
+from tico.experimental.quantization.config.ptq import PTQConfig
 from tico.experimental.quantization.ptq.wrappers.quant_module_base import (
     QuantModuleBase,
 )
@@ -34,7 +34,7 @@ class PTQWrapper(QuantModuleBase):
     def __init__(
         self,
         module: torch.nn.Module,
-        qcfg: Optional[QuantConfig] = None,
+        qcfg: Optional[PTQConfig] = None,
         *,
         fp_name: Optional[str] = None,
     ):

tico/experimental/quantization/ptq/wrappers/quant_elementwise.py

@@ -17,7 +17,7 @@ from typing import Callable, Optional
 import torch
 import torch.nn as nn
 
-from tico.experimental.quantization.ptq.quant_config import QuantConfig
+from tico.experimental.quantization.config.ptq import PTQConfig
 from tico.experimental.quantization.ptq.wrappers.quant_module_base import (
     QuantModuleBase,
 )
@@ -48,7 +48,7 @@ class QuantElementwise(QuantModuleBase):
         self,
         fp_module: nn.Module,
         *,
-        qcfg: Optional[QuantConfig] = None,
+        qcfg: Optional[PTQConfig] = None,
         fp_name: Optional[str] = None,
     ):
         super().__init__(qcfg, fp_name=fp_name)

tico/experimental/quantization/ptq/wrappers/quant_module_base.py

@@ -17,9 +17,10 @@ from typing import Iterable, Optional, Tuple
 
 import torch.nn as nn
 
+from tico.experimental.quantization.config.ptq import PTQConfig
+
 from tico.experimental.quantization.ptq.mode import Mode
 from tico.experimental.quantization.ptq.observers.base import ObserverBase
-from tico.experimental.quantization.ptq.quant_config import QuantConfig
 
 
 class QuantModuleBase(nn.Module, ABC):
@@ -29,7 +30,7 @@ class QuantModuleBase(nn.Module, ABC):
     Responsibilities
     ----------------
     • Own *one* Mode enum (`NO_QUANT / CALIB / QUANT`)
-    • Own a QuantConfig describing default / per-observer dtypes
+    • Own a PTQConfig describing default / per-observer dtypes
     • Expose a canonical lifecycle:
           enable_calibration()
           freeze_qparams()
@@ -38,10 +39,10 @@ class QuantModuleBase(nn.Module, ABC):
     """
 
     def __init__(
-        self, qcfg: Optional[QuantConfig] = None, *, fp_name: Optional[str] = None
+        self, qcfg: Optional[PTQConfig] = None, *, fp_name: Optional[str] = None
     ) -> None:
         super().__init__()
-        self.qcfg = qcfg or QuantConfig()
+        self.qcfg = qcfg or PTQConfig()
         self._mode: Mode = Mode.NO_QUANT  # default state
         self.fp_name = fp_name
 
@@ -118,9 +119,9 @@ class QuantModuleBase(nn.Module, ABC):
         Instantiate an observer named *name*.
 
         Precedence (3-tier) for keys:
-           • observer:  user > wrapper-default > QuantConfig.default_observer
-           • dtype:     user > wrapper-default > QuantConfig.default_dtype
-           • qscheme:   user > wrapper-default > QuantConfig.default_qscheme
+           • observer:  user > wrapper-default > PTQConfig.default_observer
+           • dtype:     user > wrapper-default > PTQConfig.default_dtype
+           • qscheme:   user > wrapper-default > PTQConfig.default_qscheme
 
         Other kwargs (e.g., qscheme, channel_axis, etc.) remain:
            user override > wrapper-default

tico/utils/convert.py

@@ -143,6 +143,7 @@ def traced_run_decompositions(exported_program: ExportedProgram):
         or torch.__version__.startswith("2.7")
         or torch.__version__.startswith("2.8")
         or torch.__version__.startswith("2.9")
+        or torch.__version__.startswith("2.10")
     ):
         return run_decompositions(exported_program)
     else:

{tico-0.1.0.dev251013.dist-info → tico-0.1.0.dev251014.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: tico
-Version: 0.1.0.dev251013
+Version: 0.1.0.dev251014
 Summary: Convert exported Torch module to circle
 Home-page: UNKNOWN
 License: UNKNOWN
@@ -68,7 +68,7 @@ This will generate `build` and `dist` directories in the root directory.
 **Available options**
 - `--dist` To install the package from .whl (without this option, _TICO_ is installed in an editable mode)
 - `--torch_ver <torch version>` To install a specific torch version (default: 2.6).
-  - Available <torch version>: 2.5, 2.6, 2.7, nightly
+  - Available <torch version>: 2.5, 2.6, 2.7, 2.8, nightly
 
 4. Now you can convert a torch module to a `.circle`.
 

{tico-0.1.0.dev251013.dist-info → tico-0.1.0.dev251014.dist-info}/RECORD

@@ -1,4 +1,4 @@
-tico/__init__.py,sha256=U-GkjWKgZwkndQ10KSKpe0dF_BMQeunWF7ktsiz0FGQ,1883
+tico/__init__.py,sha256=-_cU2NnQ66r8uJLRWR_AzjQdiPVMOTGKMriQY1yrnBA,1883
 tico/pt2_to_circle.py,sha256=gu3MD4Iqc0zMZcCZ2IT8oGbyj21CTSbT3Rgd9s2B_9A,2767
 tico/config/__init__.py,sha256=xZzCXjZ84qE-CsBi-dfaL05bqpQ3stKKfTXhnrJRyVs,142
 tico/config/base.py,sha256=q5xMqGxTUZs4mFqt5c7i_y9U00fYgdMGl9nUqIVMlCo,1248
@@ -44,12 +44,13 @@ tico/experimental/quantization/config/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3
 tico/experimental/quantization/config/base.py,sha256=xg_HCDSuMgYvMd6ENZe4Sm2SYJgMaCBj4cmqaz_lhAs,816
 tico/experimental/quantization/config/gptq.py,sha256=IUIEz5bLhsTXqoBCE1rfPec99zsRjwgpDbPW5YJqOPg,973
 tico/experimental/quantization/config/pt2e.py,sha256=9HCrraTGGZeKEN9puKV-ODi7ncV2Wjc3oe_JCO1D_Rs,850
+tico/experimental/quantization/config/ptq.py,sha256=uloDu-BKLJ9RussCmoLsw0Wq41zdk_iKsjdi_xqOn30,4431
 tico/experimental/quantization/config/smoothquant.py,sha256=b92dz4-MiBbkaLzXb47bVoO29d2P416woFQUZ1wpO_s,1414
 tico/experimental/quantization/evaluation/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3ouxARaxCK7bz9dn4blPQ,26
 tico/experimental/quantization/evaluation/backend.py,sha256=CZL9rZOA0t8cH7PHp6u9l7dGqWNvTj9bKOvwo0PVul0,692
 tico/experimental/quantization/evaluation/evaluate.py,sha256=kfa_GvFaX6DoSTAmuCImMJqF2jgqtnor5UpC7wVmGPI,7877
 tico/experimental/quantization/evaluation/metric.py,sha256=t9M058dOQ8iy_2PcrbNMAebBNJs8TU8USZw_nbi2iWI,5488
-tico/experimental/quantization/evaluation/utils.py,sha256=82RG_e5LuKfWo786wEZUVwXY93nNl901n04fB7D0Z6k,5909
+tico/experimental/quantization/evaluation/utils.py,sha256=n4Im3FiIVG3oVjB-wtIwV-0GUs24E6zS6Vc_cBnN5QQ,5912
 tico/experimental/quantization/evaluation/executor/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3ouxARaxCK7bz9dn4blPQ,26
 tico/experimental/quantization/evaluation/executor/backend_executor.py,sha256=3kLu3_rcsreA_NK42yRgRgubPtZmVp7QCRvaqLNw10E,1522
 tico/experimental/quantization/evaluation/executor/circle_executor.py,sha256=8o2Cb1fewPMtYd6jrn3nKHnSTGs7FOw-vD5_DZ2FVlE,2745
@@ -65,15 +66,14 @@ tico/experimental/quantization/ptq/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3oux
 tico/experimental/quantization/ptq/dtypes.py,sha256=xfCBtq6mQmUYRwsoFgII6gvRl1raQi0Inj9pznDuKwQ,2236
 tico/experimental/quantization/ptq/mode.py,sha256=lT-T8vIv8YWcwrjT7xXVhOw1g7aoAdh_3PWB-ptPKaI,1052
 tico/experimental/quantization/ptq/qscheme.py,sha256=uwhv7bCxOOXB3I-IKlRyr_u4eXOq48uIqGy4TLDqGxY,1301
-tico/experimental/quantization/ptq/quant_config.py,sha256=nm7570Y1X2mOT_8s27ilWid04otor6cVTi9GwgAEaKc,4300
 tico/experimental/quantization/ptq/examples/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3ouxARaxCK7bz9dn4blPQ,26
-tico/experimental/quantization/ptq/examples/compare_ppl.py,sha256=eVQn8-M24QkoCy_FCBQLSlUrnyqUDSkvUFpUpEdpMx4,8265
-tico/experimental/quantization/ptq/examples/debug_quant_outputs.py,sha256=Hpx_jj46WISwdVp33NrIadizVAzf1nFTXuAcHsKEQuQ,8179
+tico/experimental/quantization/ptq/examples/compare_ppl.py,sha256=2we7u4WgeK1NbYa31T2irUWE-RbTPUxRBdFtc9mPccY,8255
+tico/experimental/quantization/ptq/examples/debug_quant_outputs.py,sha256=GviYQa3MZ0-nKTKRaRsPXRI24VtWvDL3uOhOqsqxniY,8169
 tico/experimental/quantization/ptq/examples/quantize_linear.py,sha256=8zq-ZJDYgam0xQ-PbC6Xb1I7W1mv0Wi-b--IP2wwXtw,4539
 tico/experimental/quantization/ptq/examples/quantize_llama_attn.py,sha256=cVWUSSzaZWFp5QZkNkrlpHU3kXyP84QtnZbahVml_yQ,4329
 tico/experimental/quantization/ptq/examples/quantize_llama_decoder_layer.py,sha256=mBWrjkyEovYQsPC4Rrsri6Pm1rlFmDb3NiP0DQQhFyM,5751
-tico/experimental/quantization/ptq/examples/quantize_llama_mlp.py,sha256=N1qZQgt1S-xZrdv-PW7OfXEcv0gsO2q9faOF4aD-zKo,4147
-tico/experimental/quantization/ptq/examples/quantize_with_gptq.py,sha256=y-SK56j4wL-9j-0jtuOqQUq4CElZtGOETp-Tg4XivUI,10438
+tico/experimental/quantization/ptq/examples/quantize_llama_mlp.py,sha256=poP-TFmsP_Iy3K6NEu6f8UmHInaCX3wUSFZWhhqoUCQ,4137
+tico/experimental/quantization/ptq/examples/quantize_with_gptq.py,sha256=OqlaegX7ySR2PN6mSOZjcfKdxzrDe3gT_jBJg9HMzvM,10428
 tico/experimental/quantization/ptq/observers/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3ouxARaxCK7bz9dn4blPQ,26
 tico/experimental/quantization/ptq/observers/affine_base.py,sha256=e2Eba64nrxKQyE4F_WJ7WTSsk3xe6bkdGUKaoLFWGFw,4638
 tico/experimental/quantization/ptq/observers/base.py,sha256=Wons1MzpqK1mfcy-ppl-B2Dum0edXg2dWW2Lw3V18tw,3280
@@ -86,25 +86,25 @@ tico/experimental/quantization/ptq/utils/introspection.py,sha256=y2oGf7RoApMHJeX
 tico/experimental/quantization/ptq/utils/metrics.py,sha256=EW_FQmJrl9Y4esspZQ0GHfJ58RwuJUz0l8IfYq3NWY4,4461
 tico/experimental/quantization/ptq/utils/reduce_utils.py,sha256=3kWawLB91EcvvHlCrNqqfZF7tpgr22htBSA049mKw_4,973
 tico/experimental/quantization/ptq/wrappers/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3ouxARaxCK7bz9dn4blPQ,26
-tico/experimental/quantization/ptq/wrappers/ptq_wrapper.py,sha256=F9sK_DiRaXiGNHULcwIbs5EUtHz6ZJ7N4r5CWTTfhsM,2442
-tico/experimental/quantization/ptq/wrappers/quant_elementwise.py,sha256=LhEoobfvto6zKrBOKL4gmxfFFc31jHzyQV_zfps-iQM,3604
-tico/experimental/quantization/ptq/wrappers/quant_module_base.py,sha256=vkcDos_knGSS29rIZuEIWkAJLHrENbGz8nCH2-iara8,5969
+tico/experimental/quantization/ptq/wrappers/ptq_wrapper.py,sha256=5JgLwO6C3LMjVPmbZvl7cJcsJYKno91YPJZHl4HT1GQ,2432
+tico/experimental/quantization/ptq/wrappers/quant_elementwise.py,sha256=A9XYDPYqEnFmE-FGu1pS6V3LetqLr6xNXgo2TeoiF1c,3594
+tico/experimental/quantization/ptq/wrappers/quant_module_base.py,sha256=2F9NVvUP8Bs4Rs_XIPsQDUmLh-s4ZSgfNWwlXi-jLh4,5950
 tico/experimental/quantization/ptq/wrappers/registry.py,sha256=OVO5nev6J8Br9zsIX-Ut7ZgWzA9f_jk0Np9bGioXgQM,5171
 tico/experimental/quantization/ptq/wrappers/fairseq/__init__.py,sha256=Mc8FLd9DusyB_IT1vk1OYrRkngOYnYd05IvtA9ORVQc,160
 tico/experimental/quantization/ptq/wrappers/fairseq/decoder_export_single_step.py,sha256=d7ZieKiSbZ2ffkaLYMg2PJl1OyAxkKjB3OHKB4poxJs,9796
-tico/experimental/quantization/ptq/wrappers/fairseq/quant_decoder.py,sha256=CILYvxPhW7xLkroWW_hunQBGAYGexLqnPnO5xmMnK-E,17877
-tico/experimental/quantization/ptq/wrappers/fairseq/quant_decoder_layer.py,sha256=JT79shxOhDtRFgm8jrrN6HKvyVotiytLjMjAxX-Cztg,20416
-tico/experimental/quantization/ptq/wrappers/fairseq/quant_encoder.py,sha256=r9DPUAbL2KRJ8zpMJ39Y9n6Oe79nte-mFcdjG2qEP-w,13809
-tico/experimental/quantization/ptq/wrappers/fairseq/quant_encoder_layer.py,sha256=aGr80Ku75j2H-UZ0elEa0mOQEyaAs2YJ4WJCN0lonn0,6412
-tico/experimental/quantization/ptq/wrappers/fairseq/quant_mha.py,sha256=HsigmOLeacLXc46QNeFqwQ0DwKQhNrtWTKEtLJoqXoc,15562
+tico/experimental/quantization/ptq/wrappers/fairseq/quant_decoder.py,sha256=Lu6jBX1xUvpof3Q4YRh_FdwjKYbioKjvbtQQyBlA4bg,17865
+tico/experimental/quantization/ptq/wrappers/fairseq/quant_decoder_layer.py,sha256=r9-Ws-MhAgO7pXFAx_pFy42hT2syMS54cPdtsreVq_c,20406
+tico/experimental/quantization/ptq/wrappers/fairseq/quant_encoder.py,sha256=LtPJGk8avTWSNCTvzA5J88mcdT27Cg9HLmwCjrj8Lp0,13797
+tico/experimental/quantization/ptq/wrappers/fairseq/quant_encoder_layer.py,sha256=O-rkqWnTEzYG39bg9xNB3RH6T8I-cZyLgAwfWWBXFBk,6402
+tico/experimental/quantization/ptq/wrappers/fairseq/quant_mha.py,sha256=ek6lq5J0lAIpIlVQa4vri-7_Qe6Z38jE0A-lzS9UinE,15552
 tico/experimental/quantization/ptq/wrappers/llama/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3ouxARaxCK7bz9dn4blPQ,26
-tico/experimental/quantization/ptq/wrappers/llama/quant_attn.py,sha256=futw-XhAhErdaK2cZY8T3_xCxZbsj-l1dbsSbeunE_4,10403
-tico/experimental/quantization/ptq/wrappers/llama/quant_decoder_layer.py,sha256=ZImtfT2pyYyGJa0QCcHgCVootiWeflpRvLa4LisjZSY,7646
-tico/experimental/quantization/ptq/wrappers/llama/quant_mlp.py,sha256=uZMnrX66oZwxhKhcNbLXXeri-WxxRBiZnr15aBXJMm0,3562
+tico/experimental/quantization/ptq/wrappers/llama/quant_attn.py,sha256=B7SRKXclY1mdt31Yxbe8WLCxuCOTo8LVsgLQFVSrX7A,10393
+tico/experimental/quantization/ptq/wrappers/llama/quant_decoder_layer.py,sha256=hQRoccrsop4f0dj_y1sIRPcNOw2h4hBB8nZoBWpxB50,7666
+tico/experimental/quantization/ptq/wrappers/llama/quant_mlp.py,sha256=SVQsv4qwa48rLP_080YlxeEZlMkynvsFgkfLJ__F2uM,3552
 tico/experimental/quantization/ptq/wrappers/nn/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3ouxARaxCK7bz9dn4blPQ,26
-tico/experimental/quantization/ptq/wrappers/nn/quant_layernorm.py,sha256=G5Sgt-tXnzh0Rxyk-2honmZIfEQOZlRfOsoDBdSGmA4,6887
-tico/experimental/quantization/ptq/wrappers/nn/quant_linear.py,sha256=xW-VEPB7RJoslS3xLVCdhIuMjppknvpkZleRGK4JFVQ,2240
-tico/experimental/quantization/ptq/wrappers/nn/quant_silu.py,sha256=XnJDggkWUTfXC1-BLeAbcCUtp687XLIkIIbuQlqycDw,1864
+tico/experimental/quantization/ptq/wrappers/nn/quant_layernorm.py,sha256=-lhH2iR8kZZw89ejMg8i6MnFwE1RaP--xNzgiTFhWGY,6878
+tico/experimental/quantization/ptq/wrappers/nn/quant_linear.py,sha256=F2RN_lNStUwi6NQ9HOz0pdWerfMCK5JYt1NTTke49L4,2231
+tico/experimental/quantization/ptq/wrappers/nn/quant_silu.py,sha256=bIuMbAARoXhTdusPhchqj3DAL8dno8xO1mEMNhSF9OM,1854
 tico/interpreter/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3ouxARaxCK7bz9dn4blPQ,26
 tico/interpreter/infer.py,sha256=1ZFe3DVMR2mlwBosoedqoL0-CGN_01CKLgMgxuw62KA,4861
 tico/interpreter/interpreter.py,sha256=tGbluCbrehTCqBu8mtGDNzby_ieJ2ry8_RH_eC0CQxk,3828
@@ -238,7 +238,7 @@ tico/serialize/operators/utils.py,sha256=lXGpEJW1h8U_-gfc6EWjvvSiq3yJ9P-v1v3EMRT
 tico/serialize/operators/adapters/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3ouxARaxCK7bz9dn4blPQ,26
 tico/serialize/operators/adapters/llama_rmsnorm.py,sha256=6t3dhfNpR03eIjsmhymF2JKd6lCf7PvInqMf77c_BOE,1139
 tico/utils/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3ouxARaxCK7bz9dn4blPQ,26
-tico/utils/convert.py,sha256=10YufXpuqHz274ACUb1_F5594uClUFhBEh8SY6gYp7w,13809
+tico/utils/convert.py,sha256=ylEnz_1jrvWTVEItRw9xCwaMhnbjp6TSrePbXiiccbc,13857
 tico/utils/define.py,sha256=Ypgp7YffM4pgPl4Zh6TmogSn1OxGBMRw_e09qYGflZk,1467
 tico/utils/diff_graph.py,sha256=_eDGGPDPYQD4b--MXX0DLoVgSt_wLfNPt47UlolLLR4,5272
 tico/utils/dtype.py,sha256=L5Qb7qgbt0eQ5frUTvHYrRtTJb1dg4-JNEopcxCNg1U,1389
@@ -262,9 +262,9 @@ tico/utils/mx/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3ouxARaxCK7bz9dn4blPQ,26
 tico/utils/mx/elemwise_ops.py,sha256=V6glyAHsVR1joqpsgnNytatCD_ew92xNWZ19UFDoMTA,10281
 tico/utils/mx/formats.py,sha256=uzNWyu-1onUlwQfX5cZ6fZSUfHMRqorper7_T1k3jfk,3404
 tico/utils/mx/mx_ops.py,sha256=RcfUTYVi-wilGB2sC35OeARdwDqnixv7dG5iyZ-fQT8,8555
-tico-0.1.0.dev251013.dist-info/LICENSE,sha256=kp4JLII7bzRhPb0CPD5XTDZMh22BQ7h3k3B7t8TiSbw,12644
-tico-0.1.0.dev251013.dist-info/METADATA,sha256=gxBnCYAGBeSVsMkXVBKj0fDqApxxDeM5lIDdM_0ziBk,8450
-tico-0.1.0.dev251013.dist-info/WHEEL,sha256=G16H4A3IeoQmnOrYV4ueZGKSjhipXx8zc8nu9FGlvMA,92
-tico-0.1.0.dev251013.dist-info/entry_points.txt,sha256=kBKYSS_IYrSXmUYevmmepqIVPScq5vF8ulQRu3I_Zf0,59
-tico-0.1.0.dev251013.dist-info/top_level.txt,sha256=oqs7UPoNSKZEwqsX8B-KAWdQwfAa7i60pbxW_Jk7P3w,5
-tico-0.1.0.dev251013.dist-info/RECORD,,
+tico-0.1.0.dev251014.dist-info/LICENSE,sha256=kp4JLII7bzRhPb0CPD5XTDZMh22BQ7h3k3B7t8TiSbw,12644
+tico-0.1.0.dev251014.dist-info/METADATA,sha256=CrvKqaj02su0UWPLp4FUbHTBRBjna1DZeM6aKopnlhQ,8455
+tico-0.1.0.dev251014.dist-info/WHEEL,sha256=G16H4A3IeoQmnOrYV4ueZGKSjhipXx8zc8nu9FGlvMA,92
+tico-0.1.0.dev251014.dist-info/entry_points.txt,sha256=kBKYSS_IYrSXmUYevmmepqIVPScq5vF8ulQRu3I_Zf0,59
+tico-0.1.0.dev251014.dist-info/top_level.txt,sha256=oqs7UPoNSKZEwqsX8B-KAWdQwfAa7i60pbxW_Jk7P3w,5
+tico-0.1.0.dev251014.dist-info/RECORD,,