tico 0.1.0.dev251102__py3-none-any.whl → 0.1.0.dev251106__py3-none-any.whl

tico/__init__.py

@@ -29,7 +29,7 @@ __all__ = [
 ]
 
 # THIS LINE IS AUTOMATICALLY GENERATED BY setup.py
-__version__ = "0.1.0.dev251102"
+__version__ = "0.1.0.dev251106"
 
 MINIMUM_SUPPORTED_VERSION = "2.5.0"
 SECURE_TORCH_VERSION = "2.6.0"

tico/quantization/algorithm/gptq/gptq.py

@@ -36,6 +36,11 @@ class GPTQ:
         self.layer = layer
         self.dev = self.layer.weight.device
         W = layer.weight.data.clone()
+        if isinstance(self.layer, nn.Conv2d):
+            W = W.flatten(1)
+
+        if isinstance(self.layer, nn.Conv1d):
+            W = W.t()
         self.rows = W.shape[0]
         self.columns = W.shape[1]
         self.H: Optional[torch.Tensor] = torch.zeros(
@@ -48,10 +53,22 @@ class GPTQ:
         if len(inp.shape) == 2:
             inp = inp.unsqueeze(0)
         tmp = inp.shape[0]
-        if isinstance(self.layer, nn.Linear):
-            if len(inp.shape) == 3:
+        if isinstance(self.layer, nn.Linear) or isinstance(self.layer, nn.Conv1d):
+            if len(inp.shape) > 2:
                 inp = inp.reshape((-1, inp.shape[-1]))
             inp = inp.t()
+        if isinstance(self.layer, nn.Conv2d):
+            unfold = nn.Unfold(
+                self.layer.kernel_size,
+                dilation=self.layer.dilation,
+                padding=self.layer.padding,
+                stride=self.layer.stride,
+            )
+
+            inp = unfold(inp)
+            inp = inp.permute([1, 0, 2])
+            inp = inp.flatten(1)
+
         self.H *= self.nsamples / (self.nsamples + tmp)
         self.nsamples += tmp
         inp = math.sqrt(2 / self.nsamples) * inp.float()
@@ -67,6 +84,10 @@ class GPTQ:
         verbose=False,
     ):
         W = self.layer.weight.data.clone()
+        if isinstance(self.layer, nn.Conv2d):
+            W = W.flatten(1)
+        if isinstance(self.layer, nn.Conv1d):
+            W = W.t()
         W = W.float()
         tick = time.time()
         if not self.quantizer.ready():

tico/quantization/wrapq/wrappers/nn/quant_silu.py

@@ -19,14 +19,15 @@ import torch.nn as nn
 
 from tico.quantization.config.ptq import PTQConfig
 from tico.quantization.wrapq.wrappers.quant_module_base import QuantModuleBase
-from tico.quantization.wrapq.wrappers.registry import register
+from tico.quantization.wrapq.wrappers.registry import try_register
 
 
-@register(nn.SiLU)
+@try_register("torch.nn.SiLU", "transformers.activations.SiLUActivation")
 class QuantSiLU(QuantModuleBase):
     """
-    QuantSiLU — drop-in replacement for nn.SiLU that quantizes
-    both intermediate tensors:
+    QuantSiLU — drop-in quantized implementation of the SiLU operation.
+
+    This module quantizes both intermediate tensors:
         • s  = sigmoid(x)   (logistic)
         • y  = x * s        (mul)
     """

tico/quantization/wrapq/wrappers/registry.py

@@ -23,14 +23,17 @@ _WRAPPERS: Dict[Type[nn.Module], Type[QuantModuleBase]] = {}
 _IMPORT_ONCE = False
 _CORE_MODULES = (
     "tico.quantization.wrapq.wrappers.quant_elementwise",
+    ## nn ##
     "tico.quantization.wrapq.wrappers.nn.quant_layernorm",
     "tico.quantization.wrapq.wrappers.nn.quant_linear",
+    # This includes not only `nn.SiLU` but also `SiLUActivation` from transformers
+    # as they are same operation.
     "tico.quantization.wrapq.wrappers.nn.quant_silu",
-    # llama
+    ## llama ##
     "tico.quantization.wrapq.wrappers.llama.quant_attn",
     "tico.quantization.wrapq.wrappers.llama.quant_decoder_layer",
     "tico.quantization.wrapq.wrappers.llama.quant_mlp",
-    # fairseq
+    ## fairseq ##
     "tico.quantization.wrapq.wrappers.fairseq.quant_decoder_layer",
     "tico.quantization.wrapq.wrappers.fairseq.quant_encoder",
     "tico.quantization.wrapq.wrappers.fairseq.quant_encoder_layer",

{tico-0.1.0.dev251102.dist-info → tico-0.1.0.dev251106.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: tico
-Version: 0.1.0.dev251102
+Version: 0.1.0.dev251106
 Summary: Convert exported Torch module to circle
 Home-page: UNKNOWN
 License: UNKNOWN

{tico-0.1.0.dev251102.dist-info → tico-0.1.0.dev251106.dist-info}/RECORD

@@ -1,4 +1,4 @@
-tico/__init__.py,sha256=y_g2dpUeGlGMQ9CHvwpWnytUxBZg1Z4De9VFXdD4bos,1883
+tico/__init__.py,sha256=zFizlBKVmXPeQuRPHlpsmWOePEoPowkPdrlRwoxPs4k,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
@@ -52,7 +52,7 @@ tico/quantization/quantizer.py,sha256=FYNiqUqoH9vz1bda0I6yuKqJi2KdIfLEBd4EgeC-_t
 tico/quantization/quantizer_registry.py,sha256=MxVE1_hj1p8FjdAqkLzUhdez3Cqc-V25k6XKOcTkei0,2414
 tico/quantization/algorithm/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3ouxARaxCK7bz9dn4blPQ,26
 tico/quantization/algorithm/gptq/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3ouxARaxCK7bz9dn4blPQ,26
-tico/quantization/algorithm/gptq/gptq.py,sha256=HkuKv_UWs0xEdbj7zEP-65QPEtI_varmvAORFstyTic,5542
+tico/quantization/algorithm/gptq/gptq.py,sha256=x7wM9_OgOrcs6WmkVCDLn2bF7YuUAR_k6vLG2l593sk,6235
 tico/quantization/algorithm/gptq/quant.py,sha256=Rl4wAOCmlE0U09BtNCDbccaSNohRHCNLwFi3zCqZfNo,5127
 tico/quantization/algorithm/gptq/quantizer.py,sha256=OvR9sHgosGYofwYcDhye84FBl55cNY7-UlfBt9gXbDY,11734
 tico/quantization/algorithm/gptq/utils.py,sha256=leGKayf-xbSjVwwAGTA5RsxUKrhDiklOQdlsLifjdrs,1811
@@ -131,7 +131,7 @@ tico/quantization/wrapq/wrappers/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3ouxAR
 tico/quantization/wrapq/wrappers/ptq_wrapper.py,sha256=6zcVZ-vVhPCvFHQw6UlN7iizElrIHNkpAraeMaA0DDU,2388
 tico/quantization/wrapq/wrappers/quant_elementwise.py,sha256=trchhUknmZTcoCwVA62uzBP_mWuCjjuZjF0jb7TZpfA,3550
 tico/quantization/wrapq/wrappers/quant_module_base.py,sha256=SgyUlFYxDx39CAvcN2q4lsTedbEVPmetIigrllmvvD4,5915
-tico/quantization/wrapq/wrappers/registry.py,sha256=1rH28O7aWrp-uIFL7exa6rfdyEHeupzXuMZhNTW2i2k,5030
+tico/quantization/wrapq/wrappers/registry.py,sha256=QJcOD9gEGB_DJowdTTqemcRDcYxQa4tHv2CDFgZDnA0,5168
 tico/quantization/wrapq/wrappers/fairseq/__init__.py,sha256=K4R7rbxHosx9LBLk2WKlL8gFuZTYTws41TW47AsSUPM,149
 tico/quantization/wrapq/wrappers/fairseq/decoder_export_single_step.py,sha256=d7ZieKiSbZ2ffkaLYMg2PJl1OyAxkKjB3OHKB4poxJs,9796
 tico/quantization/wrapq/wrappers/fairseq/quant_decoder.py,sha256=JTCUDNEHYU5iOcbC_2mpuhvEoZqzTNIW3gPUZE1J7FE,17810
@@ -146,7 +146,7 @@ tico/quantization/wrapq/wrappers/llama/quant_mlp.py,sha256=I0EUJPnBOvaTnjT1Jk4N2
 tico/quantization/wrapq/wrappers/nn/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3ouxARaxCK7bz9dn4blPQ,26
 tico/quantization/wrapq/wrappers/nn/quant_layernorm.py,sha256=UoWWQaDqBY_bAeWRRsNl19LO331KQQLpZP9ACE-HyiU,6823
 tico/quantization/wrapq/wrappers/nn/quant_linear.py,sha256=y3exJX_Og8HIi0VdpvX4M9m8Voq0e0ndiX8G6DZflT8,2165
-tico/quantization/wrapq/wrappers/nn/quant_silu.py,sha256=6inKWfcVTlXFsnTX_6DdIChME3x0jL_urGbONjydMqw,1810
+tico/quantization/wrapq/wrappers/nn/quant_silu.py,sha256=jRbM2lCFjqAqQj3Gur4eiHs1eCoNtjejMd16VBhNZt8,1901
 tico/serialize/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3ouxARaxCK7bz9dn4blPQ,26
 tico/serialize/circle_graph.py,sha256=qvyul_HULoz7B_6RFKQ8s9RjEvMgPq-ynMVkZe8aqE4,12034
 tico/serialize/circle_mapping.py,sha256=c__AIHPi23lPugNJFolgMAKrw8j7gEeMaUQ1LAMSFnY,8542
@@ -263,9 +263,9 @@ 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.dev251102.dist-info/LICENSE,sha256=kp4JLII7bzRhPb0CPD5XTDZMh22BQ7h3k3B7t8TiSbw,12644
-tico-0.1.0.dev251102.dist-info/METADATA,sha256=JsgJk7a4VRJL_Mt9Z3_DBpgqsDWMtXTHgNRK2Bi-wTA,9730
-tico-0.1.0.dev251102.dist-info/WHEEL,sha256=G16H4A3IeoQmnOrYV4ueZGKSjhipXx8zc8nu9FGlvMA,92
-tico-0.1.0.dev251102.dist-info/entry_points.txt,sha256=kBKYSS_IYrSXmUYevmmepqIVPScq5vF8ulQRu3I_Zf0,59
-tico-0.1.0.dev251102.dist-info/top_level.txt,sha256=oqs7UPoNSKZEwqsX8B-KAWdQwfAa7i60pbxW_Jk7P3w,5
-tico-0.1.0.dev251102.dist-info/RECORD,,
+tico-0.1.0.dev251106.dist-info/LICENSE,sha256=kp4JLII7bzRhPb0CPD5XTDZMh22BQ7h3k3B7t8TiSbw,12644
+tico-0.1.0.dev251106.dist-info/METADATA,sha256=cH9ZYH9ysDRREwJcSjMV1VkLFpNjoOSxPBChSzab-A0,9730
+tico-0.1.0.dev251106.dist-info/WHEEL,sha256=G16H4A3IeoQmnOrYV4ueZGKSjhipXx8zc8nu9FGlvMA,92
+tico-0.1.0.dev251106.dist-info/entry_points.txt,sha256=kBKYSS_IYrSXmUYevmmepqIVPScq5vF8ulQRu3I_Zf0,59
+tico-0.1.0.dev251106.dist-info/top_level.txt,sha256=oqs7UPoNSKZEwqsX8B-KAWdQwfAa7i60pbxW_Jk7P3w,5
+tico-0.1.0.dev251106.dist-info/RECORD,,