tico 0.1.0.dev250703__py3-none-any.whl → 0.1.0.dev250707__py3-none-any.whl

tico/__init__.py

@@ -21,7 +21,7 @@ from tico.config import CompileConfigV1, get_default_config
 from tico.utils.convert import convert, convert_from_exported_program, convert_from_pt2
 
 # THIS LINE IS AUTOMATICALLY GENERATED BY setup.py
-__version__ = "0.1.0.dev250703"
+__version__ = "0.1.0.dev250707"
 
 MINIMUM_SUPPORTED_VERSION = "2.5.0"
 SECURE_TORCH_VERSION = "2.6.0"

tico/experimental/quantization/passes/insert_quantize_on_dtype_mismatch.py

@@ -31,6 +31,7 @@ from tico.utils.utils import quant_min_max, set_new_meta_val
 from tico.utils.validate_args_kwargs import (
     AddTensorArgs,
     BmmArgs,
+    CatArgs,
     LinearArgs,
     MulTensorArgs,
     PermuteArgs,
@@ -79,7 +80,7 @@ def _u8_to_i16(qparam: QuantParam) -> QuantParam:
     max_ = u8_scale * (255 - u8_zerop)
     min_ = u8_scale * (-u8_zerop)
 
-    abs_max = abs(max([max_, min_], key=abs))
+    abs_max = max(abs(max_), abs(min_))
     s16_scale = abs_max / 32767
     s16_zerop = 0
 
@@ -282,6 +283,35 @@ class InsertQuantizeOnDtypeMismatch(PassBase):
                 else:
                     raise NotYetSupportedError("Unsupported dtype")
 
+            elif node.target == torch.ops.aten.cat.default:
+                cat_args = CatArgs(*node.args, **node.kwargs)
+                tensors = cat_args.tensors
+
+                if any(QPARAM_KEY not in x.meta for x in tensors):
+                    continue
+
+                if QPARAM_KEY not in node.meta:
+                    continue
+
+                assert len(tensors) > 0
+                in_dtype = qparam_dtype(tensors[0])
+                if in_dtype == qparam_dtype(node):
+                    continue
+
+                if any(qparam_dtype(x) != in_dtype for x in tensors):
+                    continue
+
+                if in_dtype == "int16" and qparam_dtype(node) == "uint8":
+                    quantize = _insert_quantize_op_after(node)
+
+                    quantize.meta[QPARAM_KEY] = copy.deepcopy(node.meta[QPARAM_KEY])
+                    node.meta[QPARAM_KEY] = _u8_to_i16(node.meta[QPARAM_KEY])
+                    logger.debug(
+                        f"quantize_per_tensor.default is inserted after {node.name}."
+                    )
+                else:
+                    raise NotYetSupportedError("Unsupported dtype")
+
             elif node.target == torch.ops.aten.bmm.default:
                 bmm_args = BmmArgs(*node.args, **node.kwargs)
                 x = bmm_args.input

{tico-0.1.0.dev250703.dist-info → tico-0.1.0.dev250707.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: tico
-Version: 0.1.0.dev250703
+Version: 0.1.0.dev250707
 Summary: Convert exported Torch module to circle
 Home-page: UNKNOWN
 License: UNKNOWN
@@ -43,10 +43,7 @@ designed for optimized on-device neural network inference.
 0. Prerequisites
 
 - Python 3.10
-- [one-compiler nightly](https://github.com/Samsung/TICO/issues/2)
-  - This project depends on [ONE](https://github.com/Samsung/ONE) Compiler, and it uses 
-  nightly features that are not yet available in the official release. Until one-compiler 1.30.0
-  is released, you must use a prebuilt nighlty version of ONE Compiler.
+- [one-compiler 1.30.0](https://github.com/Samsung/ONE/releases/tag/1.30.0)
 
 We highly recommend to use a virtual env, e.g., conda.
 
@@ -69,7 +66,7 @@ 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, nightly
+  - Available <torch version>: 2.5, 2.6, 2.7, nightly
 
 4. Now you can convert a torch module to a `.circle`.
 
@@ -290,17 +287,10 @@ By default, `./ccex test` runs all modules with the `circle-interpreter` engine.
  You can override this and run tests using the `onert` runtime instead.
 
 
-##### 0. Installing ONERT Nightly
-
-Some ONERT features are only available in the nightly build until the next official release.
- To install the ONERT wheel from the issue comment:
-
-1. Download the `.whl` file linked 
-in [the relevant Github issue comment](https://github.com/Samsung/TICO/issues/2#issuecomment-2841487306).
-2. Install it with pip, for example:
+##### 0. Install ONERT
 
 ```bash
-pip install /path/to/onert_nightly.whl
+pip install onert
 ```
 
 ##### 1. Command-Line Flag

{tico-0.1.0.dev250703.dist-info → tico-0.1.0.dev250707.dist-info}/RECORD

@@ -1,4 +1,4 @@
-tico/__init__.py,sha256=P6Y7D-gIsRLMIUA9w5OkdqONBxv5lRR70wvkWFHh1wY,1743
+tico/__init__.py,sha256=G29NbaZNUeg1aAnblegJHoEefDV7Icu2V7UzEeaZf1A,1743
 tico/pt2_to_circle.py,sha256=gu3MD4Iqc0zMZcCZ2IT8oGbyj21CTSbT3Rgd9s2B_9A,2767
 tico/config/__init__.py,sha256=xZzCXjZ84qE-CsBi-dfaL05bqpQ3stKKfTXhnrJRyVs,142
 tico/config/base.py,sha256=anwOiJFkUxUi7Cef573JgQcjk6S-FSi6O_TLjYASW-g,1244
@@ -51,7 +51,7 @@ tico/experimental/quantization/evaluation/executor/circle_executor.py,sha256=eCC
 tico/experimental/quantization/evaluation/executor/triv24_executor.py,sha256=sUoXl6oOO2arAKaNjOBg7HiQja145_Jv6qgY7XtR7A8,5159
 tico/experimental/quantization/passes/__init__.py,sha256=IO6FP_xYbGy0dW0HL26GXD3ouxARaxCK7bz9dn4blPQ,26
 tico/experimental/quantization/passes/fold_quant_ops.py,sha256=iaBMyO49CwVkhebMz3rjkHWfWE2LhwH6fORe7n4S6XQ,7040
-tico/experimental/quantization/passes/insert_quantize_on_dtype_mismatch.py,sha256=lQeN6VJNYQyDeucB4KpyyWvIiuhGRq7wjIeaCKdM7ck,15462
+tico/experimental/quantization/passes/insert_quantize_on_dtype_mismatch.py,sha256=FfqTlGANcG1V64zw0MFcIxL9WafuxPINuzWohGdsYCg,16617
 tico/experimental/quantization/passes/propagate_qparam_backward.py,sha256=TGtyW0Z2qOTgVIasBdGRgbwH31YYd6ek7OvLTmCV614,3118
 tico/experimental/quantization/passes/propagate_qparam_forward.py,sha256=RhUHGCR2RpBO5KYkQ7Z8U5u7HEwDq2wdKHLKAJCi-5c,5138
 tico/experimental/quantization/passes/quantize_bias.py,sha256=ZQ3rETYStpW28JUbODRixbq5sDEOiIOB_qWA-Jzuu-Y,4337
@@ -200,9 +200,9 @@ 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.dev250703.dist-info/LICENSE,sha256=kp4JLII7bzRhPb0CPD5XTDZMh22BQ7h3k3B7t8TiSbw,12644
-tico-0.1.0.dev250703.dist-info/METADATA,sha256=b1B8214FjIYVG6Z4kL5NDM0EN-jksdYbAtiViQ-hGtU,8846
-tico-0.1.0.dev250703.dist-info/WHEEL,sha256=G16H4A3IeoQmnOrYV4ueZGKSjhipXx8zc8nu9FGlvMA,92
-tico-0.1.0.dev250703.dist-info/entry_points.txt,sha256=kBKYSS_IYrSXmUYevmmepqIVPScq5vF8ulQRu3I_Zf0,59
-tico-0.1.0.dev250703.dist-info/top_level.txt,sha256=oqs7UPoNSKZEwqsX8B-KAWdQwfAa7i60pbxW_Jk7P3w,5
-tico-0.1.0.dev250703.dist-info/RECORD,,
+tico-0.1.0.dev250707.dist-info/LICENSE,sha256=kp4JLII7bzRhPb0CPD5XTDZMh22BQ7h3k3B7t8TiSbw,12644
+tico-0.1.0.dev250707.dist-info/METADATA,sha256=DNpanGd-UTVh99jSdqMsSlyt6Z9CkqgkHi5uzLlKWSU,8244
+tico-0.1.0.dev250707.dist-info/WHEEL,sha256=G16H4A3IeoQmnOrYV4ueZGKSjhipXx8zc8nu9FGlvMA,92
+tico-0.1.0.dev250707.dist-info/entry_points.txt,sha256=kBKYSS_IYrSXmUYevmmepqIVPScq5vF8ulQRu3I_Zf0,59
+tico-0.1.0.dev250707.dist-info/top_level.txt,sha256=oqs7UPoNSKZEwqsX8B-KAWdQwfAa7i60pbxW_Jk7P3w,5
+tico-0.1.0.dev250707.dist-info/RECORD,,