Update aimet_torch.v2 apis to default in the docs (#3672)

Update v2 APIs to default in docs Signed-off-by: Hitarth Mehta <[email protected]>
quic · Dec 19, 2024 · 5494006 · 5494006
1 parent 946e769
commit 5494006
Show file tree

Hide file tree

Showing 31 changed files with 259 additions and 213 deletions.
diff --git a/Docs/apiref/torch/adaround.rst b/Docs/apiref/torch/adaround.rst
@@ -7,10 +7,6 @@ aimet_torch.adaround
 ..
   # start-after
 
-.. note::
-
-    This module is also available in the experimental :mod:`aimet_torch.v2` namespace with the same top-level API.
-
 **Top level APIs**
 
 .. autofunction:: aimet_torch.adaround.adaround_weight.Adaround.apply_adaround

diff --git a/Docs/apiref/torch/index.rst b/Docs/apiref/torch/index.rst
@@ -9,14 +9,19 @@ aimet_torch API
 
     aimet_torch.quantsim <quantsim>
     aimet_torch.adaround <adaround>
+    aimet_torch.nn <nn>
+    aimet_torch.quantization <quantization>
     aimet_torch.seq_mse <seq_mse>
+    aimet_torch.quantsim.config_utils <lpbq>
+    aimet_torch.manual_mixed_precision <mmp>
     aimet_torch.batch_norm_fold <bnf>
     aimet_torch.cross_layer_equalization <cle>
     aimet_torch.model_preparer <model_preparer>
-    aimet_torch.mixed_precision <amp>
+    aimet_torch.auto_mixed_precision <amp>
     aimet_torch.quant_analyzer <quant_analyzer>
     aimet_torch.autoquant <autoquant>
     aimet_torch.bn_reestimation <bn>
+    aimet_torch.visualization_tools <interactive_visualization>
     aimet_torch.layer_output_utils <layer_output_generation>
     aimet_torch.peft <peft_lora>
     aimet_torch.compress <compress>
@@ -25,88 +30,49 @@ aimet_torch
 ===========
 
 .. important::
+   :mod:`aimet_torch` package is upgraded to :mod:`aimet_torch.v2` with more
+   flexible, extensible, and PyTorch-friendly user interface!
 
-   :mod:`aimet_torch` package is planned to be upgraded to :mod:`aimet_torch.v2` with more flexible, extensible, and PyTorch-friendly user interface! In a future release, the core APIs of :mod:`aimet_torch` will be fully replaced with the equivalents in :mod:`aimet_torch.v2`.
-
-AIMET quantization for PyTorch models provides the following functionality.
+   aimet_torch 2 is fully backward compatible with all the public APIs of aimet_torch 1.x.,
+   please see :doc:`Migrate to aimet_torch 2 <../../quantsim/torch/migration_guide>`.
 
 - :ref:`aimet_torch.quantsim <apiref-torch-quantsim>`
+- :ref:`aimet_torch.nn <apiref-torch-nn>`
+- :ref:`aimet_torch.quantization <apiref-torch-quantization>`
 - :ref:`aimet_torch.adaround <apiref-torch-adaround>`
 - :ref:`aimet_torch.seq_mse <apiref-torch-seq-mse>`
+- :ref:`aimet_torch.quantsim.config_utils <apiref-torch-lpbq>`
 - :ref:`aimet_torch.batch_norm_fold <apiref-torch-bnf>`
 - :ref:`aimet_torch.cross_layer_equalization <apiref-torch-cle>`
 - :ref:`aimet_torch.model_preparer <apiref-torch-model-preparer>`
-- :ref:`aimet_torch.mixed_precision <api-torch-amp>`
+- :ref:`aimet_torch.mixed_precision <api-torch-mp>`
 - :ref:`aimet_torch.quant_analyzer <apiref-torch-quant-analyzer>`
 - :ref:`aimet_torch.autoquant <apiref-torch-autoquant>`
 - :ref:`aimet_torch.bn_reestimation <apiref-torch-bn>`
+- :ref:`aimet_torch.visualization_tools <api-torch-interactive-visualization>`
 - :ref:`aimet_torch.layer_output_utils <apiref-torch-layer-output-generation>`
 - :ref:`aimet_torch.peft <apiref-torch-peft-lora>`
 - :ref:`aimet_torch.compress <apiref-torch-compress>`
 
-aimet_torch.v2
+aimet_torch.v1
 ==============
 
-Introducing :mod:`aimet_torch.v2`, a future version of :mod:`aimet_torch` with more powerful
-quantization features and PyTorch-friendly user interface!
-
-What's New
-----------
-
-These are some of the powerful new features and interfaces supported in :mod:`aimet_torch.v2`
-
-- Blockwise quantization (BQ)
-- Low power blockwise quantization (LPBQ)
-- Dispatching custom quantized kernels
-
-Backwards Compatibility
------------------------
-
-Good news! :mod:`aimet_torch.v2` is carefully designed to be fully backwards-compatible with all
-previous public APIs of :mod:`aimet_torch`. All you need is drop-in replacement of import statements
-from :mod:`aimet_torch` to :mod:`aimet_torch.v2` as below!
-
-.. code-block:: diff
-
-   -from aimet_torch.quantsim import QuantizationSimModel
-   +from aimet_torch.v2.quantsim import QuantizationSimModel
-
-   -from aimet_torch.adaround.adaround_weight import Adaround, AdaroundParameters
-   +from aimet_torch.v2.adaround import Adaround, AdaroundParameters
-
-   -from aimet_torch.seq_mse import apply_seq_mse
-   +from aimet_torch.v2.seq_mse import apply_seq_mse
-
-   -from aimet_torch.quant_analyzer import QuantAnalyzer
-   +from aimet_torch.v2.quant_analyzer import QuantAnalyzer
-
-All the other APIs that didn't changed in or are orthogonal with :mod:`aimet_torch.v2` will be
-still accessible via :mod:`aimet_torch` namespace as before.
+If you still prefer to use aimet_torch 1.x, your imports should originate from the :mod:`aimet_torch.v1`
+namespace.
 
 .. toctree::
     :hidden:
 
-    aimet_torch.v2 migration guide <v2/migration_guide>
-    aimet_torch.v2.nn <v2/nn>
-    aimet_torch.v2.quantization <v2/quantization>
-    aimet_torch.v2.adaround <v2/adaround>
-    aimet_torch.v2.seq_mse <v2/seq_mse>
-    aimet_torch.v2.quantsim.config_utils <v2/lpbq>
-    aimet_torch.v2.mixed_precision <v2/mmp>
-    aimet_torch.v2.quant_analyzer <v2/quant_analyzer>
-    aimet_torch.v2.visualization_tools <v2/interactive_visualization>
-
-For more detailed information about how to migrate to :mod:`aimet_torch.v2`,
-see :ref:`aimet_torch.v2 migration guide <torch-migration-guide>`
-
-AIMET core APIs for PyTorch framework.
-
-- :ref:`aimet_torch.v2.quantsim <apiref-torch-v2-quantsim>`
-- :ref:`aimet_torch.v2.nn <apiref-torch-nn>`
-- :ref:`aimet_torch.v2.quantization <apiref-torch-quantization>`
-- :ref:`aimet_torch.v2.adaround <apiref-torch-v2-adaround>`
-- :ref:`aimet_torch.v2.seq_mse <apiref-torch-v2-seq-mse>`
-- :ref:`aimet_torch.v2.quantsim.config_utils <apiref-torch-v2-lpbq>`
-- :ref:`aimet_torch.v2.mixed_precision <api-torch-v2-mmp>`
-- :ref:`aimet_torch.v2.quant_analyzer <apiref-torch-v2-quant-analyzer>`
-- :ref:`aimet_torch.v2.visualization_tools <api-torch-v2-interactive-visualization>`
+    aimet_torch.v1.quantsim <v1/quantsim>
+    aimet_torch.v1.adaround <v1/adaround>
+    aimet_torch.v1.seq_mse <v1/seq_mse>
+    aimet_torch.v1.quant_analyzer <v1/quant_analyzer>
+    aimet_torch.v1.autoquant <v1/autoquant>
+    aimet_torch.v1.amp <v1/amp>
+
+- :ref:`aimet_torch.v1.quantsim <apiref-torch-v1-quantsim>`
+- :ref:`aimet_torch.v1.adaround <apiref-torch-v1-adaround>`
+- :ref:`aimet_torch.v1.seq_mse <apiref-torch-v1-seq-mse>`
+- :ref:`aimet_torch.v1.quant_analyzer <apiref-torch-v1-quant-analyzer>`
+- :ref:`aimet_torch.v1.autoquant <apiref-torch-v1-autoquant>`
+- :ref:`aimet_torch.v1.amp <apiref-torch-v1-amp>`
diff --git a/...ef/torch/v2/interactive_visualization.rst → ...piref/torch/interactive_visualization.rst b/...ef/torch/v2/interactive_visualization.rst → ...piref/torch/interactive_visualization.rst
@@ -1,13 +1,12 @@
-.. _api-torch-v2-interactive-visualization:
+.. _api-torch-interactive-visualization:
 
-##################################
-aimet_torch.v2.visualization_tools
-##################################
+###############################
+aimet_torch.visualization_tools
+###############################
 
 ..
   # start-after
 
-
 .. warning::
     This feature is under heavy development and API changes may occur without notice in future versions.
 

diff --git a/Docs/apiref/torch/v2/lpbq.rst → Docs/apiref/torch/lpbq.rst b/Docs/apiref/torch/v2/lpbq.rst → Docs/apiref/torch/lpbq.rst
@@ -1,8 +1,8 @@
-.. _apiref-torch-v2-lpbq:
+.. _apiref-torch-lpbq:
 
-####################################
-aimet_torch.v2.quantsim.config_utils
-####################################
+#################################
+aimet_torch.quantsim.config_utils
+#################################
 
 ..
   # start-after

diff --git a/Docs/apiref/torch/mp.rst b/Docs/apiref/torch/mp.rst
@@ -0,0 +1,46 @@
+.. _api-torch-mp:
+
+===========================
+aimet_torch.mixed_precision
+===========================
+
+..
+  # start-after mmp
+
+**Top-level API for Manual mixed precision**
+
+.. autoclass:: aimet_torch.v2.mixed_precision.MixedPrecisionConfigurator
+    :members:
+    :member-order: bysource
+
+..
+  # end-before mmp
+
+..
+  # start-after amp
+
+**Top-level API for Automatic mixed precision**
+
+.. autofunction:: aimet_torch.mixed_precision.choose_mixed_precision
+
+.. note::
+    To enable phase-3 set the attribute GreedyMixedPrecisionAlgo.ENABLE_CONVERT_OP_REDUCTION = True
+
+Currently only two candidates are supported - ((8,int), (8,int)) & ((16,int), (8,int))
+
+**Quantizer Groups definition**
+
+.. autoclass:: aimet_torch.amp.quantizer_groups.QuantizerGroup
+   :members:
+
+**CallbackFunc Definition**
+
+.. autoclass:: aimet_common.defs.CallbackFunc
+   :members:
+
+.. autoclass:: aimet_torch.amp.mixed_precision_algo.EvalCallbackFactory
+   :members:
+
+..
+  # end-before amp
+
diff --git a/Docs/apiref/torch/v2/nn.rst → Docs/apiref/torch/nn.rst b/Docs/apiref/torch/v2/nn.rst → Docs/apiref/torch/nn.rst
@@ -1,8 +1,8 @@
 .. _apiref-torch-nn:
 
-#################
-aimet_torch.v2.nn
-#################
+##############
+aimet_torch.nn
+##############
 
 .. currentmodule:: aimet_torch.v2.nn
 

diff --git a/Docs/apiref/torch/peft_lora.rst b/Docs/apiref/torch/peft_lora.rst
@@ -6,7 +6,7 @@ aimet_torch.peft
 
 This document provides steps for integrating LoRA adapters with AIMET Quantization flow. LoRA adapters
 are used to enhance the efficiency of fine-tuning large models with reduced memory usage. We will use
-:ref:`PEFT <https://huggingface.co/docs/peft/main/en/package_reference/peft_model>` library
+`PEFT <https://huggingface.co/docs/peft/main/en/package_reference/peft_model>`_ library
 from HuggingFace to instantiate our model and add adapters to it.
 
 By integrating adapters with AIMET quantization, we can perform similar functionalities as present in PEFT,

diff --git a/Docs/apiref/torch/quant_analyzer.rst b/Docs/apiref/torch/quant_analyzer.rst
@@ -31,7 +31,7 @@ QuantAnalyzer utility of your interest and pass the same object to it.
 
 .. automethod:: aimet_torch.quant_analyzer.QuantAnalyzer.export_per_layer_stats_histogram
 
-.. automethod:: aimet_torch.v1.quant_analyzer.QuantAnalyzer.export_per_layer_mse_loss
+.. automethod:: aimet_torch.quant_analyzer.QuantAnalyzer.export_per_layer_mse_loss
 
 ..
   # end-before
diff --git a/Docs/apiref/torch/v2/quantization.rst → Docs/apiref/torch/quantization.rst b/Docs/apiref/torch/v2/quantization.rst → Docs/apiref/torch/quantization.rst
@@ -1,8 +1,8 @@
 .. _apiref-torch-quantization:
 
-###########################
-aimet_torch.v2.quantization
-###########################
+########################
+aimet_torch.quantization
+########################
 
 .. currentmodule:: aimet_torch.v2.quantization
 

diff --git a/Docs/apiref/torch/quantsim.rst b/Docs/apiref/torch/quantsim.rst
@@ -7,10 +7,6 @@ aimet_torch.quantsim
 ..
   # start-after
 
-.. note::
-
-    This module is also available in the experimental :mod:`aimet_torch.v2` namespace with the same top-level API.
-
 .. autoclass:: aimet_torch.quantsim.QuantizationSimModel
 
 **The following API can be used to Compute encodings for calibration:**
@@ -27,9 +23,6 @@ aimet_torch.quantsim
 
 .. automethod:: aimet_torch.quantsim.QuantizationSimModel.export
 
-Enum Definition
-===============
-
 **Quant Scheme Enum**
 
 .. autoclass:: aimet_common.defs.QuantScheme

diff --git a/Docs/apiref/torch/seq_mse.rst b/Docs/apiref/torch/seq_mse.rst
@@ -7,10 +7,6 @@ aimet_torch.seq_mse
 ..
   # start-after
 
-.. note::
-
-    This module is also available in the experimental :mod:`aimet_torch.v2` namespace with the same top-level API.
-
 **Top level APIs**
 
 .. autofunction:: aimet_torch.seq_mse.apply_seq_mse
@@ -22,4 +18,3 @@ aimet_torch.seq_mse
 
 ..
   # end-before
-
diff --git a/Docs/apiref/torch/v1/adaround.rst b/Docs/apiref/torch/v1/adaround.rst
@@ -0,0 +1,24 @@
+.. _apiref-torch-v1-adaround:
+
+#######################
+aimet_torch.v1.adaround
+#######################
+
+..
+  # start-after
+
+.. note::
+    This module is also available in the default :mod:`aimet_torch` namespace with
+    the same top-level API.
+
+**Top level APIs**
+
+.. autofunction:: aimet_torch.v1.adaround.adaround_weight.Adaround.apply_adaround
+
+**Adaround parameters**
+
+.. autoclass:: aimet_torch.v1.adaround.adaround_weight.AdaroundParameters
+    :members:
+
+..
+  # end-before
diff --git a/Docs/apiref/torch/amp.rst → Docs/apiref/torch/v1/amp.rst b/Docs/apiref/torch/amp.rst → Docs/apiref/torch/v1/amp.rst
@@ -1,12 +1,16 @@
-.. _api-torch-amp:
+.. _apiref-torch-v1-amp:
 
-===========================
-aimet_torch.mixed_precision
-===========================
+==============================
+aimet_torch.v1.mixed_precision
+==============================
 
 ..
   # start-after
 
+.. note::
+    This module is also available in the default :mod:`aimet_torch` namespace with
+    the same top-level API.
+
 **Top-level API**
 
 .. autofunction:: aimet_torch.v1.mixed_precision.choose_mixed_precision

diff --git a/Docs/apiref/torch/v1/autoquant.rst b/Docs/apiref/torch/v1/autoquant.rst
@@ -0,0 +1,22 @@
+.. _apiref-torch-v1-autoquant:
+
+#########################
+aimet_torch.v1.auto_quant
+#########################
+
+..
+  # start-after
+
+.. note::
+    This module is also available in the default :mod:`aimet_torch` namespace with
+    the same top-level API.
+
+**Top-level API**
+
+.. autoclass:: aimet_torch.v1.auto_quant.AutoQuantWithAutoMixedPrecision
+    :members:
+    :member-order: bysource
+    :no-index:
+
+..
+  # end-before