AIMET Files

What's Changed

• Cmake cleanup by @quic-michof in https://github.com/quic/aimet/pull/3138
• Add python implementation for cle by @quic-hitameht in https://github.com/quic/aimet/pull/3143
• Remove Dequantize class by @quic-kyunggeu in https://github.com/quic/aimet/pull/3144
• Add python implementation for bn fold by @quic-hitameht in https://github.com/quic/aimet/pull/3145
• Fix bug in onnx perchannel mode with ConvTranspose by @quic-mtuttle in https://github.com/quic/aimet/pull/3149
• Fix bug when layer has 'branch' in name by @quic-mtuttle in https://github.com/quic/aimet/pull/3148
• Set save_as_external_data to True when saving a large model using onn… by @quic-huzh in https://github.com/quic/aimet/pull/3151
• Remove hardcoded test directories by @quic-michof in https://github.com/quic/aimet/pull/3146
• Add more quantized modules by @quic-kyunggeu in https://github.com/quic/aimet/pull/3156
• Remove outdated note by @quic-ipendse in https://github.com/quic/aimet/pull/3113
• Eps as minimum scale by @quic-kyunggeu in https://github.com/quic/aimet/pull/3159
• Fix aimet_torch cle acceptance test by @quic-hitameht in https://github.com/quic/aimet/pull/3161
• Fix typo in the import by @quic-hitameht in https://github.com/quic/aimet/pull/3162
• Add fn to get fp32 lora layers by @quic-mangal in https://github.com/quic/aimet/pull/3160
• Fix intermittent qsim2 test failures by @quic-mtuttle in https://github.com/quic/aimet/pull/3165
• Change error to warning for tf-torch builds by @quic-mtuttle in https://github.com/quic/aimet/pull/3164
• Open Source AMP by @quic-ykota in https://github.com/quic/aimet/pull/3167
• More fine-grained options to set/detect C++ ABI in CMakeLists.txt. by @quic-michof in https://github.com/quic/aimet/pull/3166
• Bypasss bug in torch.onnx.export by @quic-kyunggeu in https://github.com/quic/aimet/pull/3170
• Register more quantized module definitions by @quic-kyunggeu in https://github.com/quic/aimet/pull/3171
• Open source AMP docs/examples by @quic-ykota in https://github.com/quic/aimet/pull/3172
• Add python implementation for bc by @quic-hitameht in https://github.com/quic/aimet/pull/3174
• Use higher tolerance for bn fold test which requires GPU by @quic-hitameht in https://github.com/quic/aimet/pull/3173
• Fix device issue with blockwise api by @quic-klhsieh in https://github.com/quic/aimet/pull/3177
• Fix minor bug in utils.py by @quic-kyunggeu in https://github.com/quic/aimet/pull/3169
• Fix torch AMP nightly tests by @quic-ykota in https://github.com/quic/aimet/pull/3178
• Added support for GridSample Op and made ScatterND op Jit tracable by @quic-ssayanta in https://github.com/quic/aimet/pull/3182
• Issue add quantizer to scale by @quic-mangal in https://github.com/quic/aimet/pull/3179
• Fix matmul per-channel quantization axis by @quic-mtuttle in https://github.com/quic/aimet/pull/3184
• Add QuantizedGroupNorm by @quic-kyunggeu in https://github.com/quic/aimet/pull/3183
• Exclude bias from quantization for matmul add pattern by @quic-hitameht in https://github.com/quic/aimet/pull/3185
• Support layernorm in onnx-quantsim by @quic-rgite in https://github.com/quic/aimet/pull/3180
• Use 0-D scale for per-tensor quantizers by @quic-kyunggeu in https://github.com/quic/aimet/pull/3186
• Rename elementwise_ops.py to nn/modules/custom.py by @quic-kyunggeu in https://github.com/quic/aimet/pull/3188
• fixing a bug in setting default input while building cache dataset by @quic-sendilk in https://github.com/quic/aimet/pull/3190
• Fix in-place function handling of QuantizedTensorBase by @quic-kyunggeu in https://github.com/quic/aimet/pull/3191
• Fix onnx quantsim failures with large models by @quic-mtuttle in https://github.com/quic/aimet/pull/3154
• Fix failure when model has unused MatMul layers by @quic-mtuttle in https://github.com/quic/aimet/pull/3192
• Enable param renaming for export to onnx direct by @quic-klhsieh in https://github.com/quic/aimet/pull/3194
• Dispatch-based implementation of quantized modules by @quic-kyunggeu in https://github.com/quic/aimet/pull/3195
• Allow encoding inheritance of Tensor.type and Tensor.type_as by @quic-kyunggeu in https://github.com/quic/aimet/pull/3196
• RmsNorm Op Support by @quic-gsaichai in https://github.com/quic/aimet/pull/3198
• Register RmsNorm for quantization by @quic-klhsieh in https://github.com/quic/aimet/pull/3202
• Allow compilation for cuda compute 90 by @quic-mtuttle in https://github.com/quic/aimet/pull/3203
• Add GPTVQ document by @quic-geunlee in https://github.com/quic/aimet/pull/3199
• Implement QuantizedBatchNorm/InstanceNorm by @quic-kyunggeu in https://github.com/quic/aimet/pull/3204
• Migrate qmodule definitions of custom modules to nn/quantized/custom.py by @quic-kyunggeu in https://github.com/quic/aimet/pull/3206
• Add placeholders for torch.nn.modules in alphabetical order by @quic-kyunggeu in https://github.com/quic/aimet/pull/3207
• Fix TestTensorQuantizer test that compared against uninitialized data. by @quic-michof in https://github.com/quic/aimet/pull/3205
• Simplify onnx IR using onnxscript by @quic-kyunggeu in https://github.com/quic/aimet/pull/3208
• Mark lazy quantize wrapper as unquantizable by @quic-klhsieh in https://github.com/quic/aimet/pull/3212
• Expose all custom modules to v2.nn namespace by @quic-kyunggeu in https://github.com/quic/aimet/pull/3213
• Updated version to 1.34.0 by @quic-mprajapa in https://github.com/quic/aimet/pull/3097
• Implement unary qmodules by @quic-kyunggeu in https://github.com/quic/aimet/pull/3216
• Add python only implementation for weight svd by @quic-hitameht in https://github.com/quic/aimet/pull/3218
• Implement QuantizedConcat by @quic-kyunggeu in https://github.com/quic/aimet/pull/3219
• Allow saving to external data for export to onnx direct by @quic-klhsieh in https://github.com/quic/aimet/pull/3222
• Deprecate tar based rank selection algo by @quic-hitameht in https://github.com/quic/aimet/pull/3223
• Handle variadic operators properly in propagate_output_encodings by @quic-kyunggeu in https://github.com/quic/aimet/pull/3217
• Implement quantized RNN family by @quic-kyunggeu in https://github.com/quic/aimet/pull/3224
• aimet_onnx adaround debugging by @quic-mtuttle in https://github.com/quic/aimet/pull/3226
• Implement non-unary qmodules by @quic-kyunggeu in https://github.com/quic/aimet/pull/3227
• Ensure encoding symmetry in sequential MSE by @quic-kyunggeu in https://github.com/quic/aimet/pull/3225
• Introducing the Quant Stats Visualizer - an interactive tool to visualize the statistics of quantized layers by @quic-ipendse in https://github.com/quic/aimet/pull/3228
• Delete deprecated svd interface by @quic-hitameht in https://github.com/quic/aimet/pull/3229
• Optimize memory consumption of quantize functions by @quic-kyunggeu in https://github.com/quic/aimet/pull/3230
• Always compute scale in float32 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3231
• Keras QAT accuracy mismatch issue in Range Learning Scheme by @quic-ssayanta in https://github.com/quic/aimet/pull/3215
• Quantized module impl for hardswish, hardsigmoid, and unflatten by @quic-kyunggeu in https://github.com/quic/aimet/pull/3233
• Implement quantized EmbeddingBag by @quic-kyunggeu in https://github.com/quic/aimet/pull/3234
• Upgrade onnx and onnxruntime version by @quic-mtuttle in https://github.com/quic/aimet/pull/3235
• QAT memory optimization by @quic-kyunggeu in https://github.com/quic/aimet/pull/3232
• Add list of ops whose params not to be quantized (re-adding) by @quic-rgite in https://github.com/quic/aimet/pull/3237
• Exception rules for specific ops as per backend constraints by @quic-rgite in https://github.com/quic/aimet/pull/3236
• Update matmul exception logic by @quic-klhsieh in https://github.com/quic/aimet/pull/3239
• Refrain using pretrained weights by @quic-hitameht in https://github.com/quic/aimet/pull/3241
• Delete deprecated tf1.x APIs by @quic-hitameht in https://github.com/quic/aimet/pull/3242
• Ignore non-tensor inputs during encoding propagation by @quic-kyunggeu in https://github.com/quic/aimet/pull/3243
• Update state dict for v2 float quantizers by @quic-klhsieh in https://github.com/quic/aimet/pull/3245
• feat(keras): Lambda layers for ConnectedGraph by @quic-ernst in https://github.com/quic/aimet/pull/3248
• SparseConv3D support by @quic-ristha in https://github.com/quic/aimet/pull/3250
• Modify Per-channel Quantizer to Per-Tensor while use load_encodings_to_sim API with float encodings by @quic-gsaichai in https://github.com/quic/aimet/pull/3252
• Upgrading CUDA-Version to 12.1 by @quic-gsaichai in https://github.com/quic/aimet/pull/3254
• Fix minor bug in encoding propagation by @quic-kyunggeu in https://github.com/quic/aimet/pull/3246
• Pickling bugfix for torch1.x by @quic-kyunggeu in https://github.com/quic/aimet/pull/3258
• Remove aimet torch dependency from aimet onnx by @quic-hitameht in https://github.com/quic/aimet/pull/3259
• Adding AIMET example files into variant specific tarball by @quic-gsaichai in https://github.com/quic/aimet/pull/3257
• Fixed onnx cle.ipynb. by @quic-rij in https://github.com/quic/aimet/pull/3064
• SpConv3d op fixes by @quic-ssayanta in https://github.com/quic/aimet/pull/3264
• Added missing standard C++ library to dependency list by @quic-mprajapa in https://github.com/quic/aimet/pull/3262
• Adds encodings for onnx missed layers in torch encoding file. by @quic-saksrai in https://github.com/quic/aimet/pull/3261
• Allow bfloat16 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3265
• Temporarily skip pylint plugin processing in PR Jenkinsfile due to un… by @quic-bharathr in https://github.com/quic/aimet/pull/3266
• Promote aimet v2 documentation by @quic-kyunggeu in https://github.com/quic/aimet/pull/3270
• Remove aimet tensor quantizer dep from aimet onnx by @quic-hitameht in https://github.com/quic/aimet/pull/3271
• Make quantsim compatible with deepspeed zero3/zero-offload pre-partitioned models by @quic-kyunggeu in https://github.com/quic/aimet/pull/3274
• Fix adaround null pointer bug by @quic-kyunggeu in https://github.com/quic/aimet/pull/3276
• Debug inf/nan values during compute encodings by @quic-mtuttle in https://github.com/quic/aimet/pull/3275
• Disable pylint warning by @quic-hitameht in https://github.com/quic/aimet/pull/3283
• Reparametrize affine quantizer with (qmin, qmax) by @quic-kyunggeu in https://github.com/quic/aimet/pull/3282
• Optimize GPTVQ peak memory usage by @quic-geunlee in https://github.com/quic/aimet/pull/3284
• Remove randomness when initializing centroid distances by @quic-geunlee in https://github.com/quic/aimet/pull/3285
• Implement broadcastable qdq kernel by @quic-mtuttle in https://github.com/quic/aimet/pull/3288
• Fix exception rule failures from branch op by @quic-mtuttle in https://github.com/quic/aimet/pull/3287
• Error handler for onnx simplifier in AdaRound by @quic-rgite in https://github.com/quic/aimet/pull/3273
• Fix deepspeed import error by @quic-hitameht in https://github.com/quic/aimet/pull/3289
• Make unit tests torch version agnostic in aimet-onnx by @quic-hitameht in https://github.com/quic/aimet/pull/3290
• Fix export adapter weights by @quic-mangal in https://github.com/quic/aimet/pull/3293
• Implement block level sampling to collect input tensor efficiently by @quic-geunlee in https://github.com/quic/aimet/pull/3298
• Fix typehint for create_encoding_from_dict by @quic-gsaichai in https://github.com/quic/aimet/pull/3297
• Ignore torch.uint8 during quantization by @quic-gsaichai in https://github.com/quic/aimet/pull/3295
• Added documentation for SparseConvolution support and code example by @quic-ssayanta in https://github.com/quic/aimet/pull/3300
• Export weight and enc api by @quic-ristha in https://github.com/quic/aimet/pull/3302
• Util to load torch model with weights from safetensors by @quic-ristha in https://github.com/quic/aimet/pull/3304
• Enable aimet-onnx acceptance tests by @quic-hitameht in https://github.com/quic/aimet/pull/3305
• Add comments to explain the background of hacky workaround by @quic-kyunggeu in https://github.com/quic/aimet/pull/3308
• Add more argument sanity checks for affine quantizer constructor by @quic-kyunggeu in https://github.com/quic/aimet/pull/3307
• Deprecate FakeQuantizationMixin by @quic-kyunggeu in https://github.com/quic/aimet/pull/3311
• Remove duplicate TOC entries by @quic-kyunggeu in https://github.com/quic/aimet/pull/3312
• Added fixes for spconv in gpu by @quic-ssayanta in https://github.com/quic/aimet/pull/3315
• Quant Stats Visualizer PR2 by @quic-ipendse in https://github.com/quic/aimet/pull/3316
• Adding encoding analyzer input check by @quic-pdangi in https://github.com/quic/aimet/pull/3317
• Add collective api for removing encodings by @quic-pdangi in https://github.com/quic/aimet/pull/3318
• Encoding analyzer documentation by @quic-pdangi in https://github.com/quic/aimet/pull/3319
• Fix issue with passthrough ops by @quic-seokan in https://github.com/quic/aimet/pull/3321
• Make qsim export spot on weights by @quic-seokan in https://github.com/quic/aimet/pull/3322
• SpConv fixes for dlc export by @quic-ssayanta in https://github.com/quic/aimet/pull/3324
• Spacing fix by @quic-ssayanta in https://github.com/quic/aimet/pull/3325
• Disable acceptance test temporarily by @quic-hitameht in https://github.com/quic/aimet/pull/3326
• Aimet docker version fix by @quic-jpolizzi in https://github.com/quic/aimet/pull/3292
• Updated version to 1.35.0 by @quic-bharathr in https://github.com/quic/aimet/pull/3327
• Updated installation docs for Torch cuda 12.x and TF cudart 11-8 by @quic-bharathr in https://github.com/quic/aimet/pull/3328
• Add visualize_stats API docs by @quic-mtuttle in https://github.com/quic/aimet/pull/3329
• Add BQ support to onnx QcQuantizeOp by @quic-mtuttle in https://github.com/quic/aimet/pull/3330
• Remove stale code by @quic-kyunggeu in https://github.com/quic/aimet/pull/3334
• Updated cmake files to use CMAKE_INSTALL_LIBDIR instead of AIMET_INSTALL_DIR by @quic-bharathr in https://github.com/quic/aimet/pull/3333
• Updated installation instructions and fixed wheel file installation commands by @quic-bharathr in https://github.com/quic/aimet/pull/3335
• Optimize quantization functions with torch.compile by @quic-kyunggeu in https://github.com/quic/aimet/pull/3336
• Test if quantized modules are torch.jit traceable by @quic-kyunggeu in https://github.com/quic/aimet/pull/3337
• Exclude abstract nn.Module from quantization by @quic-kyunggeu in https://github.com/quic/aimet/pull/3339
• Allow shape inference of CustomMarkers by @quic-kyunggeu in https://github.com/quic/aimet/pull/3338
• Implement quantized cumsum and mm family by @quic-kyunggeu in https://github.com/quic/aimet/pull/3340
• fix: resolve merge conflict in unit test by @quic-kyunggeu in https://github.com/quic/aimet/pull/3341
• Migrate 'implements' method to base class by @quic-kyunggeu in https://github.com/quic/aimet/pull/3342
• Minor refactoring in connected graph helper function by @quic-kyunggeu in https://github.com/quic/aimet/pull/3343
• Add top-level onnx BQ API by @quic-mtuttle in https://github.com/quic/aimet/pull/3345
• Quant Stats Visualizer - Handling Histogram Observers and extracting Advanced Stats - v2 by @quic-ipendse in https://github.com/quic/aimet/pull/3347
• Removed autosummary from docs by @quic-bharathr in https://github.com/quic/aimet/pull/3349
• Quant Stats Visualizer - Adding Box Plot visualization for Advanced Stats by @quic-ipendse in https://github.com/quic/aimet/pull/3350
• Ga build docker images by @quic-hitameht in https://github.com/quic/aimet/pull/3351
• Capture all static activation tensors for quantization by @quic-rgite in https://github.com/quic/aimet/pull/3355
• Exception rule for Gather op by @quic-rgite in https://github.com/quic/aimet/pull/3356
• Soft-deprecate rounding_mode parameter by @quic-kyunggeu in https://github.com/quic/aimet/pull/3358
• Enable seq mse with lpbq by @quic-klhsieh in https://github.com/quic/aimet/pull/3360
• Quantization User Guide edits by @dwelsch-esi in https://github.com/quic/aimet/pull/3348
• Add AutoQuant with AMP by @quic-kyunggeu in https://github.com/quic/aimet/pull/3361
• Declare explicit namespace package by @quic-kyunggeu in https://github.com/quic/aimet/pull/3364
• Use formal type comparison by @quic-kyunggeu in https://github.com/quic/aimet/pull/3365
• Rmsnorm update by @quic-klhsieh in https://github.com/quic/aimet/pull/3363
• Fix input order flipping issue in connected graph for Matmul, concat, Minimum and DynamicConv2d by @quic-gsaichai in https://github.com/quic/aimet/pull/3367
• Explicitly disable quantizer if tensor is not float by @quic-ykota in https://github.com/quic/aimet/pull/3368
• Fix typos in bq docstrings by @quic-mtuttle in https://github.com/quic/aimet/pull/3369
• Use untrained resnet to preclude network connection during unit test by @quic-kyunggeu in https://github.com/quic/aimet/pull/3370
• Add warning for the future change in the default value of quant_scheme by @quic-kyunggeu in https://github.com/quic/aimet/pull/3371
• Enable batch norm fold in quantsim v2 by @quic-pdangi in https://github.com/quic/aimet/pull/3372
• Add manual mixed precision APIs by @quic-ykota in https://github.com/quic/aimet/pull/3374
• Migrate quantizer group unit tests by @quic-kyunggeu in https://github.com/quic/aimet/pull/3375
• Migrate v1 code under aimet_torch.v1 subpackage by @quic-kyunggeu in https://github.com/quic/aimet/pull/3376
• Refactor onnx quantsim export by @quic-mtuttle in https://github.com/quic/aimet/pull/3377
• Fix node name within if node by @quic-hitameht in https://github.com/quic/aimet/pull/3378
• Clean up QcQuantizeOp encoding APIs by @quic-mtuttle in https://github.com/quic/aimet/pull/3381
• Fixed issue in op name retrival by @quic-cgulecha in https://github.com/quic/aimet/pull/3383
• Support 1.0.0 encoding format in aimet_onnx by @quic-mtuttle in https://github.com/quic/aimet/pull/3386
• Resolve pylint failures by @quic-mtuttle in https://github.com/quic/aimet/pull/3385
• Add tie observers utility in aimet onnx by @quic-hitameht in https://github.com/quic/aimet/pull/3387
• Revert onnx sim._export_encodings API by @quic-mtuttle in https://github.com/quic/aimet/pull/3388
• Add deprecation warning for encoding version by @quic-klhsieh in https://github.com/quic/aimet/pull/3390
• Unpinned torch version for aimet onnx variants by @quic-bharathr in https://github.com/quic/aimet/pull/3382
• Keep AIMET custom modules as a single op in connected graph by @quic-klhsieh in https://github.com/quic/aimet/pull/3393
• MMP: Ingest user requests and preprocess by @quic-ykota in https://github.com/quic/aimet/pull/3394
• Update Pytorch AMP supported kernels logic by @quic-klhsieh in https://github.com/quic/aimet/pull/3396
• Fixed torch and ONNX installation instructions (added missing find-links option) by @quic-bharathr in https://github.com/quic/aimet/pull/3397
• fix/Fix per-channel quantization bug in multi-output for aimet_onnx by @jinhyeokparknota in https://github.com/quic/aimet/pull/3400
• Fix typo by @quic-kyunggeu in https://github.com/quic/aimet/pull/3401
• Allow defining non-leaf quantized modules by @quic-kyunggeu in https://github.com/quic/aimet/pull/3402
• Fix aimet tensorflow doc rendering by @quic-hitameht in https://github.com/quic/aimet/pull/3403
• Compression doc edits. by @dwelsch-esi in https://github.com/quic/aimet/pull/3404
• converted pipelines to use matrix w/ aimet-extra running after by @quic-twilkens in https://github.com/quic/aimet/pull/3405
• Update QAT notebook by @emmanuel-ferdman in https://github.com/quic/aimet/pull/3391
• Resolve pylint failure by @quic-hitameht in https://github.com/quic/aimet/pull/3406
• Fixed in docker usage PR Jenkinsfile and script by @quic-bharathr in https://github.com/quic/aimet/pull/3407
• Activation clamping for onnx by @quic-rgite in https://github.com/quic/aimet/pull/3409
• Support LPBQ in aimet_onnx by @quic-mtuttle in https://github.com/quic/aimet/pull/3412
• Delete FakeQuantizationMixin documentation by @quic-kyunggeu in https://github.com/quic/aimet/pull/3413
• Edited Jupyter Notebook examples by @dwelsch-esi in https://github.com/quic/aimet/pull/3414
• Add descriptive error messages when importing encodings by @quic-mtuttle in https://github.com/quic/aimet/pull/3415
• Show actual import error when libpymo import fails by @quic-mtuttle in https://github.com/quic/aimet/pull/3417
• Use unittest.mock to mock onnxscript in doc build by @quic-mtuttle in https://github.com/quic/aimet/pull/3416
• Migrate v1-specific non-public modules to aimet_torch.v1 namespace by @quic-kyunggeu in https://github.com/quic/aimet/pull/3418
• Add documentation with concrete code examples for @implements API by @quic-kyunggeu in https://github.com/quic/aimet/pull/3419
• Fix duplicate import causing pylint warning by @quic-geunlee in https://github.com/quic/aimet/pull/3420
• Added Sequential MSE for ONNX by @quic-ristha in https://github.com/quic/aimet/pull/3422
• Onnx memory optimization by @quic-rgite in https://github.com/quic/aimet/pull/3424
• Fix unit-tests & pylint errors seq-mse by @quic-rgite in https://github.com/quic/aimet/pull/3427
• Set MatMul op to use per-tensor quantization by @quic-mtuttle in https://github.com/quic/aimet/pull/3429
• Support encoding constraints in aimet onnx by @quic-mtuttle in https://github.com/quic/aimet/pull/3430
• Add code placeholder for custom quantized modules by @quic-kyunggeu in https://github.com/quic/aimet/pull/3425
• Set cmake compiler flag to Level 3 Optimization to improve performance by @quic-ssayanta in https://github.com/quic/aimet/pull/3433
• Make quantsim quantize all constants except singleton parameter by @quic-kyunggeu in https://github.com/quic/aimet/pull/3431
• Rename get_encoding (singular) to get_encodings (plural) by @quic-kyunggeu in https://github.com/quic/aimet/pull/3437
• Disable BN running stats quantization in onnx by @quic-mtuttle in https://github.com/quic/aimet/pull/3434
• Allow nested dispatch by @quic-kyunggeu in https://github.com/quic/aimet/pull/3442
• Implement encoding rescaler by @quic-akhobare in https://github.com/quic/aimet/pull/3441
• Align v2 adaround module/package names with v1 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3444
• Make second argument of compute_encodings optional by @quic-kyunggeu in https://github.com/quic/aimet/pull/3443
• Remove separate quantsim v2 doc build by @quic-mtuttle in https://github.com/quic/aimet/pull/3447
• Draft implementation of lora.QuantizedLinear by @quic-kyunggeu in https://github.com/quic/aimet/pull/3451
• Adding support for custom normalization modules by @quic-pdangi in https://github.com/quic/aimet/pull/3450
• Ensure the synchronization of parameters using zero offload by @quic-huzh in https://github.com/quic/aimet/pull/3436
• removed editable mode stage/matrix by @quic-twilkens in https://github.com/quic/aimet/pull/3446
• Add model simplification step in onnx notebooks by @quic-rgite in https://github.com/quic/aimet/pull/3454
• Fix aimet_onnx standalone bn handling by @quic-mtuttle in https://github.com/quic/aimet/pull/3448
• Use formal type comparison only by @quic-kyunggeu in https://github.com/quic/aimet/pull/3453
• Declare quantized lora layers only when peft library is present by @quic-kyunggeu in https://github.com/quic/aimet/pull/3452
• Expose quantized tensor documentation by @quic-kyunggeu in https://github.com/quic/aimet/pull/3457
• Rename ExportableQuantModule to QuantizedModuleProtocol by @quic-kyunggeu in https://github.com/quic/aimet/pull/3455
• Fix onnx BQ segfaults with large models by @quic-mtuttle in https://github.com/quic/aimet/pull/3459
• Register quantized definitions of lora layers automatically upon import aimet_torch.v2 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3456
• Fix minor bug in has_hooks by @quic-kyunggeu in https://github.com/quic/aimet/pull/3460
• Add sandbox for new documentation by @quic-kyunggeu in https://github.com/quic/aimet/pull/3458
• Add js and xml to package_data by @quic-geunlee in https://github.com/quic/aimet/pull/3462
• Fix unexpected keyword error when deepspeed is not available by @quic-kyunggeu in https://github.com/quic/aimet/pull/3461
• Improve GitHub Actions CI by @quic-bharathr in https://github.com/quic/aimet/pull/3464
• MMP: Propagate user requests upstream, apply requests to quantsim by @quic-ashvkuma in https://github.com/quic/aimet/pull/3466
• Fix import failure in PyTorch 1.x by @quic-kyunggeu in https://github.com/quic/aimet/pull/3468
• Preserve input/output structure in Torch ConnectedGraph by @quic-ashvkuma in https://github.com/quic/aimet/pull/3467
• Remove match-case syntax for compatibility with Python<3.10 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3472
• Fix pylint error by @quic-kyunggeu in https://github.com/quic/aimet/pull/3473
• Fix onnx PCQ bug by @quic-mtuttle in https://github.com/quic/aimet/pull/3476
• Add beta content by @quic-kyunggeu in https://github.com/quic/aimet/pull/3471
• Add flag for push-button transition between v1 and v2 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3479
• Add quantized definition of custom numerical operations by @quic-pdangi in https://github.com/quic/aimet/pull/3481
• Fix unit test failure with deepspeed >= 0.15.4 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3482
• Unpin docutils version by @quic-kyunggeu in https://github.com/quic/aimet/pull/3478
• Remove save to yaml ability by @quic-klhsieh in https://github.com/quic/aimet/pull/3487
• Add sphinx extensions for new doc by @quic-kyunggeu in https://github.com/quic/aimet/pull/3477
• Preliminary refactoring for export versioning by @quic-kyunggeu in https://github.com/quic/aimet/pull/3483
• Install doc edits. by @dwelsch-esi in https://github.com/quic/aimet/pull/3398
• Remove transformers from packaging dependency by @quic-kyunggeu in https://github.com/quic/aimet/pull/3489
• Remove spconv from package dependency by @quic-kyunggeu in https://github.com/quic/aimet/pull/3488
• Updated Jenkinsfile to update aimet extra jobs for release branch PRs by @quic-bharathr in https://github.com/quic/aimet/pull/3465
• Remove cylp from package dependency by @quic-kyunggeu in https://github.com/quic/aimet/pull/3491
• Remove cumm from dependency by @quic-kyunggeu in https://github.com/quic/aimet/pull/3492
• Fix type error when spconv doesn't exist by @quic-kyunggeu in https://github.com/quic/aimet/pull/3493
• Replace progressbar with tqdm by @quic-kyunggeu in https://github.com/quic/aimet/pull/3494
• Add missing dependencies by @quic-kyunggeu in https://github.com/quic/aimet/pull/3495
• Restore broken zlib dependency by @quic-kyunggeu in https://github.com/quic/aimet/pull/3497
• Remove source code incompatibility with Python<3.9 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3496
• Fix Pylint warnings in aimet_torch connectedgraph by @quic-ashvkuma in https://github.com/quic/aimet/pull/3498
• Fix tie_quantizer errors with unquantizable ops by @quic-mtuttle in https://github.com/quic/aimet/pull/3505
• Relax overly strict aimet dependencies by @quic-mtuttle in https://github.com/quic/aimet/pull/3504
• Furo theme by @dwelsch-esi in https://github.com/quic/aimet/pull/3490
• Restructure module dependencies of v1/v2 seq_mse by @quic-kyunggeu in https://github.com/quic/aimet/pull/3484
• Support all python in ubuntu2204 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3511
• Loosen sphinx version requirement in py38 fast-release Dockerfile by @quic-kyunggeu in https://github.com/quic/aimet/pull/3513
• Add PYTHON_VERSION arg to fast-release Dockerfile by @quic-kyunggeu in https://github.com/quic/aimet/pull/3512
• Prevent exception rule from applying to float quantization by @quic-geunlee in https://github.com/quic/aimet/pull/3510
• Simplify fast-release docker image by @quic-kyunggeu in https://github.com/quic/aimet/pull/3515
• Rename QuantizerProtocol and QuantizedModuleProtocol as private classes by @quic-kyunggeu in https://github.com/quic/aimet/pull/3516
• Delete mha_classifier.py by @quic-kyunggeu in https://github.com/quic/aimet/pull/3518
• Fix pylint failures from cyclic import by @quic-mtuttle in https://github.com/quic/aimet/pull/3514
• Clean up fast-release directory by @quic-kyunggeu in https://github.com/quic/aimet/pull/3517
• Debug onnx BN fold with shared stats by @quic-mtuttle in https://github.com/quic/aimet/pull/3506
• Add special handling for attribute descriptors of QuantizedTensorBase by @quic-kyunggeu in https://github.com/quic/aimet/pull/3519
• Add ONNX Batch norm folding doc by @quic-geunlee in https://github.com/quic/aimet/pull/3521
• Updating AdaRound PyTorch docs by @quic-pdangi in https://github.com/quic/aimet/pull/3522
• Unpin PyTorch tiny version by @quic-kyunggeu in https://github.com/quic/aimet/pull/3525
• Use broadcasting impl for onnx per-channel quantization by @quic-mtuttle in https://github.com/quic/aimet/pull/3524
• Add installation instructions in beta docs by @quic-hitameht in https://github.com/quic/aimet/pull/3526
• Add quantization workflow by @quic-hitameht in https://github.com/quic/aimet/pull/3527
• Add missing sphinx extension for furo theme by @quic-hitameht in https://github.com/quic/aimet/pull/3529
• Replace lapacke with eigen by @quic-kyunggeu in https://github.com/quic/aimet/pull/3530
• Remove unused imports by @quic-geunlee in https://github.com/quic/aimet/pull/3531
• Populate content for the beta docs by @quic-hitameht in https://github.com/quic/aimet/pull/3535
• Update version 1.36.0 by @quic-msarma in https://github.com/quic/aimet/pull/3533
• Removing PT-113 legacy variant by @quic-msarma in https://github.com/quic/aimet/pull/3536
• Fix custom_tensor_utils for environments without spconv by @quic-gsaichai in https://github.com/quic/aimet/pull/3538
• Upgrade onnx and torch packages in docker image for docs by @quic-hitameht in https://github.com/quic/aimet/pull/3542
• Restructure module dependencies of v1/v2 adaround by @quic-kyunggeu in https://github.com/quic/aimet/pull/3549
• Temporarily restore ExportableQuantModule for backwards-compatibility by @quic-kyunggeu in https://github.com/quic/aimet/pull/3550
• Fix to_custom_tensor fallback to take two inputs by @quic-gsaichai in https://github.com/quic/aimet/pull/3552
• Support encoding inheritance of Tensor.T, H, mT, and mH by @quic-kyunggeu in https://github.com/quic/aimet/pull/3548
• Remove redundant furo theme CSS files. by @dwelsch-esi in https://github.com/quic/aimet/pull/3553
• Add css entry for script-output by @quic-kyunggeu in https://github.com/quic/aimet/pull/3545
• Polish adaround doc by @quic-kyunggeu in https://github.com/quic/aimet/pull/3547
• API reference for aimet_torch.v2 QuantizationSimModel by @quic-kyunggeu in https://github.com/quic/aimet/pull/3546
• Update encoding rescaler packaging logic by @quic-klhsieh in https://github.com/quic/aimet/pull/3554
• Fix typo in documentation by @quic-kyunggeu in https://github.com/quic/aimet/pull/3555
• Add Search and version selection by @dwelsch-esi in https://github.com/quic/aimet/pull/3543
• Modify PyTorch QuantSim code example by @quic-geunlee in https://github.com/quic/aimet/pull/3559
• Modify ONNX QuantSim code example by @quic-geunlee in https://github.com/quic/aimet/pull/3560
• Updating AIMET version to 2.0.0 by @quic-msarma in https://github.com/quic/aimet/pull/3556
• Upgraded tf-torch-cpu pip packages in dependencies list by @quic-msarma in https://github.com/quic/aimet/pull/3558
• Fix onnx imagenet evaluator by @quic-rgite in https://github.com/quic/aimet/pull/3557
• Add pending feature guide by @quic-hitameht in https://github.com/quic/aimet/pull/3561
• Quantizer doc by @quic-kyunggeu in https://github.com/quic/aimet/pull/3564
• MMP: Input and Output APIs by @quic-ashvkuma in https://github.com/quic/aimet/pull/3562
• Let cmake use exact system default python by @quic-kyunggeu in https://github.com/quic/aimet/pull/3563
• Fix onnx torch dependency by @quic-twilkens in https://github.com/quic/aimet/pull/3567
• Avoid recursion in get_ordered_ops by @quic-mtuttle in https://github.com/quic/aimet/pull/3565
• Fixed packaging mechanism to set portable rpaths for multiple .so fil… by @quic-bharathr in https://github.com/quic/aimet/pull/3570
• Fix warnings during document build by @quic-geunlee in https://github.com/quic/aimet/pull/3573
• Fixed pylint warnings from mixed precision by @quic-ashvkuma in https://github.com/quic/aimet/pull/3574
• Cleaned up logic for input dtypes while initializing quantizers by @quic-cgulecha in https://github.com/quic/aimet/pull/3569
• Utility to load modules in runtime added by @quic-ristha in https://github.com/quic/aimet/pull/3576
• Disable pylint warning by @quic-bharathr in https://github.com/quic/aimet/pull/3577
• Fix sklearn<1.6 backwards-compatibility issue by @quic-bharathr in https://github.com/quic/aimet/pull/3578
• BatchNorm fold bug fix for layers having submodules by @quic-ssayanta in https://github.com/quic/aimet/pull/3583
• Batch size issue in the per-layer output utility by @quic-ssayanta in https://github.com/quic/aimet/pull/3581
• Restructure batch_norm_fold module dependency by @quic-kyunggeu in https://github.com/quic/aimet/pull/3579
• Ensure pip is installed in fast-release docker image by @quic-kyunggeu in https://github.com/quic/aimet/pull/3585
• BatchNorm Folding issue incase of KerasTensor in kwargs by @quic-ssayanta in https://github.com/quic/aimet/pull/3588
• MMP: Resolve Contentions and add support for logging by @quic-ykota in https://github.com/quic/aimet/pull/3586
• Unpinned deepspeed, transformers and peft packages in dev dockers by @quic-bharathr in https://github.com/quic/aimet/pull/3595
• Add scaled dot product attention by @quic-hitameht in https://github.com/quic/aimet/pull/3593
• Fix AdaRound pad attribute handling for convtranspose by @quic-mtuttle in https://github.com/quic/aimet/pull/3590
• Allow per-channel quantization with GroupedBlockwise quantizers by @quic-mtuttle in https://github.com/quic/aimet/pull/3596
• MMP Reorganization by @quic-ashvkuma in https://github.com/quic/aimet/pull/3594
• Add advanced installation instructions by @quic-hitameht in https://github.com/quic/aimet/pull/3597
• Retag image using buildx by @quic-emironov in https://github.com/quic/aimet/pull/3592
• Make bias correction v2 compatible by @quic-kyunggeu in https://github.com/quic/aimet/pull/3601
• Migrate amp to v1 namespace by @quic-kyunggeu in https://github.com/quic/aimet/pull/3602
• Restructure quantsim guide by @quic-hitameht in https://github.com/quic/aimet/pull/3599
• Fix code incompatible with python 3.8 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3605
• Supplement unknown module type error message by @quic-kyunggeu in https://github.com/quic/aimet/pull/3604
• Make internal utils compatible with v2 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3606
• Refactor layer_database to be compatible with both v1 and v2 quantsim by @quic-kyunggeu in https://github.com/quic/aimet/pull/3608
• MMP: disable pylint fstring interpolation warning by @quic-ashvkuma in https://github.com/quic/aimet/pull/3607
• Make AMP util functions v2-compatible by @quic-kyunggeu in https://github.com/quic/aimet/pull/3609
• Make onnx ConnectedGraph one-to-one with onnx model by @quic-mtuttle in https://github.com/quic/aimet/pull/3610
• Added check for pad attribute by @quic-cgulecha in https://github.com/quic/aimet/pull/3612
• Optimize aimet torch 1 0 0 encoding export by @quic-klhsieh in https://github.com/quic/aimet/pull/3611
• Remove unused constants by @quic-kyunggeu in https://github.com/quic/aimet/pull/3603
• Implement aimet_torch.v2 AMP by @quic-kyunggeu in https://github.com/quic/aimet/pull/3615
• Handling the case for cast op where the dtype of the input tensor is same as that of the cast op by @quic-ssayanta in https://github.com/quic/aimet/pull/3614
• Migrate v1-specific util functions to aimet_torch.v1.utils by @quic-kyunggeu in https://github.com/quic/aimet/pull/3616
• Pin ORT session to CPU for deriving dtype of activation tensors by @quic-rgite in https://github.com/quic/aimet/pull/3620
• Migrate common portion to aimet_torch._base.amp by @quic-kyunggeu in https://github.com/quic/aimet/pull/3619
• Resturcture autoquant dependencies by @quic-seokan in https://github.com/quic/aimet/pull/3621
• Delete blockwise quant recipes by @quic-kyunggeu in https://github.com/quic/aimet/pull/3618
• Fix cyclic import pylint warning by @quic-hitameht in https://github.com/quic/aimet/pull/3622
• Raise import error from when aimet_torch.v2 is the default API by @quic-kyunggeu in https://github.com/quic/aimet/pull/3617
• Add links to past versions of docs by @quic-hitameht in https://github.com/quic/aimet/pull/3624
• Fix bug in torch imagenet trainer by @quic-mtuttle in https://github.com/quic/aimet/pull/3625
• Give informative error message for missing CG products by @quic-mtuttle in https://github.com/quic/aimet/pull/3626
• Remove unnecessary v1 imports by @quic-kyunggeu in https://github.com/quic/aimet/pull/3628
• Migrate common components to aimet_torch._base.quantsim by @quic-kyunggeu in https://github.com/quic/aimet/pull/3629
• Refactoring: Merge LazyParamQuantizer class into LazyQuantizer by @quic-kyunggeu in https://github.com/quic/aimet/pull/3630
• Migrate v1-specific util functions to aimet_torch.v1.utils by @quic-kyunggeu in https://github.com/quic/aimet/pull/3631
• Remove libpymo from bias_correction by @quic-kyunggeu in https://github.com/quic/aimet/pull/3633
• Restructure module dependencies of v1/v2 quant analyzer by @quic-seokan in https://github.com/quic/aimet/pull/3637
• Add autoquant with amp support to quantsim v2 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3632
• Remove v1 imports from bias_correction by @quic-kyunggeu in https://github.com/quic/aimet/pull/3634
• Insulate LazyQuantWrapper from libpymo by @quic-kyunggeu in https://github.com/quic/aimet/pull/3635
• Suppress pylint warning by @quic-kyunggeu in https://github.com/quic/aimet/pull/3638
• fix: fix sim.export bug of non-leaf quantized modules by @quic-kyunggeu in https://github.com/quic/aimet/pull/3639
• Optimize onnx quantsim init latency for large graphs by @quic-mtuttle in https://github.com/quic/aimet/pull/3640
• MMP: Add support for supported_kernels validation by @quic-ykota in https://github.com/quic/aimet/pull/3641
• Remove onnxsim from aimet_torch by @quic-kyunggeu in https://github.com/quic/aimet/pull/3642
• refactor: reduce v1 imports in quantsim code by @quic-kyunggeu in https://github.com/quic/aimet/pull/3643
• refactor: declare module, named_qmodule(), and qmodule() as interface by @quic-kyunggeu in https://github.com/quic/aimet/pull/3644
• refactor: remove v1 import in v2 AMP by @quic-kyunggeu in https://github.com/quic/aimet/pull/3646
• refactor: remove v1 import in v2 quant analyzer by @quic-kyunggeu in https://github.com/quic/aimet/pull/3647
• refactor: migrate v1.nn to _base.nn by @quic-kyunggeu in https://github.com/quic/aimet/pull/3645
• refactor: rename _base/quantsim/quantsim.py to _base/quantsim.py by @quic-kyunggeu in https://github.com/quic/aimet/pull/3648
• refactor: migrate quantsim init to _base.quantsim by @quic-kyunggeu in https://github.com/quic/aimet/pull/3649
• fix: fix bug and pylint warning by @quic-kyunggeu in https://github.com/quic/aimet/pull/3650
• Updated build n test script to remove docs build from non-docs varian… by @quic-bharathr in https://github.com/quic/aimet/pull/3652
• Add release notes for past releases by @quic-hitameht in https://github.com/quic/aimet/pull/3654
• refactor: migrate quantsim export() to _base.quantsim by @quic-kyunggeu in https://github.com/quic/aimet/pull/3653
• Glossary by @dwelsch-esi in https://github.com/quic/aimet/pull/3572
• test: add import test that checks isolation of v1 subpackage by @quic-kyunggeu in https://github.com/quic/aimet/pull/3656
• Bootstrap examples for PyTorch in quick start by @quic-hitameht in https://github.com/quic/aimet/pull/3655
• refactor: articulate imports by @quic-kyunggeu in https://github.com/quic/aimet/pull/3657
• Fix the acceptance test by @quic-hitameht in https://github.com/quic/aimet/pull/3660
• Promote new docs by @quic-hitameht in https://github.com/quic/aimet/pull/3659
• Suppress pylint cyclic-import false alarm by @quic-kyunggeu in https://github.com/quic/aimet/pull/3661
• refactor: make v1 unittest subdirectory by @quic-kyunggeu in https://github.com/quic/aimet/pull/3658
• refactor: import base/affine/float subpackages in v2.quantization subpackage by @quic-kyunggeu in https://github.com/quic/aimet/pull/3651
• Unpin PyTorch version by @quic-kyunggeu in https://github.com/quic/aimet/pull/3664
• Expose public APIs for loading/saving quantsim for backwards-compatbility by @quic-kyunggeu in https://github.com/quic/aimet/pull/3665
• Remove platform tag from aimet_torch pip package by @quic-kyunggeu in https://github.com/quic/aimet/pull/3666
• Export QuantizationSimModel API in root package by @quic-kyunggeu in https://github.com/quic/aimet/pull/3668
• Enabled PR job pylint warnings parsing plugin (for legacy Jenkins instances), fixed broken links in README and Examples MD files by @quic-bharathr in https://github.com/quic/aimet/pull/3667
• Export QuantizationSimModel API in root package by @quic-kyunggeu in https://github.com/quic/aimet/pull/3671
• Update aimet_torch.v2 apis to default in the docs by @quic-hitameht in https://github.com/quic/aimet/pull/3672
• Remove USE_PYTHON_IMPL flag by @quic-pdangi in https://github.com/quic/aimet/pull/3669
• build: exclude unnecessary files from pip package by @quic-kyunggeu in https://github.com/quic/aimet/pull/3675
• build: unpin setuptools by @quic-kyunggeu in https://github.com/quic/aimet/pull/3676
• Deprecate quantsim util APIs by @quic-kyunggeu in https://github.com/quic/aimet/pull/3677
• Export v2.quantization package to root package by @quic-kyunggeu in https://github.com/quic/aimet/pull/3674
• fix: remove USE_PYTHON_IMPL flag from nightly tests by @quic-kyunggeu in https://github.com/quic/aimet/pull/3678
• Impl seqmse offload by @quic-huzh in https://github.com/quic/aimet/pull/3663
• fix: fix sequential mse runtime error upon grouped conv2d by @quic-kyunggeu in https://github.com/quic/aimet/pull/3679
• Fix bug in onnx CLS pattern matching by @quic-kyunggeu in https://github.com/quic/aimet/pull/3682
• fix: fix wildcard import error with module-level getattr function by @quic-kyunggeu in https://github.com/quic/aimet/pull/3680
• Migrate spy_auto_quant to _base.auto_quant by @quic-kyunggeu in https://github.com/quic/aimet/pull/3683
• MMP updates to align with Python 3.8 requirements by @quic-ashvkuma in https://github.com/quic/aimet/pull/3681
• SeqMse Empty Initializer Fix by @quic-alanmaha in https://github.com/quic/aimet/pull/3685
• Deduplicate CallbackFunc class definitions by @quic-kyunggeu in https://github.com/quic/aimet/pull/3686
• Fix Seq mse output tensor bug by @quic-alanmaha in https://github.com/quic/aimet/pull/3688
• Change default quant scheme by @quic-kyunggeu in https://github.com/quic/aimet/pull/3687
• BREAKING_CHANGE: switch aimet_torch default API from v1 to v2 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3689
• test: insulate PyPI wheel package tests from other build artifacts by @quic-kyunggeu in https://github.com/quic/aimet/pull/3690
• Polish quick start guide by @quic-hitameht in https://github.com/quic/aimet/pull/3691
• MMP Misc enhancements by @quic-ykota in https://github.com/quic/aimet/pull/3692
• Updated install docs to removed pip version pinning requirement by @quic-bharathr in https://github.com/quic/aimet/pull/3693
• Updated version to 2.1.0 by @quic-msarma in https://github.com/quic/aimet/pull/3694
• Fixed multiple inputs issue for Seq mse by @quic-alanmaha in https://github.com/quic/aimet/pull/3695
• Fix handling for to_custom_tensor by @quic-klhsieh in https://github.com/quic/aimet/pull/3697
• Fix wheel file names install by @quic-bharathr in https://github.com/quic/aimet/pull/3699
• Docs update for rc2 by @quic-hitameht in https://github.com/quic/aimet/pull/3700
• Fixed input encodings export of Concat module by @quic-cgulecha in https://github.com/quic/aimet/pull/3702
• Updated model connections dict to work for multi-output tensors for a layer by @quic-ssayanta in https://github.com/quic/aimet/pull/3704
• Fix AMP v2 slowdown issue by @quic-akhobare in https://github.com/quic/aimet/pull/3705
• Fix AMP v2 slowdown issue due to compute_param_encodings by @quic-seokan in https://github.com/quic/aimet/pull/3706
• Optimize GPU min/max reductions by @quic-mtuttle in https://github.com/quic/aimet/pull/3710
• Fix type hint of eval_callback in AutoQuant of TensorFlow by @quic-hyeokoo in https://github.com/quic/aimet/pull/3711
• Resolve warnings in test_quantizable_lstm_export_encodings by @quic-hyeokoo in https://github.com/quic/aimet/pull/3712
• Replace deprecated alias 'np.float' to 'np.float64' in test_qc_quantize_wrapper_keras by @quic-hyeokoo in https://github.com/quic/aimet/pull/3713
• Seq Mse : Doc string updated for loss function by @quic-ristha in https://github.com/quic/aimet/pull/3715
• Redefine symmetry as offset=-round((qmin + qmax) / 2) by @quic-kyunggeu in https://github.com/quic/aimet/pull/3718
• Rename dummy calss not to use test prefix by @quic-hyeokoo in https://github.com/quic/aimet/pull/3717
• Add release notes for next release by @quic-hitameht in https://github.com/quic/aimet/pull/3720
• Rename aimet_torch wheel file name in docs by @quic-hitameht in https://github.com/quic/aimet/pull/3721
• Heading styles. by @dwelsch-esi in https://github.com/quic/aimet/pull/3719
• Format code blocks by @dwelsch-esi in https://github.com/quic/aimet/pull/3723
• Edit install and landing pages by @dwelsch-esi in https://github.com/quic/aimet/pull/3722
• Incorporate feedback for install pages by @quic-hitameht in https://github.com/quic/aimet/pull/3725
• Collapse/expand arrows by @dwelsch-esi in https://github.com/quic/aimet/pull/3724
• Update adaround_optimizer.py by @NikilXYZ in https://github.com/quic/aimet/pull/3709
• Streamline images in AIMET docs by @quic-pdangi in https://github.com/quic/aimet/pull/3727
• Fix the documentation tab links in README.md by @quic-hitameht in https://github.com/quic/aimet/pull/3729
• Remove duplicate install instructions from README.md by @quic-hitameht in https://github.com/quic/aimet/pull/3732
• Always represent zero accurately in v1 histogram by @quic-kyunggeu in https://github.com/quic/aimet/pull/3731
• Fix histogram bug in libpymo by @quic-kyunggeu in https://github.com/quic/aimet/pull/3733
• Change background to make code listing more legible in dark mode. by @dwelsch-esi in https://github.com/quic/aimet/pull/3738
• CI 2.0: Enabled wheel file generation and unit tests by @quic-kyunggeu in https://github.com/quic/aimet/pull/3741
• Aimet fastrelease by @quic-jpolizzi in https://github.com/quic/aimet/pull/3737
• Use onnx==1.16.2 in GitHub Action PR jobs by @quic-kyunggeu in https://github.com/quic/aimet/pull/3750
• Add QuantizedGridSample by @quic-kyunggeu in https://github.com/quic/aimet/pull/3744
• Simplify graph from ONNX only export by removing offset calculation nodes by @quic-seokan in https://github.com/quic/aimet/pull/3742
• Fix exception rule failure with non-leaf qmodules by @quic-mtuttle in https://github.com/quic/aimet/pull/3748
• Fix error message in debugging environment by @quic-kyunggeu in https://github.com/quic/aimet/pull/3751
• Fix broken link to documentation by @quic-kyunggeu in https://github.com/quic/aimet/pull/3752
• Implement onnx new export using onnx only export artifacts by @quic-seokan in https://github.com/quic/aimet/pull/3753
• Edit feature guide > cross-layer equalization. by @dwelsch-esi in https://github.com/quic/aimet/pull/3749
• Edit feature guide > batch norm folding. by @dwelsch-esi in https://github.com/quic/aimet/pull/3746
• Edit Adaround feature page. by @dwelsch-esi in https://github.com/quic/aimet/pull/3740
• Optimize onnx quantsim init data type inference by @quic-mtuttle in https://github.com/quic/aimet/pull/3747
• Edit feature guide analysis tool pages. by @dwelsch-esi in https://github.com/quic/aimet/pull/3739
• Edit feature guide > sequential MSE by @dwelsch-esi in https://github.com/quic/aimet/pull/3745
• Honor torch/tensorflow/onnx versions specified in GitHub Action workflow by @quic-kyunggeu in https://github.com/quic/aimet/pull/3759
• Remove invalid deprecation message in aimet_torch.v1 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3756
• Treat nullary torch.Tensor.type() as an attribute descriptor by @quic-kyunggeu in https://github.com/quic/aimet/pull/3757
• Update aimet_torch QuantizationSimModel API reference by @quic-kyunggeu in https://github.com/quic/aimet/pull/3760
• Add additional ops to tie observers by @quic-mtuttle in https://github.com/quic/aimet/pull/3754
• Restore aimet-torch-cpu variant in PR job by @quic-kyunggeu in https://github.com/quic/aimet/pull/3761
• Remove save/load_checkpoint from v2 quantsim api reference by @quic-kyunggeu in https://github.com/quic/aimet/pull/3762
• Throw runtime error if QuantizationSimModel takes untraceable input by @quic-kyunggeu in https://github.com/quic/aimet/pull/3764
• Edit batch norm re-estimation in the feature guide. by @dwelsch-esi in https://github.com/quic/aimet/pull/3763
• Edit AutoQuant page in feature guide. by @dwelsch-esi in https://github.com/quic/aimet/pull/3758
• Fix amp miscellaneous bug by @quic-kyunggeu in https://github.com/quic/aimet/pull/3768
• Skip aimet-extra stage in Jenkins PR jobs by @quic-kyunggeu in https://github.com/quic/aimet/pull/3765
• Add pylint stage to Github Action PR job by @quic-kyunggeu in https://github.com/quic/aimet/pull/3766
• Handle variadic functions properly in error message code examples by @quic-kyunggeu in https://github.com/quic/aimet/pull/3767
• Batch unknown module error messages in one exception by @quic-kyunggeu in https://github.com/quic/aimet/pull/3769
• Fix exception rules not searching all ops by @quic-kyunggeu in https://github.com/quic/aimet/pull/3772
• Upgrade pylint to <3 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3773
• Add doc build to GitHub Action PR job by @quic-kyunggeu in https://github.com/quic/aimet/pull/3778
• Add libpymo unit tests with pytest by @quic-kyunggeu in https://github.com/quic/aimet/pull/3779
• Selectively run PR jobs only on AIMET variants that need testing by @quic-kyunggeu in https://github.com/quic/aimet/pull/3781
• Remove excessive graph nodes iteration during ConnectedGraph construction by @quic-kyunggeu in https://github.com/quic/aimet/pull/3780
• Enable onnx graph optimizations by @quic-mtuttle in https://github.com/quic/aimet/pull/3782
• Make ConnectedGraph Op.output a list of products by @quic-mtuttle in https://github.com/quic/aimet/pull/3784
• Matmul channel axis fix for tensors of rank > 2 by @quic-mtuttle in https://github.com/quic/aimet/pull/3785
• Add static tensor quantizers into quantizer groups by @quic-mtuttle in https://github.com/quic/aimet/pull/3786
• Add C++ unit tests to PR job by @quic-kyunggeu in https://github.com/quic/aimet/pull/3788
• Remove to numpy by @quic-kyunggeu in https://github.com/quic/aimet/pull/3770
• Remove redundant forward by @quic-kyunggeu in https://github.com/quic/aimet/pull/3771
• Minor miscellaneous optimization & refactoring in connected graph by @quic-kyunggeu in https://github.com/quic/aimet/pull/3789
• Replace IterFirstX with itertools.islice by @quic-kyunggeu in https://github.com/quic/aimet/pull/3792
• Unify replace_* util functions with replace_modules by @quic-kyunggeu in https://github.com/quic/aimet/pull/3790
• Update v2 quantsim to handle models with child module sharing by @quic-kyunggeu in https://github.com/quic/aimet/pull/3791
• Issue add lora v3 utils by @quic-mangal in https://github.com/quic/aimet/pull/3787
• Refactor aimet onnx BQ logic to DlQuantization by @quic-mtuttle in https://github.com/quic/aimet/pull/3783
• Edit Quantsim pages. by @dwelsch-esi in https://github.com/quic/aimet/pull/3774
• Add config files by @quic-klhsieh in https://github.com/quic/aimet/pull/3795
• Update quantsim exceptions for V81 by @quic-klhsieh in https://github.com/quic/aimet/pull/3794
• Add docs for encoding version 1.0.0 by @quic-klhsieh in https://github.com/quic/aimet/pull/3800
• Rewrite User Guide workflow and tool (feature) pages. by @dwelsch-esi in https://github.com/quic/aimet/pull/3796
• Unify get_layer_by_name with get_named_module by @quic-kyunggeu in https://github.com/quic/aimet/pull/3793
• Fix autograd in-place modification error by @quic-kyunggeu in https://github.com/quic/aimet/pull/3798
• Fix infinite recursion error in subclasses of quantized module by @quic-kyunggeu in https://github.com/quic/aimet/pull/3797
• Decide minimum scale dynamically based on target dtype by @quic-kyunggeu in https://github.com/quic/aimet/pull/3799
• Deduplicate error messages by @quic-kyunggeu in https://github.com/quic/aimet/pull/3803
• Fix MatMul exception rule in aimet_onnx by @quic-mtuttle in https://github.com/quic/aimet/pull/3801
• Disable intermediate output quantizer in MatMul Add by @quic-mtuttle in https://github.com/quic/aimet/pull/3805
• Quantize scalar activations in aimet_torch.v1 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3806
• Removed opdef parser and dependencies by @quic-cgulecha in https://github.com/quic/aimet/pull/3775
• Add QW-LoRa and QWA-LoRa to docs by @quic-ashvkuma in https://github.com/quic/aimet/pull/3810
• Add DS_Store files to .gitignore by @quic-ashvkuma in https://github.com/quic/aimet/pull/3809
• Avoid multiple instances of same quantizer group by @quic-rgite in https://github.com/quic/aimet/pull/3808
• Replace pypi release workflow using pyproject.toml file by @quic-hitameht in https://github.com/quic/aimet/pull/3812
• Add torch==2.1.2 as extra dependency by @quic-kyunggeu in https://github.com/quic/aimet/pull/3811
• Update supergroups and norm layer naming in configs by @quic-mtuttle in https://github.com/quic/aimet/pull/3813
• Specify onnxruntime as primary build-time dependency by @quic-kyunggeu in https://github.com/quic/aimet/pull/3817
• Resolve pylint warnings in torch==2.6 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3816
• Fix Adaround caching bug in torch>=2.5 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3819
• Work around spconv bug in latest pytorch by @quic-kyunggeu in https://github.com/quic/aimet/pull/3818
• Expose BatchNormFold as public API by @quic-kyunggeu in https://github.com/quic/aimet/pull/3802
• Replace get_named_module with nn.Module.get_submodule by @quic-kyunggeu in https://github.com/quic/aimet/pull/3821
• Enable adaptive attention mask for LLMs by @quic-bozhxu in https://github.com/quic/aimet/pull/3814
• Implement BQ wrapper for encoding analyzers by @quic-mtuttle in https://github.com/quic/aimet/pull/3804
• Enable load encodings for 1.0.0 aimet_onnx by @quic-klhsieh in https://github.com/quic/aimet/pull/3822
• Fix onnxruntime version in reqs.txt file by @quic-hitameht in https://github.com/quic/aimet/pull/3823
• Enable input quantizers that take external scalar parameter as input by @quic-kyunggeu in https://github.com/quic/aimet/pull/3824
• Fix sim.onnx.export bug in per-channel quantization by @quic-kyunggeu in https://github.com/quic/aimet/pull/3825
• Make encoding 1.0.0 default by @quic-klhsieh in https://github.com/quic/aimet/pull/3827
• BREAKING_CHANGE: Use per-channel quantization by default by @quic-kyunggeu in https://github.com/quic/aimet/pull/3826
• Enable tests with cuda markers by @quic-hitameht in https://github.com/quic/aimet/pull/3828
• Fix acceptance test failures by @quic-kyunggeu in https://github.com/quic/aimet/pull/3829
• Fix aimet-onnx test dependency collision by @quic-kyunggeu in https://github.com/quic/aimet/pull/3830
• Fix some more acceptance tests failures by @quic-hitameht in https://github.com/quic/aimet/pull/3831
• Make QuantizedEmbedding's output inherit weight encodings by @quic-kyunggeu in https://github.com/quic/aimet/pull/3834
• Infer symmetry from offset in sim.onnx.export by @quic-kyunggeu in https://github.com/quic/aimet/pull/3833
• Implement new encoding schema prototype by @quic-kyunggeu in https://github.com/quic/aimet/pull/3832
• Resolve pylint warning by @quic-kyunggeu in https://github.com/quic/aimet/pull/3835
• Add nightly pipeline using reusable workflow by @quic-hitameht in https://github.com/quic/aimet/pull/3837
• Implement new LPBQ encoding JSON schema by @quic-kyunggeu in https://github.com/quic/aimet/pull/3839
• Add fast-release requirements.txt for onnx-cpu variant by @quic-kyunggeu in https://github.com/quic/aimet/pull/3836
• Move config files by @quic-klhsieh in https://github.com/quic/aimet/pull/3840
• Bug fix: Use correct method to find QGs in AMP Phase-3 by @quic-ykota in https://github.com/quic/aimet/pull/3842
• MMP: Remove hardcodings, query from config instead by @quic-ykota in https://github.com/quic/aimet/pull/3844
• Expose interface to find all batch norms to fold by @quic-ykota in https://github.com/quic/aimet/pull/3843
• Export bias encodings in sim.onnx.export by @quic-kyunggeu in https://github.com/quic/aimet/pull/3846
• Implement new encoding schema prototype in aimet-onnx by @quic-kyunggeu in https://github.com/quic/aimet/pull/3845
• Implement LPBQ encoding schema 2.0.0 (beta) in aimet-onnx by @quic-kyunggeu in https://github.com/quic/aimet/pull/3847
• Allow nightly workflow to run manually from actions tab by @quic-hitameht in https://github.com/quic/aimet/pull/3848
• Redefine LPBQ output_dtype of 2.0.0 encoding schema by @quic-kyunggeu in https://github.com/quic/aimet/pull/3849
• Remove root logger config by @quic-kyunggeu in https://github.com/quic/aimet/pull/3852
• Updated version to 2.2.0 by @quic-bharathr in https://github.com/quic/aimet/pull/3850
• Enable PR and post-merge jobs for release branches by @quic-hitameht in https://github.com/quic/aimet/pull/3851
• Implement C++ Blockwise Tensor Quantizer by @quic-mtuttle in https://github.com/quic/aimet/pull/3853
• LPBQ refactoring by @quic-kyunggeu in https://github.com/quic/aimet/pull/3841
• Move _encoding_shape util function to QcQuantizeOp by @quic-kyunggeu in https://github.com/quic/aimet/pull/3854
• Derive LPBQ bias encodings from per-channel weight scale by @quic-kyunggeu in https://github.com/quic/aimet/pull/3855
• Allow not defining quantized definition of trivial modules by @quic-kyunggeu in https://github.com/quic/aimet/pull/3856
• Fix bug in create_rand_tensors_given_shapes by @quic-kyunggeu in https://github.com/quic/aimet/pull/3857
• Remove nested_map by @quic-kyunggeu in https://github.com/quic/aimet/pull/3858
• Remove get_propagated_encoding_dict by @quic-kyunggeu in https://github.com/quic/aimet/pull/3859
• Remove get_input_shape_batch_size by @quic-kyunggeu in https://github.com/quic/aimet/pull/3860
• Move v1-specific util function to v1.utils by @quic-kyunggeu in https://github.com/quic/aimet/pull/3861
• Move winnow util function to winnow_utils.py by @quic-kyunggeu in https://github.com/quic/aimet/pull/3862
• Use blockwise quantizer in aimet onnx by @quic-mtuttle in https://github.com/quic/aimet/pull/3863
• Add additional information in pyproject.toml by @quic-hitameht in https://github.com/quic/aimet/pull/3864
• Add pip index in cmake args by @quic-hitameht in https://github.com/quic/aimet/pull/3866
• Remove unnecessary dependencies by @quic-kyunggeu in https://github.com/quic/aimet/pull/3865
• Use standard onnx operator names in quantsim config file by @quic-kyunggeu in https://github.com/quic/aimet/pull/3867
• Clean up aimet-onnx adaround activation sampler unit test by @quic-kyunggeu in https://github.com/quic/aimet/pull/3868
• Implement int32 bias encoding export mechanism in aimet-onnx by @quic-kyunggeu in https://github.com/quic/aimet/pull/3870
• BREAKING_CHANGE: Use min-max as default quant scheme in aimet-onnx by @quic-kyunggeu in https://github.com/quic/aimet/pull/3869
• Add readme in pyproject.toml, do not ignore in .dockerignore by @quic-hitameht in https://github.com/quic/aimet/pull/3873
• Add support for quantsim configs with encoding constraints by @quic-gsaichai in https://github.com/quic/aimet/pull/3872
• QW-lora + QWA-lora example update by @quic-ashvkuma in https://github.com/quic/aimet/pull/3874
• Skip unit test for documentation-only PRs by @quic-kyunggeu in https://github.com/quic/aimet/pull/3875
• Add release notes for 2.1.0 release by @quic-hitameht in https://github.com/quic/aimet/pull/3879
• Introducing GraphPasses to disable quantizers for sub-graph by @quic-bhushans in https://github.com/quic/aimet/pull/3876
• add copyright for graph registry by @quic-bhushans in https://github.com/quic/aimet/pull/3886
• Improve apply_requant_mask by @quic-bozhxu in https://github.com/quic/aimet/pull/3880
• Define quantsim config file aliases by @quic-kyunggeu in https://github.com/quic/aimet/pull/3881
• Remove MPP-specific util functions by @quic-kyunggeu in https://github.com/quic/aimet/pull/3885
• Add glibc version check for v1 import by @quic-kyunggeu in https://github.com/quic/aimet/pull/3883
• Revive affine.Dequantize class by @quic-kyunggeu in https://github.com/quic/aimet/pull/3887
• Fix loading fp16 encodings in aimet-onnx by @quic-mtuttle in https://github.com/quic/aimet/pull/3888
• Fix aimet-onnx bug in 1.0.0 export for fp16 encodings by @quic-mtuttle in https://github.com/quic/aimet/pull/3891
• Use CPU when getting act shapes in AMP to avoid OOM by @quic-mtuttle in https://github.com/quic/aimet/pull/3892
• Issue with quantizer group creation by @quic-ssayanta in https://github.com/quic/aimet/pull/3890
• Fix onnx AMP changing bitwidths of frozen quantizers by @quic-mtuttle in https://github.com/quic/aimet/pull/3893
• Clean up unused metadata from onnx model exported by sim.onnx.export by @quic-kyunggeu in https://github.com/quic/aimet/pull/3894
• Autoquant adaround encodings bugfix by @quic-klhsieh in https://github.com/quic/aimet/pull/3895
• Restrict aimet_onnx AutoQuant to encoding version 0.6.1 by @quic-klhsieh in https://github.com/quic/aimet/pull/3896
• aimet_torch omniquant frameworks. by @quic-muchhsu in https://github.com/quic/aimet/pull/3897
• Keras compute loss api update by @quic-klhsieh in https://github.com/quic/aimet/pull/3884
• [hotfix] remove direct dependency from aimet-torch pypi package by @quic-kyunggeu in https://github.com/quic/aimet/pull/3899
• Implement parameter quantizer folding mechanism by @quic-kyunggeu in https://github.com/quic/aimet/pull/3900
• Let bias encoding follow the granularity of weight encoding by @quic-kyunggeu in https://github.com/quic/aimet/pull/3901
• Fix the tag for aimet-torch whl file by @quic-hitameht in https://github.com/quic/aimet/pull/3904
• Optimize v2 histogram by @quic-kyunggeu in https://github.com/quic/aimet/pull/3907
• Decrease minimum scale of int4-16 to float32.eps by @quic-kyunggeu in https://github.com/quic/aimet/pull/3908
• Add get block input output caching to omniquant. by @quic-muchhsu in https://github.com/quic/aimet/pull/3911
• Adding RMSNormalization supergroup pass by @quic-bhushans in https://github.com/quic/aimet/pull/3906
• Replace index select with vectorized elementwise kernel by @quic-kyunggeu in https://github.com/quic/aimet/pull/3912
• Update seq_mse candidate logic by @quic-klhsieh in https://github.com/quic/aimet/pull/3910
• Add common/test to github action unit test by @quic-kyunggeu in https://github.com/quic/aimet/pull/3905
• Fix minor bug in v2 util by @quic-kyunggeu in https://github.com/quic/aimet/pull/3914
• Vecotorize sequential MSE blockwise loss calculation of nn.Linear by @quic-kyunggeu in https://github.com/quic/aimet/pull/3913
• Add min_max quant scheme as a new name of tf by @quic-kyunggeu in https://github.com/quic/aimet/pull/3915
• Add v1-style TfEnhancedEncodingAnalyzer in aimet_torch.v2 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3916
• Fix tie_quantizers propagating encodings to all inputs of Gather by @quic-mtuttle in https://github.com/quic/aimet/pull/3918
• update_encoding_spec by @quic-klhsieh in https://github.com/quic/aimet/pull/3919
• Add Omniquant LET modules. by @quic-muchhsu in https://github.com/quic/aimet/pull/3920
• Re-tag aimet-torch wheel by @quic-hitameht in https://github.com/quic/aimet/pull/3921
• Add init.py in aimet-onnx by @quic-kyunggeu in https://github.com/quic/aimet/pull/3922
• Make sim.onnx.export capable of exporting LPBQ encodings by @quic-kyunggeu in https://github.com/quic/aimet/pull/3917
• Export large models with all_tensors_to_one_file by @quic-mtuttle in https://github.com/quic/aimet/pull/3924
• Blockwise let training for Omniquant by @quic-muchhsu in https://github.com/quic/aimet/pull/3927
• Release notes 2.2 by @quic-hitameht in https://github.com/quic/aimet/pull/3929
• Upgrade to onnxruntime 1.19 + cuda 12 by @quic-mtuttle in https://github.com/quic/aimet/pull/3930
• Implement onnx QDQ graph export prototype in aimet-onnx by @quic-kyunggeu in https://github.com/quic/aimet/pull/3934
• Soft-deprecate AutoQuant from aimet-onnx by @quic-kyunggeu in https://github.com/quic/aimet/pull/3925
• Upgrade onnxruntime, cuda in remaining workflows by @quic-hitameht in https://github.com/quic/aimet/pull/3935
• Integrate LET module to omniquant optimizer pipeline. by @quic-muchhsu in https://github.com/quic/aimet/pull/3936
• Optimize aimet onnx export logic by @quic-mtuttle in https://github.com/quic/aimet/pull/3942
• Update version to 2.3.0 by @quic-hitameht in https://github.com/quic/aimet/pull/3944
• Dump omniquant metadata by @quic-muchhsu in https://github.com/quic/aimet/pull/3943
• Optimize aimet onnx export further by @quic-mtuttle in https://github.com/quic/aimet/pull/3945
• Delete blosc dependency by @quic-bhushans in https://github.com/quic/aimet/pull/3947
• Pin deepspeed version temporarily by @quic-hitameht in https://github.com/quic/aimet/pull/3949
• Ignore ReLU6 in cross-layer equalization by @quic-kyunggeu in https://github.com/quic/aimet/pull/3939
• Prevent Sequential MSE from optimizing the same module multiple times by @quic-klhsieh in https://github.com/quic/aimet/pull/3951
• Blockwise sampler by @quic-ashvkuma in https://github.com/quic/aimet/pull/3946
• Remove spconv from aimet-onnx test dependency by @quic-kyunggeu in https://github.com/quic/aimet/pull/3953
• Make CallbackFunc callable by @quic-kyunggeu in https://github.com/quic/aimet/pull/3956
• Update deprecated encoding version warning message by @quic-kyunggeu in https://github.com/quic/aimet/pull/3938
• Parametrize TARGETPLATFORM to support linux/arm64 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3954
• Extend compute_encodings API of aimet-onnx by @quic-kyunggeu in https://github.com/quic/aimet/pull/3955
• Update CLS relu6 deprecation warning message by @quic-kyunggeu in https://github.com/quic/aimet/pull/3958
• Fix tf onnx API rendering in docs by @quic-hitameht in https://github.com/quic/aimet/pull/3957
• Github action to sync internal repo to public by @quic-bhushans in https://github.com/quic/aimet/pull/3959
• Re tag aimet torch whls platform indepedent by @quic-hitameht in https://github.com/quic/aimet/pull/3961
• Amp quantizer groups update by @quic-klhsieh in https://github.com/quic/aimet/pull/3962
• Fix errors in deepspeed>=0.16.5 by @quic-kyunggeu in https://github.com/quic/aimet/pull/3963

New Contributors

• @quic-ipendse made their first contribution in https://github.com/quic/aimet/pull/3113
• @dwelsch-esi made their first contribution in https://github.com/quic/aimet/pull/3348
• @jinhyeokparknota made their first contribution in https://github.com/quic/aimet/pull/3400
• @quic-twilkens made their first contribution in https://github.com/quic/aimet/pull/3405
• @emmanuel-ferdman made their first contribution in https://github.com/quic/aimet/pull/3391
• @quic-msarma made their first contribution in https://github.com/quic/aimet/pull/3533
• @quic-hyeokoo made their first contribution in https://github.com/quic/aimet/pull/3711
• @NikilXYZ made their first contribution in https://github.com/quic/aimet/pull/3709
• @quic-bozhxu made their first contribution in https://github.com/quic/aimet/pull/3814
• @quic-bhushans made their first contribution in https://github.com/quic/aimet/pull/3876

Full Changelog: https://github.com/quic/aimet/compare/1.33.0...2.11.0