google-ai-edge/ai-edge-torch issues and pull requests

Issue - State: closed - Opened by BmanClark 5 months ago - 6 comments
Labels: status:awaiting user response, type:build/install, type:support, status:stale

#73 - Revert "Change tokens from i64 to i32 (#68)"

Pull Request - State: closed - Opened by paulinesho 5 months ago

#72 - Update quantizer to new nightly

Pull Request - State: closed - Opened by paulinesho 5 months ago - 1 comment

#71 - Pin ai-edge-quantizer nightly version

Pull Request - State: closed - Opened by paulinesho 5 months ago

#70 - Update test_model_conversion.py

Pull Request - State: open - Opened by haozha111 5 months ago

#69 - Add a system overview of Generative API

Pull Request - State: closed - Opened by haozha111 5 months ago

#68 - Change tokens from i64 to i32

Pull Request - State: closed - Opened by paulinesho 5 months ago

#67 - MacOS installation failed

Issue - State: closed - Opened by YuMinHui825 5 months ago - 7 comments
Labels: status:awaiting user response, type:build/install, type:support, status:stale

#66 - Add channel last IO transformation API

Pull Request - State: closed - Opened by chunnienc 5 months ago
Labels: ci:run

#65 - Interpreter can register builtin opcode after conversion

Issue - State: open - Opened by kohheekyung 5 months ago - 1 comment

#64 - Fix BuildInterpolateCompositePass with aten decompositions

Pull Request - State: closed - Opened by chunnienc 5 months ago
Labels: ci:run

#63 - Add README for unet layers.

Pull Request - State: closed - Opened by yichunk 5 months ago
Labels: ci:run

#62 - Add fx pass to remove sdpa composite zero mask input

Pull Request - State: closed - Opened by chunnienc 5 months ago
Labels: ci:run

#61 - Wrap embedding tables with HLFB

Pull Request - State: closed - Opened by paulinesho 5 months ago

#60 - Errors with incompatible versions of torch_xla and ai-edge-torch

Issue - State: closed - Opened by ccleavinger 5 months ago - 2 comments

#59 - Add search model api

Pull Request - State: closed - Opened by wuruoyu 5 months ago
Labels: ci:run

#58 - Skip flaky quantization test temporarily

Pull Request - State: closed - Opened by paulinesho 5 months ago

#57 - Remove 5D tensor reshape in attention layer implementation.

Pull Request - State: closed - Opened by yichunk 5 months ago - 3 comments
Labels: ci:run

#56 - support batch > 1 and add batch_size to model config.

Pull Request - State: closed - Opened by haozha111 5 months ago

#55 - Add UNet model config and refactor diffusion model.

Pull Request - State: closed - Opened by yichunk 5 months ago
Labels: ci:run

#54 - Update gemma and example model to use external KV.

Pull Request - State: closed - Opened by hheydary 5 months ago

#53 - LSTMs are being unrolled and decomposed after conversion

Issue - State: open - Opened by Doomski99 5 months ago

#52 - (DO NOT SUBMIT) Add source loc to debug info

Pull Request - State: closed - Opened by chunnienc 5 months ago

#51 - Fix broken decoder conversion of SD.

Pull Request - State: closed - Opened by yichunk 5 months ago
Labels: ci:run

#50 - Add 2D blocks used in diffusion model of stable diffusion.

Pull Request - State: closed - Opened by yichunk 5 months ago
Labels: ci:run

#49 - Add cross attention layer and refactor t5_attention

Pull Request - State: closed - Opened by yichunk 5 months ago
Labels: ci:run

#48 - Fix activation_config renaming in autoencolder_loader

Pull Request - State: closed - Opened by yichunk 5 months ago
Labels: ci:run

#47 - Add back missed deleted GELU_QUICK type.

Pull Request - State: closed - Opened by yichunk 5 months ago
Labels: ci:run

#46 - Migrate quantization to use AI Edge Quantizer

Pull Request - State: closed - Opened by paulinesho 5 months ago

#45 - Update README.md

Pull Request - State: closed - Opened by haozha111 5 months ago

#44 - Phi3 conversion OOM on A100

Issue - State: open - Opened by a8nova 5 months ago - 12 comments
Labels: override-stale

#43 - Text-To-Speech and Speech-To-Text models support

Issue - State: open - Opened by bhanuprakashbv 5 months ago

#42 - Add GeGLU activation and change activation type to activation config

Pull Request - State: closed - Opened by yichunk 5 months ago - 1 comment
Labels: ci:run

#41 - Add interpolate nearest composite support

Pull Request - State: closed - Opened by vamsimanchala 5 months ago

#40 - Update tf-nightly dep

Pull Request - State: closed - Opened by majiddadashi 5 months ago
Labels: ci:run

#39 - Refactor decoder of stable diffusion

Pull Request - State: closed - Opened by yichunk 5 months ago - 1 comment
Labels: ci:run

#38 - Bug: redefinition of symbol named 'gelu_decomp_0'

Issue - State: closed - Opened by Misaka17032 5 months ago - 1 comment

#37 - update tf-nightly version so that models w/ external cache can be converted.

Pull Request - State: closed - Opened by haozha111 5 months ago

#36 - External kv experiment

Pull Request - State: closed - Opened by hheydary 5 months ago

#35 - Migrate quantization to use AI Edge Quantizer

Pull Request - State: closed - Opened by paulinesho 5 months ago

#34 - Rename the gelu composite to avoid conflict

Pull Request - State: closed - Opened by majiddadashi 5 months ago - 1 comment
Labels: ci:run

#33 - Enable min cut partitioner by default for layout transpose optimization

Pull Request - State: closed - Opened by chunnienc 5 months ago
Labels: ci:model-coverage

#32 - Enable test for model conversion.

Pull Request - State: closed - Opened by haozha111 5 months ago - 1 comment

#31 - Update README.md to point to new code location for SDPA.

Pull Request - State: closed - Opened by haozha111 6 months ago

#30 - Refactor layers for CLIP text encoder of SD model

Pull Request - State: closed - Opened by yichunk 6 months ago - 1 comment
Labels: ci:run

#29 - Wrap GELU in composite so that it is lowered to tfl.gelu

Pull Request - State: closed - Opened by majiddadashi 6 months ago
Labels: ci:run

#28 - The result of ConvTranspose2d without bias followed by BatchNorm2d is incorrect.

Issue - State: closed - Opened by AkiSakurai 6 months ago - 2 comments

#27 - Unpin tf-nightly in requirements.txt

Pull Request - State: closed - Opened by chunnienc 6 months ago - 3 comments
Labels: ci:run

#26 - Phi3 conversion to tflite

Issue - State: closed - Opened by a8nova 6 months ago - 3 comments

#25 - Update and refactor tests for nightly release (2/2)

Pull Request - State: closed - Opened by chunnienc 6 months ago
Labels: ci:run

#24 - Create attention layers with external kv cache and prepare an example.

Pull Request - State: closed - Opened by haozha111 6 months ago

#23 - Fix tflite flags test

Pull Request - State: closed - Opened by majiddadashi 6 months ago
Labels: ci:run

#22 - Enable composite for avg_pool2d in ceil_mode

Pull Request - State: closed - Opened by majiddadashi 6 months ago
Labels: ci:run

#21 - Add stable diffusion tflite inference pipeline

Pull Request - State: closed - Opened by yichunk 6 months ago
Labels: ci:run

#20 - Converted model numerically different from the original torch model

Issue - State: closed - Opened by vmanoj1996 6 months ago - 5 comments
Labels: status:awaiting user response, status:more data needed, type:precision/accuracy, status:stale