Code

Unifying the FX and TS Frontends#1404

Code

Merged

narendasan merged 23 commits intomasterpytorch/TensorRT:masterfrom shared_corepytorch/TensorRT:shared_coreCopy head branch name to clipboardNov 21, 2022

+1,704-281

Commits

Commits on Oct 13, 2022

• feat: Adding profiling support to the runtime

  Show description for e84c8aa
  narendasan

  committed
  e84c8aa

  Copy full SHA for e84c8aa

• refactor: A new TRTModule implementation using the internal runtime which should give TS for free

  Show description for 65a7c13
  narendasan

  committed
  65a7c13

  Copy full SHA for 65a7c13

• feat: let Input generate random tensors following the spec

  Show description for 808f3e2
  narendasan

  committed
  808f3e2

  Copy full SHA for 808f3e2

Commits on Nov 3, 2022

• feat!(//core/runtime): Allow the Runtime to use binding names to align I/O

  Show description for 0f003b8
  narendasan

  committed
  0f003b8

  Copy full SHA for 0f003b8

Commits on Nov 4, 2022

• fix(//core/runtime): Resolving some issues with the runtime ABI

  Show description for df3ac77
  narendasan

  committed
  df3ac77

  Copy full SHA for df3ac77

• feat(//core/runtime): Adding a TRT layer profiler

  Show description for e804455
  narendasan

  committed
  e804455

  Copy full SHA for e804455

• feat(//py): Exposed the new runtime in Python

  Show description for bbaf152
  narendasan

  committed
  bbaf152

  Copy full SHA for bbaf152

Commits on Nov 11, 2022

• feat(//py/torch_tensorrt/fx): Compliant TRTModule implementation based

  Show description for 71872df
  narendasan

  committed
  71872df

  Copy full SHA for 71872df

• refactor: CUDADevice -> RTDevice for better distinction from compile

  Show description for 10afcb2
  narendasan

  committed
  10afcb2

  Copy full SHA for 10afcb2

• feat(//examples): Demo that you can compile using FX then deploy in

  Show description for d14c7c4
  narendasan

  committed
  d14c7c4

  Copy full SHA for d14c7c4

• refactor(//py/torch_tensorrt): Updates to existing APIs for use in fx

  Show description for 7df9032
  narendasan

  committed
  7df9032

  Copy full SHA for 7df9032

Commits on Nov 18, 2022

• feat(//core/runtime): Encode TRT engine in base64 instead of raw bytes

  Show description for e3b01f3
  narendasan

  committed
  e3b01f3

  Copy full SHA for e3b01f3

• feat(//py/torch_tensorrt/fx): Adding the option to use the experimental

  Show description for ea270e3
  narendasan

  committed
  ea270e3

  Copy full SHA for ea270e3

• fix(//core/runtime): Fixing a bug where if an exception is thrown in

  Show description for c7e757b
  narendasan

  committed
  c7e757b

  Copy full SHA for c7e757b

• feat(//py/torch_tensorrt/TRTModule): Allow state_dict extraction

  Show description for 2e299db
  narendasan

  committed
  2e299db

  Copy full SHA for 2e299db

• Merge remote-tracking branch 'origin/master' into shared_core

  Show description for e9ae43e
  narendasan

  committed
  e9ae43e

  Copy full SHA for e9ae43e

• chore: Addressing merge conflicts

  Show description for 418bc6f
  narendasan

  committed
  418bc6f

  Copy full SHA for 418bc6f

• chore: lint

  Show description for 131759a
  narendasan

  committed
  131759a

  Copy full SHA for 131759a

• chore: remove print statements

  Show description for 4c544a3
  narendasan

  committed
  4c544a3

  Copy full SHA for 4c544a3

• fix: Fix cmake build

  Show description for bdc48d4
  narendasan

  committed
  bdc48d4

  Copy full SHA for bdc48d4

Commits on Nov 21, 2022

• refactor: Add a suffix to the TRTModuleNext class while it's

  Show description for 58681f9
  narendasan

  committed
  58681f9

  Copy full SHA for 58681f9

• docs: Update docs and examples

  Show description for 71082d3
  narendasan

  committed
  71082d3

  Copy full SHA for 71082d3

• refactor: Reorder the API since everything but the engine is optional

  Show description for e782cc9
  narendasan

  committed
  e782cc9

  Copy full SHA for e782cc9