flexflow/FlexFlow

Issues

• Issue with FlexFlow LLM Compilation and Generation

  #1444 opened 7 months ago by QAZWSX0827
  2

• CUDA testing support in `proj`

  #1537 opened 3 months ago by lockshaw
  2

• Fix input, weight, noop in local execution

  #1422 opened 8 months ago by reyna-abhyankar
  0

• Add tests for managed_ff_stream and handle

  #1435 opened 7 months ago by oOTigger
  0

• Check whether an invocation is valid against a signature

  #1442 opened 7 months ago by reyna-abhyankar
  0

• Update rewriting search in unity algorithm

  #1501 opened 5 months ago by lockshaw
  0

• Fix embedding kernel refactor

  #1443 opened 7 months ago by reyna-abhyankar
  0

• Fix handling for device specific and binding arbitrary device specific types

  #1462 opened 6 months ago by reyna-abhyankar
  0

• Update files in `kernels` so that the `CMakeLists.txt` src pattern can be changed to `src/cuda/*.cu`

  #1502 opened 5 months ago by lockshaw
  0

• Move `OpTaskSignature` to dtgen, away from visitable

  #1465 opened 6 months ago by reyna-abhyankar
  1

• Add `hash` and `fmt` for `TaskBinding` in local execution

  #1503 opened 5 months ago by reyna-abhyankar
  0

• Fix profiling wrapper to avoid resource handle issues

  #1504 opened 5 months ago by reyna-abhyankar
  0

• Move `TensorSlotsBackingWithoutAddresses` over to dtgen

  #1468 opened 6 months ago by reyna-abhyankar
  0

• Add kernels/local-execution support for new `BatchNormAttrs` fields

  #1505 opened 5 months ago by lockshaw
  0

• Add a `SubstitutionBuilder` to make creating `Substitution`s less verbose and error-prone

  #1473 opened 6 months ago by lockshaw
  0

• Figure out what to do with `LazyLabelledDataflowGraph`

  #1513 opened 5 months ago by lockshaw
  0

• Improve unit tests for `ParallelComputationGraphBuilder` and `ComputationGraphBuilder`

  #1474 opened 6 months ago by lockshaw
  0

• Rename `filtermap_keys` and `filtermap_values` to `filtrans_keys` and `filtrans_values` for consistency

  #1514 opened 5 months ago by lockshaw
  0

• Add ability to document dtgen structs using doxygen

  #1475 opened 6 months ago by lockshaw
  0

• Implement `as_dot` for graphs in `utils/graph` in a less hacky way

  #1476 opened 6 months ago by lockshaw
  0

• Add a function in `op-attrs` that takes an `UnmappedOpCostEstimateKey` and generates an `OperatorTaskSpace`

  #1520 opened 4 months ago by lockshaw
  0

• Implement `is_valid_substitution`

  #1477 opened 6 months ago by lockshaw
  0

• Factor out common n-dimensional coordinate/index primitives

  #1528 opened 4 months ago by lockshaw
  0

• Replace/simplify DimOrdered

  #1483 opened 6 months ago by lockshaw
  0

• Add ability to get `vector` of all enum values to `proj dtgen`

  #1478 opened 6 months ago by lockshaw
  0

• Remove inheritance structure from graph objects

  #1484 opened 6 months ago by lockshaw
  0

• Add `num_inputs` check to `get_output_shapes(PCGOperatorAttrs, std::vector<ParallelTensorShape>)`

  #1496 opened 5 months ago by lockshaw
  0

• CUDA GPU CI for `repo-refactor`

  #1536 opened 3 months ago by lockshaw
  0

• `CHECK_VALID_OP_ATTR` is not that necessary anymore and should be removed

  #1498 opened 5 months ago by lockshaw
  0

• Add intermediate interface between `ComputationGraphBuilder` and the raw graph interface for testing

  #1499 opened 5 months ago by lockshaw
  0

• Add weight handling for SP decomposition of PCGs

  #1500 opened 5 months ago by lockshaw
  0

• Standardize kernel function signatures

  #1540 opened 3 months ago by lockshaw
  0

• Performance issue when batch_size is 32

  #1529 opened 4 months ago by letheantest
  0

• Stuck During Tree-Based Speculative Decoding with OPT Model

  #1526 opened 4 months ago by SeungjaeLim
  0

• Error when I use larger batch size for spec-infer

  #1491 opened 5 months ago by lhr-30
  1

• Add function to divide list of PCG inputs into inputs and weights in `op-attrs`

  #1469 opened 4 months ago by lockshaw
  1

• Tokenizer not optional

  #1515 opened 4 months ago by stelleg
  2

• multinode python: Legion error 67 alongside NCCL errors.

  #1480 opened 5 months ago by stelleg
  5

• cuIpcGetMemHandle triggered CUDA out of memory when I use flexflow on one gpu

  #1497 opened 5 months ago by Spacecat-zwh
  4

• Questions about the measurement of the latency

  #1454 opened 7 months ago by QAZWSX0827
  2

• it is possible to print the simulated running time and simualtion time cost？

  #1463 opened 6 months ago by tyn513
  0

• Inference with TP>1 and dissagregated qkv projection & attention operator

  #1447 opened 6 months ago by yingchen21
  1

• Issue with C++ inference for model meta-llama/Llama-2-70b-hf

  #1452 opened 7 months ago by DDDDDYTS
  0

• Add support for Tiktoken tokenizer in Request Manager

  #1438 opened 7 months ago by Flechman
  0

• Issue with debugging using cuda-gdb

  #1451 opened 7 months ago by Liu-Weijie
  0

• example inference test in cpp_inference_tests.sh does not terminate

  #1440 opened 7 months ago by Bob-Chen222
  0

• Sorry, it was a typo

  #1445 opened 7 months ago by QAZWSX0827
  0

• How to enable reduction parallel in substitutions?

  #1431 opened 7 months ago by weilinquan
  0

• How to enable reduction parallel in substitutions?

  #1432 opened 7 months ago by weilinquan
  0

• Request for Graph Pruning Algorithm Code Location in FlexLLM

  #1425 opened 8 months ago by zbtrs
  0