Pinned issues
Issues
- 3
Workload: Finally GPT2 Inference
#196 opened by hikettei - 0
Optimize: Dont lower the cached schedule-item
#228 opened by hikettei - 0
BugFix: Batch_Norm Scheduling with JIT=1
#225 opened by hikettei - 0
Optimize: LayerNorm = 1 Kernels
#224 opened by hikettei - 1
Fix: ConvND Failing Case
#203 opened by hikettei - 0
Fix Scheduler Workload
#222 opened by hikettei - 1
Beautiful docs
#159 opened by hikettei - 0
ShapeTracker: ConvND = 1 Kernels
#204 opened by hikettei - 0
- 0
Enhancement: Beautiful DOT=1, DOT=2
#221 opened by hikettei - 0
CI: ccl-bin is failing w/o saying EXIT=1
#187 opened by hikettei - 0
New Models
#217 opened by hikettei - 0
Optimize: Module is used to cache the lowered aasm
#192 opened by hikettei - 0
BugFix: Support and test the full dynamic shape compilation (test-dynamic-shape.lisp)
#211 opened by hikettei - 1
JIT: O(n) time scheduler
#191 opened by hikettei - 0
TODO: Add asin/acos/atan, asinh/acosh/atanh
#206 opened by hikettei - 0
Enhancement: Add rearrange
#143 opened by hikettei - 2
Plans for rewriting Caten/ajit
#145 opened by hikettei - 1
Rewrite iseq.lisp
#169 opened by hikettei - 0
TODO: Simplify threefry2x32 kernel
#209 opened by abourramouss - 0
Implement RoPE in Caten
#195 opened by hikettei - 0
Opt: Bring back WMMA
#198 opened by hikettei - 0
Fix ConvND Scheduling
#201 opened by hikettei - 1
Enhancement: AOT Shape Tracker
#200 opened by hikettei - 0
Opt: remove unused allocations with JIT=1 autodiff
#197 opened by hikettei - 5
Optimize: O(N) Compilation Time in Transformer
#148 opened by hikettei - 0
Enhancement: defmodule/defclass is always AOT.
#132 opened by hikettei - 0
CI: TODO list for caten/benchmarks running on CI
#180 opened by hikettei - 2
- 0
- 1
Enhancement: XXX-Style Render
#137 opened by hikettei - 1
TODO: Schedule Group Partitioning
#140 opened by hikettei - 0
Cannot lower (!softmax (make-tensor `(10)))
#173 opened by hikettei - 0
[TODO] Ideas for the high-level APIs
#150 opened by hikettei - 0
Refactor: Move float features to ./caten/common
#149 opened by hikettei - 1
- 1
Fix: ConvND Scheduling
#122 opened by hikettei - 3
Optimize: Scalar Promotion
#135 opened by hikettei - 1
Scheduler: Support Symbolic Increment
#136 opened by hikettei - 0
Implement Bitnet and Sparse Kernel
#147 opened by hikettei - 0
- 0
- 0
BugFix: 3D !tril/!triu should be in-place
#125 opened by hikettei - 0
Fix: !randint dtype inference
#129 opened by hikettei - 0
Feature: MLIR Renderer
#127 opened by hikettei - 1
Optimize: Broadcast+Matmul Fusion
#113 opened by hikettei - 0
- 0
Fix: ConvND Shape Inference
#118 opened by hikettei - 2
- 1
Enhancement: EXPR Simplifier
#111 opened by hikettei