tinygrad/tinygrad

Issues

• scheduler matcher speedup

  #7840 opened 17 days ago by Qazalin
  0

• delete LazyBuffer, everything is UOp

  #7697 opened 24 days ago by Qazalin
  2

• Use SUPPORT_BF16=0 with llama3.py leading to Segmentation fault

  #7543 opened a month ago by FFAMax
  3

• Proposal to switch `Allocator` with `Buffer` in all devices

  #7787 opened 20 days ago by geohot
  7

• suttle bug with duplicate indices in setitem

  #7739 opened 19 days ago by ttomsa
  8

• Import Device not working with pip3 install tinygrad

  #7777 opened 20 days ago by stringbuild3r
  0

• HCQ Read Timeout on NVIDIA, No error on Metal

  #7727 opened 23 days ago by rggs
  5

• Software

  #7768 opened 21 days ago by Apdure
  1

• alu of tensors on different devices give incomprehensible errors

  #7699 opened 24 days ago by chenyuxyz
  1

• ptxas/nvdisasm: Unsupported .version 8.5; current version is '7.5'

  #7729 opened 23 days ago by FFAMax
  1

• AMD unsupported Arch: gfx096

  #7704 opened 22 days ago by FlorianHeigl
  1

• `np.float_` was removed in the NumPy 2.0 release when I try to go through examples

  #7594 opened 24 days ago by benjamin-asdf
  4

• Tensor(numpy).realize() takes 0.85ms to schedule

  #7698 opened 24 days ago by nimlgen
  0

• Big graph

  #7044 opened 24 days ago by Qazalin
  0

• Tensor.gradient should be a function (method)

  #7183 opened 2 months ago by geohot
  2

• AMD gfx1010: "Unsupported arch: {self.arch}

  #7504 opened a month ago by FFAMax
  3

• QCOM backend: openpilot DM ONNX silently produces wrong outputs

  #7606 opened a month ago by haraschax
  1

• jit behaves differently based on output assignment

  #7637 opened a month ago by chenyuxyz
  1

• Bump mypy to 1.12.1

  #7176 opened 2 months ago by geohot
  3

• Does CLANG work with CPU SIMD?

  #7573 opened a month ago by chosen-ox
  0

• better generate dataset process

  #7558 opened a month ago by chenyuxyz
  2

• CLANG has different casting behavior with intermediate inf

  #7516 opened a month ago by chenyuxyz
  1

• TRANSCENDENTAL=2 Tensor([11], dtype='half').exp() returns inf

  #7421 opened a month ago by chenyuxyz
  3

• CUDA_VISIBLE_DEVICES: should it me added to env_vars doc?

  #7502 opened a month ago by FFAMax
  1

• examples/llama3.py: just stuck nowhere

  #7503 opened a month ago by FFAMax
  0

• Feature request: add support OPENCL_VISIBLE_DEVICES

  #7505 opened a month ago by FFAMax
  0

• "Caused GPU Hang Error" on M2 Max 64GB

  #7172 opened a month ago by rggs
  7

• Overlapped setitem failed on some devices

  #7161 opened 2 months ago by MaxWolf-01
  1

• VIZ=1 with big graphs

  #7435 opened a month ago by Qazalin
  0

• AMD GPU support on OpenBSD?

  #7420 opened a month ago by nalzok
  0

• simple retry for fetch

  #7189 opened 2 months ago by Qazalin
  5

• Is it possible specify GPU?

  #7318 opened a month ago by FFAMax
  0

• Use of SUPPORT_BF16=0 CLANG=1 with llama3.py seg faults (and draft fix)

  #7376 opened a month ago by stmuk
  0

• VIZ UI improvements

  #7067 opened a month ago by Qazalin
  0

• Finish #7179 adding 1B llama 3.2 support in GGUF

  #7180 opened a month ago by geohot
  0

• NV vs CUDA fuzz diffs

  #7133 opened 2 months ago by nimlgen
  0

• Question on export_model

  #7244 opened a month ago by zcbenz
  0

• canonicalize shapetracker

  #7237 opened 2 months ago by chenyuxyz
  0

• [bug] bf16 exp precision error

  #7219 opened 2 months ago by ignaciosica
  0

• Adding CUDA_HOME for conda env ?

  #7038 opened 2 months ago by keithyau
  2

• Use commit hash to get PR title in process replay

  #7211 opened 2 months ago by Qazalin
  0

• Invalid Metal library in conda env on M3 MacBook

  #7202 opened 2 months ago by sravell
  1

• Move fetch to Tensor.fetch (staticmethod)

  #7182 opened 2 months ago by geohot
  0

• Add tests on Fedora linux

  #7197 opened 2 months ago by bkmgit
  0

• Add Simple Retry Mechanism for Fetch Requests

  #7195 opened 2 months ago by abbasabro
  0

• No CI testing of the float sym_infer pathway

  #7181 opened 2 months ago by geohot
  0

• Creating tensors from blob with QCOM=1 fails

  #7151 opened 2 months ago by haraschax
  2

• EfficientNet Example from the Documentation Doesn't Work

  #7113 opened 2 months ago by AdrianoWeid
  2

• Running python3 -m pytest test/ uninstalled Cuda runtime 12.4 and installed 12.1

  #7091 opened 2 months ago by keithyau
  0

• Metal backend does not like .min() or .max() on uchar tensors

  #7036 opened 2 months ago by StanislavNikolov
  2