-
Notifications
You must be signed in to change notification settings - Fork 141
Issues: pytorch/ao
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Author
Label
Projects
Milestones
Assignee
Sort
Issues list
Make float8_dynamic_activation_float8_weight work with Good for newcomers
quantize
weights_only
load
float8
good first issue
#1118
opened Oct 18, 2024 by
drisspg
int4wo can't use same packed weight for cpu and cuda
performance
quantize
#1117
opened Oct 18, 2024 by
HDCharles
Named Symbol not found (torchchat #1298)
good first issue
Good for newcomers
#1110
opened Oct 17, 2024 by
mikekgfb
Assertion information on precompute_float8_dynamic_scale_for_fsdp is misleading
float8
#1108
opened Oct 17, 2024 by
goldhuang
MX-scale discrepancy during quantization and dequantization
mx
#1104
opened Oct 17, 2024 by
mariosfourn
[Bug] ERR: subclass doesn't implement <function multi_head_attention_forward>
autoquant
#1103
opened Oct 17, 2024 by
dgcnz
[MX | Triton] Create MX matmul op using new New feature or request
good first issue
Good for newcomers
mx
scaled_dot
op in Triton
enhancement
#1084
opened Oct 15, 2024 by
drisspg
Int8DynActInt4WeightQATQuantizer doesn't support qwen series
#1080
opened Oct 15, 2024 by
elfisworking
[ROCm] torchao.float8 should work properly on ROCm
float8
module: rocm
#1066
opened Oct 12, 2024 by
OrenLeung
How to use float8 with SM89 hardware - i.e. NVIDIA A6000 ADA?
float8
question
Further information is requested
#1057
opened Oct 11, 2024 by
vgoklani
Why is the inference speed of the quantized model using QAT so slow?
performance
qat
#1050
opened Oct 10, 2024 by
elfisworking
Should we require a specific version of
lm_eval
to simplify torchao/_models/_eval.py
?
#1025
opened Oct 7, 2024 by
ringohoffman
enable all the most recent ruff linter rules on torchao/float8 code
#1015
opened Oct 4, 2024 by
vkuzo
[easy] delete
torchao/float8/float8_aten_api.py
and move the functionality to float8_ops.py
#1014
opened Oct 4, 2024 by
vkuzo
Add weight tensor-wise scaling for INT8 quantized and mixed-precision training
enhancement
New feature or request
good first issue
Good for newcomers
#1010
opened Oct 4, 2024 by
gau-nernst
Previous Next
ProTip!
Find all open issues with in progress development work with linked:pr.