-
Notifications
You must be signed in to change notification settings - Fork 2.2k
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
Error: make: *** [Makefile:203: train_gpt2cu] Error 255 #359
Comments
you need to either disable BF16 ( |
similar ERROR
|
Try upgrading your Cuda version to 12.4.1? |
This basically means functions are not available for computation capability <8.0.. Note, the header source is dependent on cuda tool kit version. Things that cannot be compiled in 12.1 may be compilable in 12.4 (this is the case for me). |
By default, make
# It is the same as:
PRECISION=BF16 make Compile with other options can also solve this issue. PRECISION=FP16 make
# or
PRECISION=FP32 make Related code in # Precision settings, default to bf16 but ability to override
PRECISION ?= BF16
VALID_PRECISIONS := FP32 FP16 BF16
ifeq ($(filter $(PRECISION),$(VALID_PRECISIONS)),)
$(error Invalid precision $(PRECISION), valid precisions are $(VALID_PRECISIONS))
endif
ifeq ($(PRECISION), FP32)
PFLAGS = -DENABLE_FP32
else ifeq ($(PRECISION), FP16)
PFLAGS = -DENABLE_FP16
else
PFLAGS = -DENABLE_BF16
endif |
Environment:
I encounter an error when I execute:
Warring and error message:
This problem or question might seem kind of stupid since I'm a beginner in CUDA and C. I would appreciate it if anyone could provide me with some solutions or suggestions.
The text was updated successfully, but these errors were encountered: