Open userbox020 opened 4 months ago
I get a similar error when using the prebuilt rocm wheel exllamav2-0.0.13.post1+rocm5.6-cp311-cp311-linux_x86_64.whl
ROCR_VISIBLE_DEVICES=1 python examples/chat.py -m ../../../models/exl2/tinyllama-1B-4.0bpw -mode llama (exl2)
perl: warning: Setting locale failed.
perl: warning: Please check that your locale settings:
LANGUAGE = "",
LC_ALL = (unset),
LC_TIME = "en_DE.UTF-8",
LANG = "en_US.UTF-8"
are supported and installed on your system.
perl: warning: Falling back to a fallback locale ("en_US.UTF-8").
perl: warning: Setting locale failed.
perl: warning: Please check that your locale settings:
LANGUAGE = "",
LC_ALL = (unset),
LC_TIME = "en_DE.UTF-8",
LANG = "en_US.UTF-8"
are supported and installed on your system.
perl: warning: Falling back to a fallback locale ("en_US.UTF-8").
-- Model: ../../../models/exl2/tinyllama-1B-4.0bpw
-- Options: []
-- Loading model...
Traceback (most recent call last):
File "/home/sinan/ml/llm/inference/exl2/exllamav2/examples/chat.py", line 87, in <module>
model, tokenizer = model_init.init(args, allow_auto_split = True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/sinan/.conda/envs/exl2/lib/python3.11/site-packages/exllamav2/model_init.py", line 101, in init
model.load(split)
File "/home/sinan/.conda/envs/exl2/lib/python3.11/site-packages/exllamav2/model.py", line 248, in load
for item in f: return item
File "/home/sinan/.conda/envs/exl2/lib/python3.11/site-packages/exllamav2/model.py", line 266, in load_gen
module.load()
File "/home/sinan/.conda/envs/exl2/lib/python3.11/site-packages/exllamav2/attn.py", line 189, in load
self.q_proj.load()
File "/home/sinan/.conda/envs/exl2/lib/python3.11/site-packages/exllamav2/linear.py", line 45, in load
if w is None: w = self.load_weight()
^^^^^^^^^^^^^^^^^^
File "/home/sinan/.conda/envs/exl2/lib/python3.11/site-packages/exllamav2/module.py", line 97, in load_weight
qtensors["q_perm"] = torch.argsort(qtensors["q_invperm"]).to(torch.int)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: HIP error: invalid device function
HIP kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing HIP_LAUNCH_BLOCKING=1.
Compile with `TORCH_USE_HIP_DSA` to enable device-side assertions.
@SinanAkkoyun I think new mesa drivers 24.1 solve the issue, havent check yet
Any updates?
Im using rocm5.6 and installing enviroment with ooba one click install and im getting the follow error when loading models
However can run llamacpp models on same gpu and environment without any errors