Required library version not found: libbitsandbytes_cuda100.so. #82

junwycresta · 2022-10-31T21:57:53Z

CUDA SETUP: Required library version not found: libbitsandbytes_cuda100.so. Maybe you need to compile it from source?
CUDA SETUP: Defaulting to libbitsandbytes.so...

================================================ERROR=====================================
CUDA SETUP: CUDA detection failed! Possible reasons:

CUDA driver not installed
CUDA not installed
You have multiple conflicting CUDA libraries
Required library not pre-compiled for this bitsandbytes release!
CUDA SETUP: If you compiled from source, try again with make CUDA_VERSION=DETECTED_CUDA_VERSION for example, make CUDA_VERSION=113.
================================================================================

CUDA SETUP: Something unexpected happened. Please compile from source:
git clone [email protected]:TimDettmers/bitsandbytes.git
cd bitsandbytes
CUDA_VERSION=100
python setup.py install
CUDA SETUP: Required library version not found: libbitsandbytes_cuda100.so. Maybe you need to compile it from source?
CUDA SETUP: Defaulting to libbitsandbytes.so...

================================================ERROR=====================================
CUDA SETUP: CUDA detection failed! Possible reasons:

CUDA driver not installed
CUDA not installed
You have multiple conflicting CUDA libraries
Required library not pre-compiled for this bitsandbytes release!
CUDA SETUP: If you compiled from source, try again with make CUDA_VERSION=DETECTED_CUDA_VERSION for example, make CUDA_VERSION=113.
================================================================================

CUDA SETUP: Something unexpected happened. Please compile from source:
git clone [email protected]:TimDettmers/bitsandbytes.git
cd bitsandbytes
CUDA_VERSION=100
python setup.py install
CUDA SETUP: Required library version not found: libbitsandbytes_cuda100.so. Maybe you need to compile it from source?
CUDA SETUP: Defaulting to libbitsandbytes.so...

================================================ERROR=====================================
CUDA SETUP: CUDA detection failed! Possible reasons:

CUDA driver not installed
CUDA not installed
You have multiple conflicting CUDA libraries
Required library not pre-compiled for this bitsandbytes release!
CUDA SETUP: If you compiled from source, try again with make CUDA_VERSION=DETECTED_CUDA_VERSION for example, make CUDA_VERSION=113.
================================================================================

CUDA SETUP: Something unexpected happened. Please compile from source:
git clone [email protected]:TimDettmers/bitsandbytes.git
cd bitsandbytes
CUDA_VERSION=100
python setup.py install
CUDA SETUP: Something unexpected happened. Please compile from source:
git clone [email protected]:TimDettmers/bitsandbytes.git
cd bitsandbytes
CUDA_VERSION=100
python setup.py install
╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮
│ :1 in │
│ :18 in init │
│ :31 in _load_dependencies │
│ │
│ /root/venv/lib/python3.7/site-packages/transformers/pipelines/init.py:727 in pipeline │
│ │
│ 724 │ │ framework=framework, │
│ 725 │ │ task=task, │
│ 726 │ │ **hub_kwargs, │
│ ❱ 727 │ │ **model_kwargs, │
│ 728 │ ) │
│ 729 │ │
│ 730 │ model_config = model.config │
│ │
│ /root/venv/lib/python3.7/site-packages/transformers/pipelines/base.py:257 in │
│ infer_framework_load_model │
│ │
│ 254 │ │ │ │ ) │
│ 255 │ │ │ │
│ 256 │ │ │ try: │
│ ❱ 257 │ │ │ │ model = model_class.from_pretrained(model, **kwargs) │
│ 258 │ │ │ │ if hasattr(model, "eval"): │
│ 259 │ │ │ │ │ model = model.eval() │
│ 260 │ │ │ │ # Stop loading on the first successful load. │
│ │
│ /root/venv/lib/python3.7/site-packages/transformers/models/auto/auto_factory.py:464 in │
│ from_pretrained │
│ │
│ 461 │ │ elif type(config) in cls._model_mapping.keys(): │
│ 462 │ │ │ model_class = _get_model_class(config, cls._model_mapping) │
│ 463 │ │ │ return model_class.from_pretrained( │
│ ❱ 464 │ │ │ │ pretrained_model_name_or_path, *model_args, config=config, **hub_kwargs, │
│ 465 │ │ │ ) │
│ 466 │ │ raise ValueError( │
│ 467 │ │ │ f"Unrecognized configuration class {config.class} for this kind of AutoM │
│ │
│ /root/venv/lib/python3.7/site-packages/transformers/modeling_utils.py:2231 in from_pretrained │
│ │
│ 2228 │ │ │ model = cls(config, *model_args, **model_kwargs) │
│ 2229 │ │ │
│ 2230 │ │ if load_in_8bit: │
│ ❱ 2231 │ │ │ from .utils.bitsandbytes import get_keys_to_not_convert, replace_8bit_linear │
│ 2232 │ │ │ │
│ 2233 │ │ │ logger.info("Detected 8-bit loading: activating 8-bit loading for this model │
│ 2234 │
│ │
│ /root/venv/lib/python3.7/site-packages/transformers/utils/bitsandbytes.py:10 in │
│ │
│ 7 │ import torch │
│ 8 │ import torch.nn as nn │
│ 9 │ │
│ ❱ 10 │ import bitsandbytes as bnb │
│ 11 │
│ 12 if is_accelerate_available(): │
│ 13 │ from accelerate import init_empty_weights │
│ │
│ /root/venv/lib/python3.7/site-packages/bitsandbytes/init.py:6 in │
│ │
│ 3 # This source code is licensed under the MIT license found in the │
│ 4 # LICENSE file in the root directory of this source tree. │
│ 5 │
│ ❱ 6 from .autograd._functions import ( │
│ 7 │ MatmulLtState, │
│ 8 │ bmm_cublas, │
│ 9 │ matmul, │
│ │
│ /root/venv/lib/python3.7/site-packages/bitsandbytes/autograd/_functions.py:5 in │
│ │
│ 2 import warnings │
│ 3 │
│ 4 import torch │
│ ❱ 5 import bitsandbytes.functional as F │
│ 6 │
│ 7 from dataclasses import dataclass │
│ 8 from functools import reduce # Required in Python 3 │
│ │
│ /root/venv/lib/python3.7/site-packages/bitsandbytes/functional.py:13 in │
│ │
│ 10 from typing import Tuple │
│ 11 from torch import Tensor │
│ 12 │
│ ❱ 13 from .cextension import COMPILED_WITH_CUDA, lib │
│ 14 from functools import reduce # Required in Python 3 │
│ 15 │
│ 16 # math.prod not compatible with python < 3.8 │
│ │
│ /root/venv/lib/python3.7/site-packages/bitsandbytes/cextension.py:121 in │
│ │
│ 118 │ │ raise RuntimeError(''' │
│ 119 │ │ CUDA Setup failed despite GPU being available. Inspect the CUDA SETUP outputs to │
│ 120 │ │ If you cannot find any issues and suspect a bug, please open an issue with detal │
│ ❱ 121 │ │ https:/TimDettmers/bitsandbytes/issues''') │
│ 122 │ lib.cadam32bit_g32 │
│ 123 │ lib.get_context.restype = ct.c_void_p │
│ 124 │ lib.get_cusparse.restype = ct.c_void_p │
╰──────────────────────────────────────────────────────────────────────────────────────────────────╯
RuntimeError:
CUDA Setup failed despite GPU being available. Inspect the CUDA SETUP outputs to fix your environment!
If you cannot find any issues and suspect a bug, please open an issue with detals about your environment:
https:/TimDettmers/bitsandbytes/issues

The text was updated successfully, but these errors were encountered:

junwycresta · 2022-10-31T21:59:21Z

I tried to follow the instructions to compile the lib from source

CUDA SETUP: Something unexpected happened. Please compile from source:
git clone [email protected]:TimDettmers/bitsandbytes.git
cd bitsandbytes
CUDA_VERSION=100
python setup.py install

but the issue remains

TimDettmers · 2022-11-01T01:17:36Z

Thank you for reporting the problem.

Currently, CUDA 10.0 is not supported. CUDA 10.2 is supported, though.

CUDA 10.0 was supported in the past but required the maintenance of additional code since 10.0 does not support all the features of CUDA 10.2. If possible, upgrade to a different CUDA version. This should resolve the issue.

I will review the inclusion of CUDA 10.0 at a later time again.

junwycresta · 2022-11-01T16:13:30Z

Thanks a lot for the reply, I have CUDA 11.6 installed.

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 510.47.03    Driver Version: 510.47.03    CUDA Version: 11.6     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Tesla T4            On   | 00000000:00:1E.0 Off |                    0 |
| N/A   30C    P8    13W /  70W |      2MiB / 15360MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

CUDA_VERSION=100 is from the error message. I tried to update the version to my version

git clone [[email protected]](mailto:[email protected]):TimDettmers/bitsandbytes.git
cd bitsandbytes
CUDA_VERSION=116
python setup.py install

But still seeing the error message.

tomaarsen · 2022-11-09T14:31:13Z

Are you getting the same error message?

Teale001 · 2022-11-11T04:32:21Z

i get the same error

junwycresta · 2022-11-14T00:46:29Z

yes same error message.

giantvision · 2022-11-24T13:05:04Z

Any progress?

Titus-von-Koeller · 2022-11-24T13:49:04Z

Hey all, Tim is busy writing a new paper and I am completely swamped at work. I should be able to look into this in 7-14days from now. We're doing this in our free time. Thanks for your patience, Titus

…

On Thu, Nov 24, 2022, 14:05 yang ***@***.***> wrote: Any progress? — Reply to this email directly, view it on GitHub <#82 (comment)>, or unsubscribe <https:/notifications/unsubscribe-auth/ACFBEO7GQXEKGRRXBTJ2D2TWJ5RYXANCNFSM6AAAAAARTQEL4M> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

TimDettmers · 2023-01-03T13:58:37Z

Currently, CUDA 10.0 is still not supported. If you have the same error message after installing CUDA 11.6 and you get the message

CUDA_VERSION=100
then the CUDA setup still detects only the CUDA 10.0 version. You can try to:

uninstall CUDA (try conda list | grep cuda to see if old versions are installed)
Check if your path variable still have the path to the old CUDA version (you can try printenv | grep cuda)

I will update the CUDA install to throw an error instead to make it clearer what is going on.

github-actions · 2023-12-20T16:13:35Z

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

TimDettmers added the enhancement New feature or request label Nov 1, 2022

TimDettmers self-assigned this Nov 1, 2022

TimDettmers added a commit that referenced this issue Jan 3, 2023

Added error+instructions for unsupported CUDA 10.0 version #82

211ad59

github-actions bot closed this as completed Dec 29, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Required library version not found: libbitsandbytes_cuda100.so. #82

Required library version not found: libbitsandbytes_cuda100.so. #82

junwycresta commented Oct 31, 2022

junwycresta commented Oct 31, 2022

TimDettmers commented Nov 1, 2022

junwycresta commented Nov 1, 2022 •

edited

Loading

tomaarsen commented Nov 9, 2022

Teale001 commented Nov 11, 2022

junwycresta commented Nov 14, 2022

giantvision commented Nov 24, 2022

Titus-von-Koeller commented Nov 24, 2022 via email

TimDettmers commented Jan 3, 2023

github-actions bot commented Dec 20, 2023

Required library version not found: libbitsandbytes_cuda100.so. #82

Required library version not found: libbitsandbytes_cuda100.so. #82

Comments

junwycresta commented Oct 31, 2022

junwycresta commented Oct 31, 2022

TimDettmers commented Nov 1, 2022

junwycresta commented Nov 1, 2022 • edited Loading

tomaarsen commented Nov 9, 2022

Teale001 commented Nov 11, 2022

junwycresta commented Nov 14, 2022

giantvision commented Nov 24, 2022

Titus-von-Koeller commented Nov 24, 2022 via email

TimDettmers commented Jan 3, 2023

github-actions bot commented Dec 20, 2023

junwycresta commented Nov 1, 2022 •

edited

Loading