Any pointers would be appreciated. After too many hours I finally got jaxlib to install with miniforge after watching this (Mac M1 Monterey Installing Miniforge and Anaconda/Miniconda Side-by-Side - YouTube), but no luck yet with trax.
I have not tried trax
, but went through the process with other libs for other deeplearning courses. Are you running the Apple M1-specific version of tensorflow?
Not yet, havenβt installed tensorflow on this machine at all
I just followed this tutorial to install the HuggingFace transformers
library for Course 4, Week 3 ungraded labs. In the process, I installed tensorflow for the Apple Silicon, M1 mac: From Zero to Transformers on Apple M1 Chip | by Alfonso Farruggia | Medium
Havenβt tried again to install trax, however.
Just tried installing trax again with no success. With miniforge on an environment where tensorflow runs:
conda install trax
yields
Collecting
g package metadata (current_repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Collecting package metadata (repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
PackagesNotFoundError: The following packages are not available from current channels:
- trax
Current channels:
- https://conda.anaconda.org/conda-forge/osx-arm64
- https://conda.anaconda.org/conda-forge/noarch
To search for alternate channels that may provide the conda package you're
looking for, navigate to
https://anaconda.org
and use the search bar at the top of the page.
Then
pip3 install trax==1.3.9
yields
Collecting trax==1.3.9
Using cached trax-1.3.9-py2.py3-none-any.whl (629 kB)
Requirement already satisfied: absl-py in /Users/myuser/miniforgeenvs/tensorflowenv/lib/python3.9/site-packages (from trax==1.3.9) (1.0.0)
Collecting matplotlib
Downloading matplotlib-3.5.1-cp39-cp39-macosx_11_0_arm64.whl (7.2 MB)
ββββββββββββββββββββββββββββββββββββββββ 7.2/7.2 MB 9.3 MB/s eta 0:00:00
Collecting jaxlib
Downloading jaxlib-0.1.76-cp39-none-macosx_11_0_arm64.whl (57.3 MB)
ββββββββββββββββββββββββββββββββββββββββ 57.3/57.3 MB 19.2 MB/s eta 0:00:00
Requirement already satisfied: six in /Users/myuser/miniforgeenvs/tensorflowenv/lib/python3.9/site-packages (from trax==1.3.9) (1.15.0)
Collecting tensorflow-datasets
Downloading tensorflow_datasets-4.5.2-py3-none-any.whl (4.2 MB)
ββββββββββββββββββββββββββββββββββββββββ 4.2/4.2 MB 26.4 MB/s eta 0:00:00
Collecting jax
Downloading jax-0.2.28.tar.gz (887 kB)
ββββββββββββββββββββββββββββββββββββββββ 887.3/887.3 KB 20.6 MB/s eta 0:00:00
Preparing metadata (setup.py) ... done
Collecting funcsigs
Using cached funcsigs-1.0.2-py2.py3-none-any.whl (17 kB)
Requirement already satisfied: numpy in /Users/myuser/miniforgeenvs/tensorflowenv/lib/python3.9/site-packages (from trax==1.3.9) (1.21.5)
Collecting gym
Using cached gym-0.21.0-py3-none-any.whl
Collecting t5
Using cached t5-0.9.3-py3-none-any.whl (153 kB)
Collecting psutil
Using cached psutil-5.9.0.tar.gz (478 kB)
Preparing metadata (setup.py) ... done
Collecting gin-config
Using cached gin_config-0.5.0-py3-none-any.whl (61 kB)
Requirement already satisfied: scipy in /Users/myuser/miniforgeenvs/tensorflowenv/lib/python3.9/site-packages (from trax==1.3.9) (1.7.3)
ERROR: Could not find a version that satisfies the requirement tensorflow-text (from trax) (from versions: none)
ERROR: No matching distribution found for tensorflow-text
looks like this continues to be a problem for everyone. some people are trying to compile their own, but that doesnβt seem straightforward at allβ¦
opened ~4 weeks ago
I got jax
and jaxlib
on my M1 mac but no joy with tensorflow-text
or trax
.
UPDATE: got tensorflow-text installed in a python 3.8 virtual env
tensorflow-deps 2.8.0 0 apple
tensorflow-hub 0.12.0 pypi_0 pypi
tensorflow-macos 2.8.0 pypi_0 pypi
tensorflow-metal 0.4.0 pypi_0 pypi
tensorflow-text 2.8.1 pypi_0 pypi
still no joy on trax
Yes I was finally able to install TRAX on my M1 mac!
First I installed Tensorflow-text - using the correct version to match my tensorflow installation and the python version i have installed.
After I installed that - trax installed easily and error free!
You can find the versions for M1 for tensorflow and tensorflow addons below!
Iβve created a requirements.txt to fix problems withβ¦pretty old packages used by the notebooks in the course:
Not sure Iβve dealt with all the issues but at least it works with C3_W1 assignment notebook.
I realize this is an old thread, and my M1 is already something of a relic. However, for anyone coming along this trail, the requirements.txt provided above was a big help, so thanks @Jeongho_Jang
Still, in August 2024 I needed to make a few changes in order to get this to run.
The suggested level of seqio
is <= 0.0.16 but this lead to some incompatibilities for me further down the dependency ladder in clu
. Resolved by using seqio==0.0.8
Additionally, I had to downlevel from default for two packages:
numpy==1.26.0
scipy==1.11.3
Here are the key players that enabled me to successfully run the (old?) trax-based code locally on arm64 / Apple M1
tensorboard 2.8.0
tensorflow-datasets 4.9.3
tensorflow-macos 2.8.0
tensorflow-metadata 1.15.0
tensorflow-text 2.8.2
keras 2.8.0
keras-preprocessing 1.1.2
tf-keras 2.15.0
jax 0.2.28
jaxlib 0.1.76
ipython 8.18.1
python 3.9.19
trax 1.3.9