mozilla / firefox-translations-training

Training pipelines for Firefox Translations neural machine translation models
https://mozilla.github.io/firefox-translations-training/
Mozilla Public License 2.0
149 stars 31 forks source link

TARGET not found #259

Open AmitMY opened 11 months ago

AmitMY commented 11 months ago

When I run

make run TARGET=merge_corpus

(or any target)

taken from https://github.com/mozilla/firefox-translations-training/blob/main/docs/snakemake.md

I see

echo "Running with config configs/config.prod.yml and profile local"
source ../mambaforge/etc/profile.d/conda.sh ; conda activate ; conda activate snakemake
chmod +x profiles/local/*
export SNAKEMAKE_OUTPUT_CACHE=../cache; snakemake \
  --profile=profiles/local \
  --configfile configs/config.prod.yml \
  merge_corpus
Running with config configs/config.prod.yml and profile local
Traceback (most recent call last):
  File "/mambaforge/envs/snakemake/bin/snakemake", line 10, in <module>
    sys.exit(main())
  File "/mambaforge/envs/snakemake/lib/python3.10/site-packages/snakemake/__init__.py", line 2755, in main
    success = snakemake(
  File "/mambaforge/envs/snakemake/lib/python3.10/site-packages/snakemake/__init__.py", line 500, in snakemake
    update_config(overwrite_config, load_configfile(f))
  File "/mambaforge/envs/snakemake/lib/python3.10/site-packages/snakemake/io.py", line 1618, in load_configfile
    config = _load_configfile(configpath)
  File "/mambaforge/envs/snakemake/lib/python3.10/site-packages/snakemake/io.py", line 1583, in _load_configfile
    obj = open(configpath_or_obj, encoding="utf-8")
FileNotFoundError: [Errno 2] No such file or directory: 'merge_corpus'
Makefile:69: recipe for target 'run' failed
make: *** [run] Error 1
eu9ene commented 10 months ago

Interesting, dry-run works fine on my local machine make dry-run TARGET=merge_corpus. You can also try using absolute path to the expected output file.

AmitMY commented 10 months ago

Tried that, without success.

You can see my exact machine config here: https://github.com/sign-language-processing/signbank-plus/blob/main/signbank_plus/nmt/bergamot/docker/Dockerfile

The RUN make test TARGET=train_vocab fails (and I also tried with the file path). not sure what is the difference.

AmitMY commented 9 months ago

Anything else I can check? running the test until train_vocab inside my docker build will accelerate running experiments for me