rmenegaux
commented
2 years ago Hi!
I ran a quick test on the master branch and the quantize command you indicate works, so the problem must be elsewhere.
The std::bad_alloc error is a bummer usually, but here are 2 possible causes I can think of:
- You are running out of memory (how much RAM does your CPU have?)
- The model you are trying to quantize is itself faulty. Does it work, i.e. can you make predictions with it?
Hope this helps
RA-Tsagairt
Hello,
i have trained several models using your code but they are all really large file sizes. To shrink their memory footprint I am using the command:
python fastDNA/fdna.py -quantize -model_name ~/path_to_model -train_fasta ~/path_to_train_fasta
but it returns the error:
terminate called after throwing an instance of 'std::bad_alloc' what(): std::bad_alloc Aborted (core dumped) Traceback (most recent call last): File "fastDNA/fdna.py", line 163, in
ft.quantize(model_path, args.train_fasta)
File "fastDNA/python/fastDNA/fastDNA.py", line 67, in quantize
return timed_check_call(command, name=readout, verbose=self.verbose)
File "fastDNA/python/fastDNA/fastDNA.py", line 13, in timed_check_call
subprocess.check_call(command, shell=True)
File "/home/secuser/anaconda3/envs/bio_glove/lib/python3.7/subprocess.py", line 363, in check_call
raise CalledProcessError(retcode, cmd)
subprocess.CalledProcessError: Command 'fastDNA/fastdna quantize -input /home/secuser/Rob/bio-predictor-development/fasta_files/25_10_21/train_bf10_C100_L500.fasta -output /home/secuser/Rob/bio-predictor-development/fast_dna/output/models/fdna_k14_d50_e500_lr0.1_n4.0 -qnorm' returned non-zero exit status 134.
I have tried this with various models and their respective training files; all returns the same error. Model sizes are ~20GB.