Closed acha21 closed 3 years ago
Can you tell me what version of python you're on?
I use 3.8.10
I'll change every command in the README to have their double-hypenated versions to comply with higher Python versions.
In the meantime, you should try using --model
instead of -m
(rather than --m
)
Bug description When I use following command you provided in projects/hallucination/README.md, I cannot train retrieval-based models. When I use the option --m instead of -m, it works.
Reproduction steps
For example, Run the following: parlai train_model -m rag -t wizard_of_wikipedia \ --rag-model-type token --rag-retriever-type dpr --dpr-model-file zoo:hallucination/multiset_dpr/hf_bert_base.cp \ --generation-model bart -o arch/bart_large \ --retriever-debug-index compressed \ --batchsize 2 --fp16 True --gradient-clip 0.1 --label-truncate 128 \ --log-every-n-secs 30 --lr-scheduler reduceonplateau --lr-scheduler-patience 1 \ --model-parallel True --optimizer adam --text-truncate 512 --truncate 512 \ -lr 1e-05 -vmm min -veps 0.25 -vme 1000 -vmt ppl -vp 5
Expected behavior The options should be overridden by argument provided in command line
Logs
Additional context Add any other context about the problem here. (like proxy settings, network setup, overall goals, etc.)