Opt Presets

Opt presets are a way to provide multiple options on the command line as shorthand. Opt presets are bundled with ParlAI and may be used by simply invoking the -o preset_name option within any ParlAI command.

You may also define your own options by placing them in ~/.parlai/opt_presets/. For example, creating ~/.parlai/opt_presets/myfolder/mypreset.opt allows you to invoke it via -o myfolder/mypreset. These preset files are simple json files containing a dictionary of files. For example:

{
    "inference": "beam",
    "beam_size": 10,
}

List of presets

The following is a list of all options presets bundled with the latest version of ParlAI.

Preset name

Description

Expansion

arch/bart_large

Architecture parameters (number layers, etc.) for BART-Large. See Lewis et. al. (2019

--activation gelu --attention-dropout 0.0 --dict-file zoo:bart/bart_large/model.dict --dict-tokenizer gpt2 --dropout 0.1 --embedding-size 1024 --embeddings-scale False --ffn-size 4096 --force-fp16-tokens True --fp16 True --init-model zoo:bart/bart_large/model --learn-positional-embeddings True --model bart --n-decoder-layers 12 --n-encoder-layers 12 --n-heads 16 --n-positions 1024 --variant bart

arch/blenderbot_3B

Architecture parameters (number layers, etc) for BlenderBot 3B. See Roller et al. (2020)

--activation gelu --attention-dropout 0 --embedding-size 2560 --ffn-size 10240 --label-truncate 128 --model transformer/generator --n-decoder-layers 24 --n-encoder-layers 2 --n-heads 32 --n-positions 128 --relu-dropout 0 --text-truncate 128 --truncate 128 --variant prelayernorm

arch/r2c2_base_3B

Architecture parameters for R2C2 Base 3B model. SeeShuster et al. (2022)

--activation gelu --delimiter \n --dict-tokenizer gpt2 --embedding-size 2048 --embeddings-scale True --ffn-size 8192 --force-fp16-tokens True --history-add-global-end-token end --init-model zoo:seeker/r2c2_base_3B/model --label-truncate 1024 --learn-positional-embeddings True --model bart --n-decoder-layers 22 --n-encoder-layers 22 --n-heads 32 --n-layers 22 --n-positions 1024 --n-segments 0 --output-scaling 1 --share-word-embeddings True --split-lines True --text-truncate 1024 --truncate 1024 --variant prelayernorm

arch/r2c2_base_400M

Architecture parameters for R2C2 Base 400M model. SeeShuster et al. (2022)

--activation gelu --attention-dropout 0.1 --delimiter \n --dict-tokenizer gpt2 --dropout 0.1 --embedding-size 1024 --embeddings-scale True --ffn-size 4096 --force-fp16-tokens True --history-add-global-end-token None --init-model zoo:seeker/r2c2_base_400M/model --label-truncate 1024 --learn-positional-embeddings True --model bart --n-decoder-layers 12 --n-encoder-layers 12 --n-heads 16 --n-layers 12 --n-positions 1024 --n-segments 0 --output-scaling 1 --rank-candidates False --relu-dropout 0 --share-word-embeddings True --split-lines True --text-truncate 1024 --truncate 1024 --variant prelayernorm

gen/blenderbot

Beam search parameters for BlenderBot. SeeRoller et al. (2020)

--beam-context-block-ngram 3 --beam-block-ngram 3 --beam-size 10 --inference beam --beam-min-length 20 --beam-block-full-context False

gen/meena

Inference parameters for the Sample & Rank procedure of Meena. See Adiwardana et al. (2020).

--beam-size 20 --inference topk --topk 40

gen/seeker_dialogue

Generation parameters for SeeKeR, Dialogue. SeeShuster et al. (2022)

--beam-disregard-knowledge-for-context-blocking False --datatype valid --drm-beam-block-full-context True --drm-beam-block-ngram 3 --drm-beam-context-block-ngram 3 --drm-beam-min-length 20 --drm-beam-size 10 --drm-inference beam --drm-message-mutators None --drm-model projects.seeker.agents.seeker:ComboFidSearchQueryAgent --exclude-context-in-krm-context-blocking False --include-knowledge-in-krm-context-blocking True --inject-query-string None --knowledge-response-control-token None --krm-beam-block-ngram 3 --krm-beam-context-block-ngram 3 --krm-beam-min-length 1 --krm-beam-size 3 --krm-doc-chunks-ranker woi_chunk_retrieved_docs --krm-inference beam --krm-message-mutators None --krm-model projects.seeker.agents.seeker:ComboFidSearchQueryAgent --krm-n-ranked-doc-chunks 1 --krm-rag-retriever-type search_engine --krm-search-query-generator-model-file '' --krm-search-server --loglevel debug --min-knowledge-length-when-search 10 --model projects.seeker.agents.seeker:SeekerAgent --model-file zoo:seeker/seeker_dialogue_3B/model --sdm-beam-block-ngram -1 --sdm-beam-min-length 1 --sdm-beam-size 1 --sdm-history-size 1 --sdm-inference greedy --sdm-model projects.seeker.agents.seeker:ComboFidSearchQueryAgent --search-decision always --search-decision-control-token __is-search-required__ --search-decision-do-search-reply __do-search__ --search-decision-dont-search-reply __do-not-search__ --search-query-control-token __generate-query__ --search-server None --sqm-beam-block-ngram -1 --sqm-beam-min-length 2 --sqm-beam-size 1 --sqm-inference beam --sqm-model projects.seeker.agents.seeker:ComboFidSearchQueryAgent

gen/seeker_lm

Generation parameters for SeeKeR, Language Model. SeeShuster et al. (2022)

--beam-disregard-knowledge-for-context-blocking True --datatype valid --drm-beam-block-full-context True --drm-beam-block-ngram 3 --drm-beam-context-block-ngram 3 --drm-beam-min-length 20 --drm-beam-size 10 --drm-inference beam --drm-message-mutators None --drm-model projects.seeker.agents.gpt2_seeker:GPT2ComboSearchQueryAgent --exclude-context-in-krm-context-blocking False --include-knowledge-in-krm-context-blocking True --inject-query-string None --knowledge-response-control-token None --krm-beam-block-ngram 3 --krm-beam-context-block-ngram -1 --krm-beam-min-length 1 --krm-beam-size 3 --krm-doc-chunks-ranker woi_chunk_retrieved_docs --krm-inference beam --krm-message-mutators None --krm-model projects.seeker.agents.gpt2_seeker:GPT2ComboSearchQueryAgent --krm-n-ranked-doc-chunks 1 --krm-rag-retriever-type search_engine --krm-search-query-generator-model-file '' --krm-search-server --loglevel debug --min-knowledge-length-when-search 1 --model projects.seeker.agents.gpt2_seeker:GPT2SeekerAgent --model-file zoo:seeker/seeker_lm_xl/model --sdm-beam-block-ngram -1 --sdm-beam-min-length 1 --sdm-beam-size 1 --sdm-history-size 1 --sdm-inference greedy --search-decision always --search-decision-control-token __is-search-required__ --search-decision-do-search-reply __do-search__ --search-decision-dont-search-reply __do-not-search__ --search-query-control-token __generate-query__ --search-server None --sqm-beam-block-ngram -1 --sqm-beam-min-length 2 --sqm-beam-size 1 --sqm-inference beam --sqm-model projects.seeker.agents.gpt2_seeker:GPT2ComboSearchQueryAgent