Tweet
Tweet

Slide 1

Slide 1 text

RedshiftML in Cookpad 2021/12/11 Redshift MLϋϯζΦϯ + re:Invent re:Cap Analytic s Cookpad R& D Yusuke Fukasawa

Slide 2

Slide 2 text

ΞδΣϯμ ● ࣗݾ঺հ ● Cookpad ʹ͓͚Δ RedshiftML ͷ࢖͍ํ ● Cookpad ʹ͓͚Δ RedshiftML Λ࢖͍͜ͳͨ͢Ίͷ޻෉

Slide 3

Slide 3 text

ࣗݾ঺հ ● Yusuke Fukasawa (@fufufukakaka ) ● ݚڀ։ൃ෦ɻςʔϒϧσʔλɾࣗવݴޠॲཧɾϨίϝϯυʹڵຯ͕͋Δ ● લ৬͸੨͍Rͷձࣾ ● ͜ͷҰ೥͸৽ଔΤϯδχΞ࠾༻୲౰΋݉೚

Slide 4

Slide 4 text

ΫοΫύουʹ͍ͭͯ https://speakerdeck.com/cookpadhr/cookpad-introduction

Slide 5

Slide 5 text

ΫοΫύουͰͷ R&D https://research.cookpad.com/

Slide 6

Slide 6 text

Cookpad ʹ͓͚Δ RedshiftML ͷ࢖͍ํ

Slide 7

Slide 7 text

Cookpad ͱ RedshiftML ● RedshiftML ͷ૝ఆσʔλ͸ςʔϒϧσʔλ(ߏ଄Խσʔλ ) ● ͕ɺΫοΫύουʹ͓͚Δ ML λεΫͷຆͲ͕ඇߏ଄Խσʔλ(ݴޠɾը ૾)ʹର͢Δ΋ͷͰ͋ΔͨΊɺexample case ʹͦͷ··Ԋ͑ͳ͍ ● ඇߏ଄Խσʔλ(ಛʹݴޠ)ʹରͯ͠ RedshiftML Λӡ༻Ͱ͖Δ͔Ͳ͏͔ ○ ͲͷΑ͏ͳλεΫΛ RedshiftML ʹҠ͍ͯ͘͠ͱྑͦ͞͏͔ʁ ● ·ͣɺϨγϐͷΧςΰϦ෼ྨύΠϓϥΠϯΛ RedshiftML ্Ͱ࣮ݱͰ͖Δ ͔ɺʹ͍ͭͯݕূΛ࣮ࢪ

Slide 8

Slide 8 text

ϨγϐΧςΰϦ෼ྨʹ͍ͭͯ ● ΫοΫύουͰ͸༷ʑͳ৔໘ͰϨγϐ σʔλʹର͢Δ෼ྨ͕ߦΘΕ͍ͯΔ ● e.g. MYϑΥϧμʹอଘͨ͠Ϩγϐʹ ͍ͭͯߜΓࠐΉػೳ ○ called asʮ͓·͔ͤ੔ཧʯ ● ϨγϐͷλΠτϧ͕ओͳೖྗͱͳΓɺ ֤ΧςΰϦ͝ͱͷೋ஋෼ྨث͕ଘࡏ

Slide 9

Slide 9 text

ϨγϐΧςΰϦ෼ྨʹ͍ͭͯ ● ೉қ౓ͷ௿͍λεΫ׌ͭ਺͕ଟ͍ͷ ͰɺSQL͚ͩͰ׬݁Ͱ͖Δ RedshiftML ͸ັྗతʂ

Slide 10

Slide 10 text

ϕΫλϥΠζ ௒ϔϧγʔʂαϯυΠον ● ϨγϐλΠτϧ͸ͦͷ·· RedshiftML ͷೖྗʹͰ͖ͳ͍ ● ݫີʹݴ͑͹ɺจࣈྻΛೖྗʹ͢Δ͜ͱ͸Ͱ͖·͕͢ɺςΩετʹಛԽ ͨ͠લॲཧ͕ͳ͍ ○ word2vecͰͷϕΫλϥΠζɺbag-of-words ͳͲ͸ͳ͍ ● ͳͷͰɺ༧ΊςΩετΛςʔϒϧσʔλʹม׵͢Δ ● Πϝʔδˣ 0.23, 0.45, 0.67, 0.91, ...

Slide 11

Slide 11 text

ϕΫλϥΠζ ● Tokenize By MeCab(ipadic) ● Vectorize By fasttext (trained by cookpad recipe title texts) ● 1 sentence → 100 dim vector ● ͜ΕΛ csv ʹ͠·͢ logger.info("Load latest fasttext") embeddings_model = fasttext.FastText.load_model("fasttext.model") logger.info("Write vectors to CSV.GZ") with gzip.open(output_file_name, "wb") as file: for id, segmented_text in zip(ids, segmented_texts): tokens = segmented_text.split() embed = np.zeros(embeddings_model.get_dimension()) for token in tokens: embed += embeddings_model[token] embed /= len(tokens) _embed: List[float] = embed.tolist() _embed.insert(0, id) file.write(",".join([str(dim) for dim in _embed]).encode("utf-8")) file.write("\n".encode("utf-8"))

Slide 12

Slide 12 text

ϕΫλϥΠζ ● ϕΫτϧͷͨΊͷςʔϒϧΛ࡞Γ ·͢ ● ࠓճ͸100࣍ݩͱͨ͠ͷͰɺΧϥ Ϝ͸100ݸ͋Γ·͢ CREATE TABLE research.recipe_title_vectors ( dim_0 float, dim_1 float, ..., dim_99 float, label int )

Slide 13

Slide 13 text

Vector ςʔϒϧ ● ઌఔ༻ҙͨ͠csvΛςʔϒϧʹྲྀ͠ࠐΈ·͢ ● ͜ΕͰςΩετʹର͢Δ෼ྨ͕Ͱ͖·͢

Slide 14

Slide 14 text

Ϟσϧֶश ● CREATE MODEL ͷඞཁͳه ड़ΛຒΊ·͢ ○ ઃఆΛ΄΅ॻ͔ͣʹ Auto Ͱͷ࣮ߦ΋Մೳ ● bread_prediction_examp l ● ͋ΔϨγϐ͕ύϯྉཧ͔ Ͳ͏͔ͷڭࢣσʔλΛ· ͱΊͨςʔϒϧ DROP MODEL IF EXISTS research.bread_prediction_model; CREATE MODEL research.bread_prediction_model from ( select label, dim_0, dim_1, ..., dim_99 from research.recipe_title_vectors as v join ( select recipe_id, label from research.bread_prediction_examples where is_train = TRUE ) as id_label on id_label.recipe_id = v.recipe_id ) TARGET label function redshiftml_fn_bread_prediction IAM_ROLE 'arn:aws:iam::xxxxxxxxxxxx:role/RedshiftSystemAccess' MODEL_TYPE XGBOOST PROBLEM_TYPE BINARY_CLASSIFICATION OBJECTIVE 'Accuracy' SETTINGS ( S3_BUCKET 'xxxxx' );

Slide 15

Slide 15 text

ਫ਼౓Λ֬ೝ ● ݱࡏ production Ͱಈ͍͍ͯΔϞσϧ͸୯ ޠස౓+SVCͰɺF1͕90 % ● ରͯ͠ RedshiftML(fasttext+xgboost)Ͱ͸ F1 95 % ● ਫ਼౓తʹ͸΄΅มΘΒͣ(एׯྑ͘ͳͬ ͨ)Ͱɺஔ͖׵͑ͯ΋໰୊ͳͦ͞͏Ͱ͋Δ ͜ͱΛ֬ೝ svc redshiftml

Slide 16

Slide 16 text

όονਤ ● ಉֶ࣌श਺ʹ4ͱ͍͏ ੍ݶ͕͋ΔͨΊɺ9ݸ ͷϞσϧΛ3ݸͣͭʹ ෼ֶ͚ͯशΛεέ δϡʔϦϯά ● ֶश͸ຖ݄ɺਪ࿦͸ ຖ೔࣮ࢪ

Slide 17

Slide 17 text

RedshiftML Λ࢖͍͜ͳͨ͢Ίͷ޻෉

Slide 18

Slide 18 text

RedshiftML ͷͰ͖ͳ͍͜ͱ ● ඇߏ଄Խσʔλʹର͢Δߴ౓ͳϞσϧ͸૊Ίͳ͍ ● ྫ͑͹ɺը૾ʹରͯ͠CNNΛ૊ΜͩΓɺςΩετʹରͯ͠຋༁Ϟσϧɺ NERϞσϧΛ૊ΜͩΓ ● ඇߏ଄ԽσʔλΛҰ౓ςʔϒϧܗࣜʹམͱ͞ͳ͚Ε͹͍͚ͳ͍࣌఺Ͱɺ Ͱ ͖Δ͜ͱ͸͔ͳΓ؆୯ͳϞσϧΛ૊Ή͜ͱʹݶఆ͞ΕΔ

Slide 19

Slide 19 text

RedshiftML ͷྑ͍ͱ͜Ζ ● ϞσϧΛ૊Ή࡞ۀ͕શͯ SQL Ͱ׬݁͢Δ ● ਪ࿦ίετ͕͔͔Βͳ্͍ʹ͔ͳΓૣ͍ ○ Ϩγϐ300ສ݅ʹର͢Δਪ࿦͕Ұ෼ఔ౓Ͱ׬ྃ ● Ϟσϧͷύϥϝʔλαʔν΋΍ͬͯ͘ΕΔ ● ֶशͨ͠Ϟσϧ͸୭Ͱ΋ SQL Λॻ͘͜ͱͰར༻Մೳ ○ grant จͰݖݶΛ։์͢Δ͜ͱ͕ඞཁ

Slide 20

Slide 20 text

RedshiftML ΛԿͱ͔ͯ͠࢖͍͜ͳ͢ ● RedshiftML Λຊ൪ӡ༻͢Δ্Ͱରॲͨ͜͠ͱ 1. Ϟσϧ͕σάϨʔγϣϯ͍ͯ͠Δ͔Ͳ͏͔͔֬ΊΔͨΊʹɺΦϑϥΠϯς ετͷϝτϦΫεΛ௥੻͍ͨ͠ 2. ਪ࿦όονΛಈ͔͢લʹʮRedshiftML ͷ͜ͷϞσϧΛ࢖͑Δ͔Ͳ͏͔ʯ Λ֬ೝͰ͖ΔΑ͏ʹ͍ͨ͠

Slide 21

Slide 21 text

RedshiftML ͷ ϝτϦΫεΛ௥੻͢Δ

Slide 22

Slide 22 text

RedshiftML Ϟσϧͷ ϝτϦΫε֬ೝ ● RedshiftML Ͱ͸ Optimize ʹࢦఆ ͨ͠ϝτϦΫε͔͠ SageMaker ଆ͔Βฦͬͯ͜ͳ͍ ● ͨͱ͑͹ Accuracy Λࢦඪͱͯ͠ AutoPilot ʹ౤͛ΔͱɺͦΕ͔͠ Θ͔Βͳ͍(ෳ਺ࢦఆ͸Ͱ͖ͳ͍) ͜Ε

Slide 23

Slide 23 text

RedshiftML ϞσϧͷϝτϦΫε֬ೝ ● ͜ͷ໰୊Λղܾ͢ΔͨΊʹ؆୯ͳΞϓϦΛॻ͖·ͨ͠ˣ

Slide 24

Slide 24 text

MetricsTracer ΋ͱ΋ͱ͸ ී௨ͷػցֶशϓϩδΣΫτ ͷΦϑϥΠϯςετͰͷ ϝτϦΫεΛه࿥ɾूܭ͢Δ ͨΊʹ࡞ͬͨ΋ͷ ࠓճɺRedshiftMLͷͨΊʹ গ֦͠ுͨ͠ NextJS + Chakra-UI

Slide 25

Slide 25 text

MetricsTracer: RedshiftMLϞσϧৄࡉ

Slide 26

Slide 26 text

MetricsTracer ͕΍Δ͜ͱ ● ֤ϞσϧΛҰཡͰ͖Δϖʔδ ● ֤Ϟσϧͷৄࡉ ○ ςετσʔλʹରͯ͠ͷϝτϦΫεΛ஝ੵɾදࣔ ○ Ϟσϧ͕࠶ֶश͞ΕͨΒͦΕΛݕ஌ͯ͠ϝτϦΫεΛܭࢉ͠ɺ஝ੵ͠·͢ ○ ͲΜͳSQLͰਪ࿦͕Ͱ͖Δ͔ɺαϯϓϧͷSQLΛදࣔ͠·͢ ● RedshiftML ͚ͩͰ͸ΧόʔͰ͖ͳ͍ʮΦϑϥΠϯςετͰͷϝτϦΫε௥੻ʯΛ࣮ݱ͠Α͏ͱ ͨࣾ͠಺πʔϧͰ͢ ○ RedshiftML ͷΦϑϥΠϯςετΛߦ͏όον͕ฒߦͯ͠ಈ͍͍ͯ·͢ ● ࠓͷͱ͜Ζ͸໰୊ͳ͘ಈ͍͍ͯ·͢

Slide 27

Slide 27 text

RedshiftML ͷΦϑϥΠϯςετ 1 . ఆظతʹϞσϧঢ়ଶΛ໰ ͍߹ΘͤΔ (2,3࣌ؒʹҰճ) 2 . Ϟσϧͷঢ়ଶΛฦ͢ 3 . ΋͠Ϟσϧͷঢ়ଶ͕લճ NOT READY Ͱ ৽ͨʹ READY ʹͳͬͨͷͰ͋Ε͹(retrain ) ςετσʔλʹରͯ͠ͷ༧ଌ݁ՌΛܭࢉ͢ΔΫΤ ϦΛ౤͛Δ 4 . ༧ଌ݁Ռ͔Β ϝτϦΫεΛܭࢉ͢Δ

Slide 28

Slide 28 text

RedshiftML ͷ ਪ࿦όονΛ੍ޚ͢Δ

Slide 29

Slide 29 text

RedshiftML Ϟσϧͷঢ়ଶ֬ೝ: લఏ ● ΫοΫύουʹ͸ศརͳ SQL όον࣮ߦج൫͕͋Δ ● Bricolageɾkuroko2 ͱ͍͏γεςϜΛհͯ͠ SQL ΛόονॲཧͰྲྀ͢͜ ͱʹؔͯ͠͸ຆͲखؒͳ࣮͘ߦՄೳ ● ͜Ε͸ RedshiftML ͱ͸ඇৗʹ૬ੑ͕Α͘ɺϞσϧֶश͔Βਪ࿦·Ͱ Bricolage ʹ৐ͤΔ͚ͩͰ׬݁ ● ͔͠͠ɺঢ়ଶ؅ཧΛߟ͑ΔͱɺҰखؒඞཁ

Slide 30

Slide 30 text

RedshiftML Ϟσϧͷঢ়ଶ֬ೝ ● CREATE MODEL จΛ࣮ߦ͢Δͱɺֶश͕੒ޭ͢Δ͠ͳ͍ͷҎલʹֶश σʔλΛ SageMaker ʹૹͬͨ࣌఺ͰͦͷΫΤϦ͸׬ྃ͠·͢ ● ඇಉظʹֶशδϣϒ͕ಈ͘ͷͰɺֶशδϣϒ͕׬ྃͨ͠Βશ݅ਪ࿦όον Λճ͢ʂͱ͍ͬͨґଘؔ܎Λ૊Ή͜ͱ͕গ͠೉͍͠Ͱ͢

Slide 31

Slide 31 text

RedshiftML Ϟσϧͷঢ়ଶ֬ೝ ● શ݅ਪ࿦όον͕ಈ͘௚લʹɺ୭͔͕ϞσϧֶशΛෆҙʹಈ͔ͨ͠Γ͢Δ ͱ(͋·Γͳ͍ͱࢥ͍·͕͢)ɺશ݅ਪ࿦όονʹεςʔλενΣοΫػߏ ͕ͳ͍৔߹ɺෆҙʹશ݅ਪ࿦όον͕མͪͯ͠·͏ɺͱ͍͏͜ͱʹͳΓ· ͢ ● ͜Ε͸ਫ਼ਆӴੜ্(΋ͪΖΜӡ༻্΋)͋·ΓΑ͋͘Γ·ͤΜ

Slide 32

Slide 32 text

RedshiftML Ϟσϧͷঢ়ଶ֬ೝ ● ͦ͜ͰɺRedshiftML ͷϞσϧঢ়ଶΛνΣοΫͯ͠OKͳΒਪ࿦όονΛಈ͔͢ɺͱ͍͏੍ޚΛ ߦ͏ΞϓϦέʔγϣϯΛॻ͖·ͨ͠ ● ʮRedshiftml-Batchʯ

Slide 33

Slide 33 text

Redshiftml-Batch ͕΍Δ͜ͱ ● StatusChecker.py

Slide 34

Slide 34 text

Redshiftml-Batch ͕΍Δ͜ͱ ● BatchRunner.py

Slide 35

Slide 35 text

Redshiftml-Batch ͕΍Δ͜ͱ ● run.py s3 s3

Slide 36

Slide 36 text

·ͱΊ

Slide 37

Slide 37 text

RedshiftML in Cookpad: ·ͱΊ ● Cookpad ͰͷϝΠϯλʔήοτ: ςΩετσʔλ΁ͷ׆༻ʹτϥΠ ● ಛ௃ྔ࡞੒ύΠϓϥΠϯͳͲ͕ผͰඞཁͰ͸͋Δ΋ͷͷɺظ଴Ҏ্ͷਫ਼౓ ͕ಘΒΕΔ͜ͱ͕Θ͔ͬͨ ● ҰํͰɺΦϑϥΠϯςετͰͷϝτϦΫε௥੻ɺਪ࿦όον࣮ߦ࣌ͷঢ়ଶ ֬ೝͳͲͷิ׬͢΂͖఺΋Θ͔ͬͨ ● Cookpad Ͱ͸ͦΕΛิ͏ࣾ಺πʔϧΛ࣮૷ͯ͠೔ʑͷӡ༻ʹ଱͑͏Δ඼࣭ Λ୲อ͠Α͏ͱ͍ͯ͠·͢