search

SAI990323 / TALLRec

Apache License 2.0
199 stars 32 forks source link

Question about 'sample' settings #40

Closed Shiruiruiwang closed 11 months ago

Shiruiruiwang commented 11 months ago

I noticed that you have set sample=64 in the shell file, which controls the training dataset only containing 64 instructions according to this code

train_data["train"] = train_data["train"].shuffle(seed=seed).select(range(sample)) if sample > -1 else train_data["train"].shuffle(seed=seed)

in https://github.com/SAI990323/TALLRec/blob/a10cdbcc13e47c9f86a669c3345825255664dc8b/finetune_rec.py#L212C5-L212C5

I am confused about why only using 64 samples of total around 20k samples (for book dataset)? Would this setting stay consistent for other comparison methods?

Shiruiruiwang commented 11 months ago

Sorry for my careless check in closed issues, now I have acquired the answer from #16 . Thanks for your excellent work!!