Evaluation Metrics

[MaxyLee]

Hi, there. I'm replicating DM-GAN on CUB but got bad results(R-Precision: 57.07(±0.71); FID: 25.68) after training 700 epochs. Besides, I got even worse results(R-Precision: 35.45(±0.64); FID: 42.57) using the checkpoint(bird_DMGAN.pth) you provided. By way of generation, I generated 30,000 images using the scripts provided. Can you please help me with this issue?
btw, I have no access to the inception model for IS evaluation on google drive, can you please provide another download link?
Thanks!
Here is my config:

CONFIG_NAME: 'DMGAN'

DATASET_NAME: 'birds'
DATA_DIR: '../data/birds'
GPU_ID: 7
WORKERS: 1

B_VALIDATION: True  # True  # False
TREE:
    BRANCH_NUM: 3


TRAIN:
    FLAG: False
    NET_G: '../models/bird_DMGAN.pth'
    B_NET_D: False
    BATCH_SIZE: 10
    NET_E: '../DAMSMencoders/bird/text_encoder200.pth'


GAN:
    DF_DIM: 32
    GF_DIM: 64
    Z_DIM: 100
    R_NUM: 2

TEXT:
    EMBEDDING_DIM: 256
    CAPTIONS_PER_IMAGE: 10
    WORDS_NUM: 25