Open JR-Wang opened 7 months ago
It seems the .yml file put in 'options' dir isn't the exact file you use. I downloaded the span.zip you provided and found in SPANX4_CN48/test_AttLi_x4_20231123_15530.log
network_g:[ type: AttLiP18 num_in_ch: 3 num_out_ch: 3 feature_channels: 48 upscale: 4 img_range: 255.0 rgb_mean: [0.4488, 0.4371, 0.404] ] network_t:[ type: SRFormer upscale: 4 in_chans: 3 img_size: 64 window_size: 22 img_range: 1.0 depths: [6, 6, 6, 6, 6, 6] embed_dim: 180 num_heads: [6, 6, 6, 6, 6, 6] mlp_ratio: 2 upsampler: pixelshuffle resi_connection: 1conv ]
but still I didn't find class AttLiP18 or class SRFormer in code file. Which should I do to reproduce your performance?
Agree to add one, I also encountered this problem
Just load the model using the model class defined in span_arch.py. No yml needed.
It seems the .yml file put in 'options' dir isn't the exact file you use. I downloaded the span.zip you provided and found in SPANX4_CN48/test_AttLi_x4_20231123_15530.log
network_g:[ type: AttLiP18 num_in_ch: 3 num_out_ch: 3 feature_channels: 48 upscale: 4 img_range: 255.0 rgb_mean: [0.4488, 0.4371, 0.404] ] network_t:[ type: SRFormer upscale: 4 in_chans: 3 img_size: 64 window_size: 22 img_range: 1.0 depths: [6, 6, 6, 6, 6, 6] embed_dim: 180 num_heads: [6, 6, 6, 6, 6, 6] mlp_ratio: 2 upsampler: pixelshuffle resi_connection: 1conv ]
but still I didn't find class AttLiP18 or class SRFormer in code file. Which should I do to reproduce your performance?