CodeFlow-HumanML3D / metadata /best_metrics.json
AmberJar's picture
Upload CodeFlow HumanML3D release weights
09154c8 verified
{
"best_fid": {
"checkpoint": "/iridisfs/scratch/pf2m24/projects/Umdd/momask-codes/checkpoints/t2m/codeflow_part_structured_newvqtop3_w150_p192_h1152_d6s12_drop005_w0_b64_lr1e4_e600_eval10from0_seed42_test_h200_blossom_g1_20260601/model/best_fid.pt",
"epoch": 300,
"fid": 0.05680781579934546,
"selection_config": {
"best_checkpoint_limit": 3,
"cond_scale": 6.0,
"dataset_name": "t2m",
"decode_mode": "nearest",
"eval_steps": 96,
"full_eval_seed": 42,
"latent_norm_mode": "codebook",
"latent_offset": 0.0,
"repeat_times": 1,
"sampling_method": "ode",
"sampling_schedule": "uniform",
"sde_gamma": 0.0,
"split": "test",
"terminal_mode": "tied_logits",
"unit_length": 4,
"weight_source": "ema"
},
"step": 114900,
"top3": 0.8678879310344828
},
"best_fid_top3": [
{
"checkpoint": "/iridisfs/scratch/pf2m24/projects/Umdd/momask-codes/checkpoints/t2m/codeflow_part_structured_newvqtop3_w150_p192_h1152_d6s12_drop005_w0_b64_lr1e4_e600_eval10from0_seed42_test_h200_blossom_g1_20260601/model/best_fid.pt",
"epoch": 300,
"fid": 0.05680781579934546,
"selection_config": {
"best_checkpoint_limit": 3,
"cond_scale": 6.0,
"dataset_name": "t2m",
"decode_mode": "nearest",
"eval_steps": 96,
"full_eval_seed": 42,
"latent_norm_mode": "codebook",
"latent_offset": 0.0,
"repeat_times": 1,
"sampling_method": "ode",
"sampling_schedule": "uniform",
"sde_gamma": 0.0,
"split": "test",
"terminal_mode": "tied_logits",
"unit_length": 4,
"weight_source": "ema"
},
"step": 114900,
"top3": 0.8678879310344828
},
{
"checkpoint": "/iridisfs/scratch/pf2m24/projects/Umdd/momask-codes/checkpoints/t2m/codeflow_part_structured_newvqtop3_w150_p192_h1152_d6s12_drop005_w0_b64_lr1e4_e600_eval10from0_seed42_test_h200_blossom_g1_20260601/model/best_fid_rank2.pt",
"epoch": 290,
"fid": 0.058189920626574576,
"selection_config": {
"best_checkpoint_limit": 3,
"cond_scale": 6.0,
"dataset_name": "t2m",
"decode_mode": "nearest",
"eval_steps": 96,
"full_eval_seed": 42,
"latent_norm_mode": "codebook",
"latent_offset": 0.0,
"repeat_times": 1,
"sampling_method": "ode",
"sampling_schedule": "uniform",
"sde_gamma": 0.0,
"split": "test",
"terminal_mode": "tied_logits",
"unit_length": 4,
"weight_source": "ema"
},
"step": 111070,
"top3": 0.8730603448275862
},
{
"checkpoint": "/iridisfs/scratch/pf2m24/projects/Umdd/momask-codes/checkpoints/t2m/codeflow_part_structured_newvqtop3_w150_p192_h1152_d6s12_drop005_w0_b64_lr1e4_e600_eval10from0_seed42_test_h200_blossom_g1_20260601/model/best_fid_rank3.pt",
"epoch": 320,
"fid": 0.06242718631276034,
"selection_config": {
"best_checkpoint_limit": 3,
"cond_scale": 6.0,
"dataset_name": "t2m",
"decode_mode": "nearest",
"eval_steps": 96,
"full_eval_seed": 42,
"latent_norm_mode": "codebook",
"latent_offset": 0.0,
"repeat_times": 1,
"sampling_method": "ode",
"sampling_schedule": "uniform",
"sde_gamma": 0.0,
"split": "test",
"terminal_mode": "tied_logits",
"unit_length": 4,
"weight_source": "ema"
},
"step": 122560,
"top3": 0.8633620689655173
}
],
"best_top3": {
"checkpoint": "/iridisfs/scratch/pf2m24/projects/Umdd/momask-codes/checkpoints/t2m/codeflow_part_structured_newvqtop3_w150_p192_h1152_d6s12_drop005_w0_b64_lr1e4_e600_eval10from0_seed42_test_h200_blossom_g1_20260601/model/best_top3.pt",
"epoch": 290,
"fid": 0.058189920626574576,
"selection_config": {
"best_checkpoint_limit": 3,
"cond_scale": 6.0,
"dataset_name": "t2m",
"decode_mode": "nearest",
"eval_steps": 96,
"full_eval_seed": 42,
"latent_norm_mode": "codebook",
"latent_offset": 0.0,
"repeat_times": 1,
"sampling_method": "ode",
"sampling_schedule": "uniform",
"sde_gamma": 0.0,
"split": "test",
"terminal_mode": "tied_logits",
"unit_length": 4,
"weight_source": "ema"
},
"step": 111070,
"top3": 0.8730603448275862
},
"best_top3_top3": [
{
"checkpoint": "/iridisfs/scratch/pf2m24/projects/Umdd/momask-codes/checkpoints/t2m/codeflow_part_structured_newvqtop3_w150_p192_h1152_d6s12_drop005_w0_b64_lr1e4_e600_eval10from0_seed42_test_h200_blossom_g1_20260601/model/best_top3.pt",
"epoch": 290,
"fid": 0.058189920626574576,
"selection_config": {
"best_checkpoint_limit": 3,
"cond_scale": 6.0,
"dataset_name": "t2m",
"decode_mode": "nearest",
"eval_steps": 96,
"full_eval_seed": 42,
"latent_norm_mode": "codebook",
"latent_offset": 0.0,
"repeat_times": 1,
"sampling_method": "ode",
"sampling_schedule": "uniform",
"sde_gamma": 0.0,
"split": "test",
"terminal_mode": "tied_logits",
"unit_length": 4,
"weight_source": "ema"
},
"step": 111070,
"top3": 0.8730603448275862
},
{
"checkpoint": "/iridisfs/scratch/pf2m24/projects/Umdd/momask-codes/checkpoints/t2m/codeflow_part_structured_newvqtop3_w150_p192_h1152_d6s12_drop005_w0_b64_lr1e4_e600_eval10from0_seed42_test_h200_blossom_g1_20260601/model/best_top3_rank2.pt",
"epoch": 140,
"fid": 0.1756792861862806,
"selection_config": {
"best_checkpoint_limit": 3,
"cond_scale": 6.0,
"dataset_name": "t2m",
"decode_mode": "nearest",
"eval_steps": 96,
"full_eval_seed": 42,
"latent_norm_mode": "codebook",
"latent_offset": 0.0,
"repeat_times": 1,
"sampling_method": "ode",
"sampling_schedule": "uniform",
"sde_gamma": 0.0,
"split": "test",
"terminal_mode": "tied_logits",
"unit_length": 4,
"weight_source": "ema"
},
"step": 53620,
"top3": 0.8726293103448276
},
{
"checkpoint": "/iridisfs/scratch/pf2m24/projects/Umdd/momask-codes/checkpoints/t2m/codeflow_part_structured_newvqtop3_w150_p192_h1152_d6s12_drop005_w0_b64_lr1e4_e600_eval10from0_seed42_test_h200_blossom_g1_20260601/model/best_top3_rank3.pt",
"epoch": 120,
"fid": 0.20938363776417646,
"selection_config": {
"best_checkpoint_limit": 3,
"cond_scale": 6.0,
"dataset_name": "t2m",
"decode_mode": "nearest",
"eval_steps": 96,
"full_eval_seed": 42,
"latent_norm_mode": "codebook",
"latent_offset": 0.0,
"repeat_times": 1,
"sampling_method": "ode",
"sampling_schedule": "uniform",
"sde_gamma": 0.0,
"split": "test",
"terminal_mode": "tied_logits",
"unit_length": 4,
"weight_source": "ema"
},
"step": 45960,
"top3": 0.8726293103448276
}
],
"selection_config": {
"best_checkpoint_limit": 3,
"cond_scale": 6.0,
"dataset_name": "t2m",
"decode_mode": "nearest",
"eval_steps": 96,
"full_eval_seed": 42,
"latent_norm_mode": "codebook",
"latent_offset": 0.0,
"repeat_times": 1,
"sampling_method": "ode",
"sampling_schedule": "uniform",
"sde_gamma": 0.0,
"split": "test",
"terminal_mode": "tied_logits",
"unit_length": 4,
"weight_source": "ema"
}
}