Partial diffusion with sequences does not work

Hi,

I am trying to partially diffuse a complex, with the config below. However, this fails at the 45th line of run_inference.py (sampler = iu.sampler_selector(conf)). What could be the problem?

Thank you for the help!

These were the command line inputs:
./RFdiffusion/run_inference.py inference.output_prefix=outputs/inference.num_designs=1 inference.input_pdb=input.pdb diffuser.partial_T=20 'contigmap.provide_seq=[0-177]' ppi.hotspot_res=[A123,A126,A130] 'contigmap.contigs=[178-178 80-80]' inference.dump_pdb=True inference.dump_pdb_path='/dev/shm'

{'inference': {'input_pdb': 'input.pdb', 'num_designs': 1, 'design_startnum': 0, 'ckpt_override_path': None, 'symmetry': None, 'recenter': True, 'radius': 10.0, 'model_only_neighbors': False, 'output_prefix': 'outputs/', 'write_trajectory': True, 'scaffold_guided': False, 'model_runner': 'SelfConditioning', 'cautious': True, 'align_motif': True, 'symmetric_self_cond': True, 'final_step': 1, 'deterministic': False, 'trb_save_ckpt_path': None, 'dump_pdb': True, 'dump_pdb_path': '/dev/shm'}, 'contigmap': {'contigs': ['A1-178 80-80'], 'inpaint_seq': None, 'provide_seq': ['0-177'], 'length': None}, 'model': {'n_extra_block': 4, 'n_main_block': 32, 'n_ref_block': 4, 'd_msa': 256, 'd_msa_full': 64, 'd_pair': 128, 'd_templ': 64, 'n_head_msa': 8, 'n_head_pair': 4, 'n_head_templ': 4, 'd_hidden': 32, 'd_hidden_templ': 32, 'p_drop': 0.15, 'SE3_param_full': {'num_layers': 1, 'num_channels': 32, 'num_degrees': 2, 'n_heads': 4, 'div': 4, 'l0_in_features': 8, 'l0_out_features': 8, 'l1_in_features': 3, 'l1_out_features': 2, 'num_edge_features': 32}, 'SE3_param_topk': {'num_layers': 1, 'num_channels': 32, 'num_degrees': 2, 'n_heads': 4, 'div': 4, 'l0_in_features': 64, 'l0_out_features': 64, 'l1_in_features': 3, 'l1_out_features': 2, 'num_edge_features': 64}, 'd_time_emb': None, 'd_time_emb_proj': None, 'freeze_track_motif': False, 'use_motif_timestep': False}, 'diffuser': {'T': 50, 'b_0': 0.01, 'b_T': 0.07, 'schedule_type': 'linear', 'so3_type': 'igso3', 'crd_scale': 0.25, 'partial_T': 20, 'so3_schedule_type': 'linear', 'min_b': 1.5, 'max_b': 2.5, 'min_sigma': 0.02, 'max_sigma': 1.5}, 'denoiser': {'noise_scale_ca': 1, 'final_noise_scale_ca': 1, 'ca_noise_schedule_type': 'constant', 'noise_scale_frame': 1, 'final_noise_scale_frame': 1, 'frame_noise_schedule_type': 'constant'}, 'ppi': {'hotspot_res': ['A123', 'A126', 'A130']}, 'potentials': {'guiding_potentials': None, 'guide_scale': 10, 'guide_decay': 'constant', 'olig_inter_all': None, 'olig_intra_all': None, 'olig_custom_contact': None, 'substrate': None}, 'contig_settings': {'ref_idx': None, 'hal_idx': None, 'idx_rf': None, 'inpaint_seq_tensor': None}, 'preprocess': {'sidechain_input': False, 'motif_sidechain_input': True, 'd_t1d': 22, 'd_t2d': 44, 'prob_self_cond': 0.0, 'str_self_cond': False, 'predict_previous': False}, 'logging': {'inputs': False}, 'scaffoldguided': {'scaffoldguided': False, 'target_pdb': False, 'target_path': None, 'scaffold_list': None, 'scaffold_dir': None, 'sampled_insertion': 0, 'sampled_N': 0, 'sampled_C': 0, 'ss_mask': 0, 'systematic': False, 'target_ss': None, 'target_adj': None, 'mask_loops': True, 'contig_crop': None}}

Start with run_inference.py at line 45, where sampler_selector(conf) is called, and reproduce the command using the supplied partial_T, provide_seq, contigs, and hotspot settings. The report does not include the exception or traceback, so capture that first and inspect the resulting configuration. Done means the partial-diffusion sequence command proceeds past sampler selection or produces a clearly diagnosed configuration error.