Model Parameters¶

AlphaFold 3 layer names, shapes, and dtypes are documented in the table below. This can be used for example to generate random parameters for AlphaFold 3 performance optimisation on new accelerators without having to obtain the official parameters. It is important to not generate zero-only parameters for performance optimisations as accelerators often have shortcuts for zero-only arguments (e.g. 0 * tensor can be optimised to a no-op).

Producing random parameters could be done similarly to the following snippet:

```py from alphafold3.model import params import numpy as np import zstandard

parameters = ... # Data from the parameters schema.

with zstandard.open('random_weights.bin.zst', 'wb') as compressed: for scope_name, shape, dtype in parameters: if scope_name == 'meta:identifier': # The identifier can be all zeros. arr = np.zeros(shape=shape, dtype=dtype) else: # Do not use all-zero params, instead sample uniformly between -1 and 1. arr = np.random.uniform(low=-1, high=1, size=shape).astype(dtype) scope_name = scope_name.split(':') compressed.write(params.encode_record(*scope_name, arr)) ```

Parameters Schema¶

name=__meta__:__identifier__ dtype=uint8 shape=(64,) name=diffuser/~/diffusion_head/diffusion_atom_features_layer_norm:scale dtype=float32 shape=(128,) name=diffuser/~/diffusion_head/diffusion_atom_features_to_position_update:weights dtype=float32 shape=(128, 3) name=diffuser/~/diffusion_head/diffusion_atom_positions_to_features:weights dtype=float32 shape=(3, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderadaptive_zero_cond:bias dtype=float32 shape=(3, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderadaptive_zero_cond:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderffw_adaptive_zero_cond:bias dtype=float32 shape=(3, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderffw_adaptive_zero_cond:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderffw_single_cond_bias:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderffw_single_cond_layer_norm:scale dtype=float32 shape=(3, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderffw_single_cond_scale:bias dtype=float32 shape=(3, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderffw_single_cond_scale:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderffw_transition1:weights dtype=float32 shape=(3, 128, 512) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderffw_transition2:weights dtype=float32 shape=(3, 256, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decodergating_query:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderk_projection:weights dtype=float32 shape=(3, 128, 4, 32) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderksingle_cond_bias:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderksingle_cond_layer_norm:scale dtype=float32 shape=(3, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderksingle_cond_scale:bias dtype=float32 shape=(3, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderksingle_cond_scale:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderq_projection:bias dtype=float32 shape=(3, 4, 32) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderq_projection:weights dtype=float32 shape=(3, 128, 4, 32) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderqsingle_cond_bias:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderqsingle_cond_layer_norm:scale dtype=float32 shape=(3, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderqsingle_cond_scale:bias dtype=float32 shape=(3, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderqsingle_cond_scale:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decodertransition2:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/__layer_stack_with_per_layer/diffusion_atom_transformer_decoderv_projection:weights dtype=float32 shape=(3, 128, 4, 32) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/pair_input_layer_norm:scale dtype=float32 shape=(16,) name=diffuser/~/diffusion_head/diffusion_atom_transformer_decoder/pair_logits_projection:weights dtype=float32 shape=(16, 3, 4) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderadaptive_zero_cond:bias dtype=float32 shape=(3, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderadaptive_zero_cond:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderffw_adaptive_zero_cond:bias dtype=float32 shape=(3, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderffw_adaptive_zero_cond:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderffw_single_cond_bias:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderffw_single_cond_layer_norm:scale dtype=float32 shape=(3, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderffw_single_cond_scale:bias dtype=float32 shape=(3, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderffw_single_cond_scale:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderffw_transition1:weights dtype=float32 shape=(3, 128, 512) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderffw_transition2:weights dtype=float32 shape=(3, 256, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encodergating_query:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderk_projection:weights dtype=float32 shape=(3, 128, 4, 32) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderksingle_cond_bias:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderksingle_cond_layer_norm:scale dtype=float32 shape=(3, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderksingle_cond_scale:bias dtype=float32 shape=(3, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderksingle_cond_scale:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderq_projection:bias dtype=float32 shape=(3, 4, 32) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderq_projection:weights dtype=float32 shape=(3, 128, 4, 32) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderqsingle_cond_bias:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderqsingle_cond_layer_norm:scale dtype=float32 shape=(3, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderqsingle_cond_scale:bias dtype=float32 shape=(3, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderqsingle_cond_scale:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encodertransition2:weights dtype=float32 shape=(3, 128, 128) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/__layer_stack_with_per_layer/diffusion_atom_transformer_encoderv_projection:weights dtype=float32 shape=(3, 128, 4, 32) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/pair_input_layer_norm:scale dtype=float32 shape=(16,) name=diffuser/~/diffusion_head/diffusion_atom_transformer_encoder/pair_logits_projection:weights dtype=float32 shape=(16, 3, 4) name=diffuser/~/diffusion_head/diffusion_embed_pair_distances_1:weights dtype=float32 shape=(1, 16) name=diffuser/~/diffusion_head/diffusion_embed_pair_distances:weights dtype=float32 shape=(1, 16) name=diffuser/~/diffusion_head/diffusion_embed_pair_offsets_1:weights dtype=float32 shape=(3, 16) name=diffuser/~/diffusion_head/diffusion_embed_pair_offsets_valid:weights dtype=float32 shape=(1, 16) name=diffuser/~/diffusion_head/diffusion_embed_pair_offsets:weights dtype=float32 shape=(3, 16) name=diffuser/~/diffusion_head/diffusion_embed_ref_atom_name:weights dtype=float32 shape=(256, 128) name=diffuser/~/diffusion_head/diffusion_embed_ref_charge:weights dtype=float32 shape=(1, 128) name=diffuser/~/diffusion_head/diffusion_embed_ref_element:weights dtype=float32 shape=(128, 128) name=diffuser/~/diffusion_head/diffusion_embed_ref_mask:weights dtype=float32 shape=(1, 128) name=diffuser/~/diffusion_head/diffusion_embed_ref_pos:weights dtype=float32 shape=(3, 128) name=diffuser/~/diffusion_head/diffusion_embed_trunk_pair_cond:weights dtype=float32 shape=(128, 16) name=diffuser/~/diffusion_head/diffusion_embed_trunk_single_cond:weights dtype=float32 shape=(384, 128) name=diffuser/~/diffusion_head/diffusion_lnorm_trunk_pair_cond:scale dtype=float32 shape=(128,) name=diffuser/~/diffusion_head/diffusion_lnorm_trunk_single_cond:scale dtype=float32 shape=(384,) name=diffuser/~/diffusion_head/diffusion_pair_mlp_1:weights dtype=float32 shape=(16, 16) name=diffuser/~/diffusion_head/diffusion_pair_mlp_2:weights dtype=float32 shape=(16, 16) name=diffuser/~/diffusion_head/diffusion_pair_mlp_3:weights dtype=float32 shape=(16, 16) name=diffuser/~/diffusion_head/diffusion_project_atom_features_for_aggr:weights dtype=float32 shape=(128, 768) name=diffuser/~/diffusion_head/diffusion_project_token_features_for_broadcast:weights dtype=float32 shape=(768, 128) name=diffuser/~/diffusion_head/diffusion_single_to_pair_cond_col_1:weights dtype=float32 shape=(128, 16) name=diffuser/~/diffusion_head/diffusion_single_to_pair_cond_col:weights dtype=float32 shape=(128, 16) name=diffuser/~/diffusion_head/diffusion_single_to_pair_cond_row_1:weights dtype=float32 shape=(128, 16) name=diffuser/~/diffusion_head/diffusion_single_to_pair_cond_row:weights dtype=float32 shape=(128, 16) name=diffuser/~/diffusion_head/noise_embedding_initial_norm:scale dtype=float32 shape=(256,) name=diffuser/~/diffusion_head/noise_embedding_initial_projection:weights dtype=float32 shape=(256, 384) name=diffuser/~/diffusion_head/output_norm:scale dtype=float32 shape=(768,) name=diffuser/~/diffusion_head/pair_cond_initial_norm:scale dtype=float32 shape=(267,) name=diffuser/~/diffusion_head/pair_cond_initial_projection:weights dtype=float32 shape=(267, 128) name=diffuser/~/diffusion_head/pair_transition_0ffw_layer_norm:offset dtype=float32 shape=(128,) name=diffuser/~/diffusion_head/pair_transition_0ffw_layer_norm:scale dtype=float32 shape=(128,) name=diffuser/~/diffusion_head/pair_transition_0ffw_transition1:weights dtype=float32 shape=(128, 512) name=diffuser/~/diffusion_head/pair_transition_0ffw_transition2:weights dtype=float32 shape=(256, 128) name=diffuser/~/diffusion_head/pair_transition_1ffw_layer_norm:offset dtype=float32 shape=(128,) name=diffuser/~/diffusion_head/pair_transition_1ffw_layer_norm:scale dtype=float32 shape=(128,) name=diffuser/~/diffusion_head/pair_transition_1ffw_transition1:weights dtype=float32 shape=(128, 512) name=diffuser/~/diffusion_head/pair_transition_1ffw_transition2:weights dtype=float32 shape=(256, 128) name=diffuser/~/diffusion_head/single_cond_embedding_norm:scale dtype=float32 shape=(384,) name=diffuser/~/diffusion_head/single_cond_embedding_projection:weights dtype=float32 shape=(384, 768) name=diffuser/~/diffusion_head/single_cond_initial_norm:scale dtype=float32 shape=(831,) name=diffuser/~/diffusion_head/single_cond_initial_projection:weights dtype=float32 shape=(831, 384) name=diffuser/~/diffusion_head/single_transition_0ffw_layer_norm:offset dtype=float32 shape=(384,) name=diffuser/~/diffusion_head/single_transition_0ffw_layer_norm:scale dtype=float32 shape=(384,) name=diffuser/~/diffusion_head/single_transition_0ffw_transition1:weights dtype=float32 shape=(384, 1536) name=diffuser/~/diffusion_head/single_transition_0ffw_transition2:weights dtype=float32 shape=(768, 384) name=diffuser/~/diffusion_head/single_transition_1ffw_layer_norm:offset dtype=float32 shape=(384,) name=diffuser/~/diffusion_head/single_transition_1ffw_layer_norm:scale dtype=float32 shape=(384,) name=diffuser/~/diffusion_head/single_transition_1ffw_transition1:weights dtype=float32 shape=(384, 1536) name=diffuser/~/diffusion_head/single_transition_1ffw_transition2:weights dtype=float32 shape=(768, 384) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformeradaptive_zero_cond:bias dtype=float32 shape=(6, 4, 768) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformeradaptive_zero_cond:weights dtype=float32 shape=(6, 4, 384, 768) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformerffw_adaptive_zero_cond:bias dtype=float32 shape=(6, 4, 768) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformerffw_adaptive_zero_cond:weights dtype=float32 shape=(6, 4, 384, 768) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformerffw_single_cond_bias:weights dtype=float32 shape=(6, 4, 384, 768) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformerffw_single_cond_layer_norm:scale dtype=float32 shape=(6, 4, 384) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformerffw_single_cond_scale:bias dtype=float32 shape=(6, 4, 768) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformerffw_single_cond_scale:weights dtype=float32 shape=(6, 4, 384, 768) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformerffw_transition1:weights dtype=float32 shape=(6, 4, 768, 3072) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformerffw_transition2:weights dtype=float32 shape=(6, 4, 1536, 768) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformergating_query:weights dtype=float32 shape=(6, 4, 768, 768) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformerk_projection:weights dtype=float32 shape=(6, 4, 768, 16, 48) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformerq_projection:bias dtype=float32 shape=(6, 4, 16, 48) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformerq_projection:weights dtype=float32 shape=(6, 4, 768, 16, 48) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformersingle_cond_bias:weights dtype=float32 shape=(6, 4, 384, 768) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformersingle_cond_layer_norm:scale dtype=float32 shape=(6, 4, 384) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformersingle_cond_scale:bias dtype=float32 shape=(6, 4, 768) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformersingle_cond_scale:weights dtype=float32 shape=(6, 4, 384, 768) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformertransition2:weights dtype=float32 shape=(6, 4, 768, 768) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/__layer_stack_with_per_layer/transformerv_projection:weights dtype=float32 shape=(6, 4, 768, 16, 48) name=diffuser/~/diffusion_head/transformer/__layer_stack_with_per_layer/pair_logits_projection:weights dtype=float32 shape=(6, 128, 4, 16) name=diffuser/~/diffusion_head/transformer/pair_input_layer_norm:scale dtype=float32 shape=(128,) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_attention1/act_norm:offset dtype=float32 shape=(4, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_attention1/act_norm:scale dtype=float32 shape=(4, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_attention1/gating_query:weights dtype=bfloat16 shape=(4, 128, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_attention1/k_projection:weights dtype=bfloat16 shape=(4, 4, 32, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_attention1/output_projection:weights dtype=bfloat16 shape=(4, 128, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_attention1/pair_bias_projection:weights dtype=bfloat16 shape=(4, 128, 4) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_attention1/q_projection:weights dtype=bfloat16 shape=(4, 4, 32, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_attention1/v_projection:weights dtype=bfloat16 shape=(4, 128, 4, 32) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_attention2/act_norm:offset dtype=float32 shape=(4, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_attention2/act_norm:scale dtype=float32 shape=(4, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_attention2/gating_query:weights dtype=bfloat16 shape=(4, 128, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_attention2/k_projection:weights dtype=bfloat16 shape=(4, 4, 32, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_attention2/output_projection:weights dtype=bfloat16 shape=(4, 128, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_attention2/pair_bias_projection:weights dtype=bfloat16 shape=(4, 128, 4) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_attention2/q_projection:weights dtype=bfloat16 shape=(4, 4, 32, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_attention2/v_projection:weights dtype=bfloat16 shape=(4, 128, 4, 32) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_transition/input_layer_norm:offset dtype=float32 shape=(4, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_transition/input_layer_norm:scale dtype=float32 shape=(4, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_transition/transition1:weights dtype=bfloat16 shape=(4, 128, 1024) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/pair_transition/transition2:weights dtype=bfloat16 shape=(4, 512, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/single_attention_gating_query:weights dtype=bfloat16 shape=(4, 384, 384) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/single_attention_k_projection:weights dtype=bfloat16 shape=(4, 384, 16, 24) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/single_attention_layer_norm:offset dtype=float32 shape=(4, 384) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/single_attention_layer_norm:scale dtype=float32 shape=(4, 384) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/single_attention_q_projection:bias dtype=bfloat16 shape=(4, 16, 24) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/single_attention_q_projection:weights dtype=bfloat16 shape=(4, 384, 16, 24) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/single_attention_transition2:weights dtype=bfloat16 shape=(4, 384, 384) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/single_attention_v_projection:weights dtype=bfloat16 shape=(4, 384, 16, 24) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/single_pair_logits_norm:offset dtype=float32 shape=(4, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/single_pair_logits_norm:scale dtype=float32 shape=(4, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/single_pair_logits_projection:weights dtype=bfloat16 shape=(4, 128, 16) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/single_transition/input_layer_norm:offset dtype=float32 shape=(4, 384) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/single_transition/input_layer_norm:scale dtype=float32 shape=(4, 384) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/single_transition/transition1:weights dtype=bfloat16 shape=(4, 384, 3072) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/single_transition/transition2:weights dtype=bfloat16 shape=(4, 1536, 384) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/triangle_multiplication_incoming/center_norm:offset dtype=float32 shape=(4, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/triangle_multiplication_incoming/center_norm:scale dtype=float32 shape=(4, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/triangle_multiplication_incoming/gate:weights dtype=bfloat16 shape=(4, 128, 256) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/triangle_multiplication_incoming/gating_linear:weights dtype=bfloat16 shape=(4, 128, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/triangle_multiplication_incoming/left_norm_input:offset dtype=float32 shape=(4, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/triangle_multiplication_incoming/left_norm_input:scale dtype=float32 shape=(4, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/triangle_multiplication_incoming/output_projection:weights dtype=bfloat16 shape=(4, 128, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/triangle_multiplication_incoming/projection:weights dtype=bfloat16 shape=(4, 128, 256) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/triangle_multiplication_outgoing/center_norm:offset dtype=float32 shape=(4, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/triangle_multiplication_outgoing/center_norm:scale dtype=float32 shape=(4, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/triangle_multiplication_outgoing/gate:weights dtype=bfloat16 shape=(4, 128, 256) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/triangle_multiplication_outgoing/gating_linear:weights dtype=bfloat16 shape=(4, 128, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/triangle_multiplication_outgoing/left_norm_input:offset dtype=float32 shape=(4, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/triangle_multiplication_outgoing/left_norm_input:scale dtype=float32 shape=(4, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/triangle_multiplication_outgoing/output_projection:weights dtype=bfloat16 shape=(4, 128, 128) name=diffuser/confidence_head/__layer_stack_no_per_layer/confidence_pairformer/triangle_multiplication_outgoing/projection:weights dtype=bfloat16 shape=(4, 128, 256) name=diffuser/confidence_head/~_embed_features/distogram_feat_project:weights dtype=bfloat16 shape=(39, 128) name=diffuser/confidence_head/~_embed_features/left_target_feat_project:weights dtype=bfloat16 shape=(447, 128) name=diffuser/confidence_head/~_embed_features/right_target_feat_project:weights dtype=bfloat16 shape=(447, 128) name=diffuser/confidence_head/experimentally_resolved_ln:offset dtype=float32 shape=(384,) name=diffuser/confidence_head/experimentally_resolved_ln:scale dtype=float32 shape=(384,) name=diffuser/confidence_head/experimentally_resolved_logits:weights dtype=float32 shape=(384, 24, 2) name=diffuser/confidence_head/left_half_distance_logits:weights dtype=float32 shape=(128, 64) name=diffuser/confidence_head/logits_ln:offset dtype=float32 shape=(128,) name=diffuser/confidence_head/logits_ln:scale dtype=float32 shape=(128,) name=diffuser/confidence_head/pae_logits_ln:offset dtype=float32 shape=(128,) name=diffuser/confidence_head/pae_logits_ln:scale dtype=float32 shape=(128,) name=diffuser/confidence_head/pae_logits:weights dtype=float32 shape=(128, 64) name=diffuser/confidence_head/plddt_logits_ln:offset dtype=float32 shape=(384,) name=diffuser/confidence_head/plddt_logits_ln:scale dtype=float32 shape=(384,) name=diffuser/confidence_head/plddt_logits:weights dtype=float32 shape=(384, 24, 50) name=diffuser/distogram_head/half_logits:weights dtype=float32 shape=(128, 64) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderadaptive_zero_cond:bias dtype=float32 shape=(3, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderadaptive_zero_cond:weights dtype=float32 shape=(3, 128, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderffw_adaptive_zero_cond:bias dtype=float32 shape=(3, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderffw_adaptive_zero_cond:weights dtype=float32 shape=(3, 128, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderffw_single_cond_bias:weights dtype=float32 shape=(3, 128, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderffw_single_cond_layer_norm:scale dtype=float32 shape=(3, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderffw_single_cond_scale:bias dtype=float32 shape=(3, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderffw_single_cond_scale:weights dtype=float32 shape=(3, 128, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderffw_transition1:weights dtype=float32 shape=(3, 128, 512) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderffw_transition2:weights dtype=float32 shape=(3, 256, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encodergating_query:weights dtype=float32 shape=(3, 128, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderk_projection:weights dtype=float32 shape=(3, 128, 4, 32) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderksingle_cond_bias:weights dtype=float32 shape=(3, 128, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderksingle_cond_layer_norm:scale dtype=float32 shape=(3, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderksingle_cond_scale:bias dtype=float32 shape=(3, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderksingle_cond_scale:weights dtype=float32 shape=(3, 128, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderq_projection:bias dtype=float32 shape=(3, 4, 32) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderq_projection:weights dtype=float32 shape=(3, 128, 4, 32) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderqsingle_cond_bias:weights dtype=float32 shape=(3, 128, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderqsingle_cond_layer_norm:scale dtype=float32 shape=(3, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderqsingle_cond_scale:bias dtype=float32 shape=(3, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderqsingle_cond_scale:weights dtype=float32 shape=(3, 128, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encodertransition2:weights dtype=float32 shape=(3, 128, 128) name=diffuser/evoformer_conditioning_atom_transformer_encoder/__layer_stack_with_per_layer/evoformer_conditioning_atom_transformer_encoderv_projection:weights dtype=float32 shape=(3, 128, 4, 32) name=diffuser/evoformer_conditioning_atom_transformer_encoder/pair_input_layer_norm:scale dtype=float32 shape=(16,) name=diffuser/evoformer_conditioning_atom_transformer_encoder/pair_logits_projection:weights dtype=float32 shape=(16, 3, 4) name=diffuser/evoformer_conditioning_embed_pair_distances_1:weights dtype=float32 shape=(1, 16) name=diffuser/evoformer_conditioning_embed_pair_distances:weights dtype=float32 shape=(1, 16) name=diffuser/evoformer_conditioning_embed_pair_offsets_1:weights dtype=float32 shape=(3, 16) name=diffuser/evoformer_conditioning_embed_pair_offsets_valid:weights dtype=float32 shape=(1, 16) name=diffuser/evoformer_conditioning_embed_pair_offsets:weights dtype=float32 shape=(3, 16) name=diffuser/evoformer_conditioning_embed_ref_atom_name:weights dtype=float32 shape=(256, 128) name=diffuser/evoformer_conditioning_embed_ref_charge:weights dtype=float32 shape=(1, 128) name=diffuser/evoformer_conditioning_embed_ref_element:weights dtype=float32 shape=(128, 128) name=diffuser/evoformer_conditioning_embed_ref_mask:weights dtype=float32 shape=(1, 128) name=diffuser/evoformer_conditioning_embed_ref_pos:weights dtype=float32 shape=(3, 128) name=diffuser/evoformer_conditioning_pair_mlp_1:weights dtype=float32 shape=(16, 16) name=diffuser/evoformer_conditioning_pair_mlp_2:weights dtype=float32 shape=(16, 16) name=diffuser/evoformer_conditioning_pair_mlp_3:weights dtype=float32 shape=(16, 16) name=diffuser/evoformer_conditioning_project_atom_features_for_aggr:weights dtype=float32 shape=(128, 384) name=diffuser/evoformer_conditioning_single_to_pair_cond_col_1:weights dtype=float32 shape=(128, 16) name=diffuser/evoformer_conditioning_single_to_pair_cond_col:weights dtype=float32 shape=(128, 16) name=diffuser/evoformer_conditioning_single_to_pair_cond_row_1:weights dtype=float32 shape=(128, 16) name=diffuser/evoformer_conditioning_single_to_pair_cond_row:weights dtype=float32 shape=(128, 16) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_attention1/act_norm:offset dtype=float32 shape=(48, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_attention1/act_norm:scale dtype=float32 shape=(48, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_attention1/gating_query:weights dtype=bfloat16 shape=(48, 128, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_attention1/k_projection:weights dtype=bfloat16 shape=(48, 4, 32, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_attention1/output_projection:weights dtype=bfloat16 shape=(48, 128, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_attention1/pair_bias_projection:weights dtype=bfloat16 shape=(48, 128, 4) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_attention1/q_projection:weights dtype=bfloat16 shape=(48, 4, 32, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_attention1/v_projection:weights dtype=bfloat16 shape=(48, 128, 4, 32) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_attention2/act_norm:offset dtype=float32 shape=(48, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_attention2/act_norm:scale dtype=float32 shape=(48, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_attention2/gating_query:weights dtype=bfloat16 shape=(48, 128, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_attention2/k_projection:weights dtype=bfloat16 shape=(48, 4, 32, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_attention2/output_projection:weights dtype=bfloat16 shape=(48, 128, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_attention2/pair_bias_projection:weights dtype=bfloat16 shape=(48, 128, 4) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_attention2/q_projection:weights dtype=bfloat16 shape=(48, 4, 32, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_attention2/v_projection:weights dtype=bfloat16 shape=(48, 128, 4, 32) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_transition/input_layer_norm:offset dtype=float32 shape=(48, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_transition/input_layer_norm:scale dtype=float32 shape=(48, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_transition/transition1:weights dtype=bfloat16 shape=(48, 128, 1024) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/pair_transition/transition2:weights dtype=bfloat16 shape=(48, 512, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/single_attention_gating_query:weights dtype=bfloat16 shape=(48, 384, 384) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/single_attention_k_projection:weights dtype=bfloat16 shape=(48, 384, 16, 24) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/single_attention_layer_norm:offset dtype=float32 shape=(48, 384) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/single_attention_layer_norm:scale dtype=float32 shape=(48, 384) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/single_attention_q_projection:bias dtype=bfloat16 shape=(48, 16, 24) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/single_attention_q_projection:weights dtype=bfloat16 shape=(48, 384, 16, 24) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/single_attention_transition2:weights dtype=bfloat16 shape=(48, 384, 384) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/single_attention_v_projection:weights dtype=bfloat16 shape=(48, 384, 16, 24) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/single_pair_logits_norm:offset dtype=float32 shape=(48, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/single_pair_logits_norm:scale dtype=float32 shape=(48, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/single_pair_logits_projection:weights dtype=bfloat16 shape=(48, 128, 16) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/single_transition/input_layer_norm:offset dtype=float32 shape=(48, 384) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/single_transition/input_layer_norm:scale dtype=float32 shape=(48, 384) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/single_transition/transition1:weights dtype=bfloat16 shape=(48, 384, 3072) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/single_transition/transition2:weights dtype=bfloat16 shape=(48, 1536, 384) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/triangle_multiplication_incoming/center_norm:offset dtype=float32 shape=(48, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/triangle_multiplication_incoming/center_norm:scale dtype=float32 shape=(48, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/triangle_multiplication_incoming/gate:weights dtype=bfloat16 shape=(48, 128, 256) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/triangle_multiplication_incoming/gating_linear:weights dtype=bfloat16 shape=(48, 128, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/triangle_multiplication_incoming/left_norm_input:offset dtype=float32 shape=(48, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/triangle_multiplication_incoming/left_norm_input:scale dtype=float32 shape=(48, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/triangle_multiplication_incoming/output_projection:weights dtype=bfloat16 shape=(48, 128, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/triangle_multiplication_incoming/projection:weights dtype=bfloat16 shape=(48, 128, 256) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/triangle_multiplication_outgoing/center_norm:offset dtype=float32 shape=(48, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/triangle_multiplication_outgoing/center_norm:scale dtype=float32 shape=(48, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/triangle_multiplication_outgoing/gate:weights dtype=bfloat16 shape=(48, 128, 256) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/triangle_multiplication_outgoing/gating_linear:weights dtype=bfloat16 shape=(48, 128, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/triangle_multiplication_outgoing/left_norm_input:offset dtype=float32 shape=(48, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/triangle_multiplication_outgoing/left_norm_input:scale dtype=float32 shape=(48, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/triangle_multiplication_outgoing/output_projection:weights dtype=bfloat16 shape=(48, 128, 128) name=diffuser/evoformer/__layer_stack_no_per_layer_1/trunk_pairformer/triangle_multiplication_outgoing/projection:weights dtype=bfloat16 shape=(48, 128, 256) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/msa_attention1/act_norm:offset dtype=float32 shape=(4, 64) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/msa_attention1/act_norm:scale dtype=float32 shape=(4, 64) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/msa_attention1/gating_query:weights dtype=bfloat16 shape=(4, 64, 64) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/msa_attention1/output_projection:weights dtype=bfloat16 shape=(4, 64, 64) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/msa_attention1/pair_logits:weights dtype=bfloat16 shape=(4, 128, 8) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/msa_attention1/pair_norm:offset dtype=float32 shape=(4, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/msa_attention1/pair_norm:scale dtype=float32 shape=(4, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/msa_attention1/v_projection:weights dtype=bfloat16 shape=(4, 64, 8, 8) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/msa_transition/input_layer_norm:offset dtype=float32 shape=(4, 64) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/msa_transition/input_layer_norm:scale dtype=float32 shape=(4, 64) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/msa_transition/transition1:weights dtype=bfloat16 shape=(4, 64, 512) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/msa_transition/transition2:weights dtype=bfloat16 shape=(4, 256, 64) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/outer_product_mean:output_b dtype=bfloat16 shape=(4, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/outer_product_mean:output_w dtype=bfloat16 shape=(4, 32, 32, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/outer_product_mean/layer_norm_input:offset dtype=float32 shape=(4, 64) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/outer_product_mean/layer_norm_input:scale dtype=float32 shape=(4, 64) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/outer_product_mean/left_projection:weights dtype=bfloat16 shape=(4, 64, 32) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/outer_product_mean/right_projection:weights dtype=bfloat16 shape=(4, 64, 32) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_attention1/act_norm:offset dtype=float32 shape=(4, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_attention1/act_norm:scale dtype=float32 shape=(4, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_attention1/gating_query:weights dtype=bfloat16 shape=(4, 128, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_attention1/k_projection:weights dtype=bfloat16 shape=(4, 4, 32, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_attention1/output_projection:weights dtype=bfloat16 shape=(4, 128, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_attention1/pair_bias_projection:weights dtype=bfloat16 shape=(4, 128, 4) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_attention1/q_projection:weights dtype=bfloat16 shape=(4, 4, 32, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_attention1/v_projection:weights dtype=bfloat16 shape=(4, 128, 4, 32) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_attention2/act_norm:offset dtype=float32 shape=(4, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_attention2/act_norm:scale dtype=float32 shape=(4, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_attention2/gating_query:weights dtype=bfloat16 shape=(4, 128, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_attention2/k_projection:weights dtype=bfloat16 shape=(4, 4, 32, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_attention2/output_projection:weights dtype=bfloat16 shape=(4, 128, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_attention2/pair_bias_projection:weights dtype=bfloat16 shape=(4, 128, 4) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_attention2/q_projection:weights dtype=bfloat16 shape=(4, 4, 32, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_attention2/v_projection:weights dtype=bfloat16 shape=(4, 128, 4, 32) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_transition/input_layer_norm:offset dtype=float32 shape=(4, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_transition/input_layer_norm:scale dtype=float32 shape=(4, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_transition/transition1:weights dtype=bfloat16 shape=(4, 128, 1024) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/pair_transition/transition2:weights dtype=bfloat16 shape=(4, 512, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/triangle_multiplication_incoming/center_norm:offset dtype=float32 shape=(4, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/triangle_multiplication_incoming/center_norm:scale dtype=float32 shape=(4, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/triangle_multiplication_incoming/gate:weights dtype=bfloat16 shape=(4, 128, 256) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/triangle_multiplication_incoming/gating_linear:weights dtype=bfloat16 shape=(4, 128, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/triangle_multiplication_incoming/left_norm_input:offset dtype=float32 shape=(4, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/triangle_multiplication_incoming/left_norm_input:scale dtype=float32 shape=(4, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/triangle_multiplication_incoming/output_projection:weights dtype=bfloat16 shape=(4, 128, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/triangle_multiplication_incoming/projection:weights dtype=bfloat16 shape=(4, 128, 256) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/triangle_multiplication_outgoing/center_norm:offset dtype=float32 shape=(4, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/triangle_multiplication_outgoing/center_norm:scale dtype=float32 shape=(4, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/triangle_multiplication_outgoing/gate:weights dtype=bfloat16 shape=(4, 128, 256) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/triangle_multiplication_outgoing/gating_linear:weights dtype=bfloat16 shape=(4, 128, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/triangle_multiplication_outgoing/left_norm_input:offset dtype=float32 shape=(4, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/triangle_multiplication_outgoing/left_norm_input:scale dtype=float32 shape=(4, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/triangle_multiplication_outgoing/output_projection:weights dtype=bfloat16 shape=(4, 128, 128) name=diffuser/evoformer/__layer_stack_no_per_layer/msa_stack/triangle_multiplication_outgoing/projection:weights dtype=bfloat16 shape=(4, 128, 256) name=diffuser/evoformer/~_relative_encoding/position_activations:weights dtype=bfloat16 shape=(139, 128) name=diffuser/evoformer/bond_embedding:weights dtype=bfloat16 shape=(1, 128) name=diffuser/evoformer/extra_msa_target_feat:weights dtype=bfloat16 shape=(447, 64) name=diffuser/evoformer/left_single:weights dtype=bfloat16 shape=(447, 128) name=diffuser/evoformer/msa_activations:weights dtype=bfloat16 shape=(34, 64) name=diffuser/evoformer/prev_embedding_layer_norm:offset dtype=float32 shape=(128,) name=diffuser/evoformer/prev_embedding_layer_norm:scale dtype=float32 shape=(128,) name=diffuser/evoformer/prev_embedding:weights dtype=bfloat16 shape=(128, 128) name=diffuser/evoformer/prev_single_embedding_layer_norm:offset dtype=float32 shape=(384,) name=diffuser/evoformer/prev_single_embedding_layer_norm:scale dtype=float32 shape=(384,) name=diffuser/evoformer/prev_single_embedding:weights dtype=bfloat16 shape=(384, 384) name=diffuser/evoformer/right_single:weights dtype=bfloat16 shape=(447, 128) name=diffuser/evoformer/single_activations:weights dtype=bfloat16 shape=(447, 384) name=diffuser/evoformer/template_embedding/output_linear:weights dtype=bfloat16 shape=(64, 128) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_attention1/act_norm:offset dtype=float32 shape=(2, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_attention1/act_norm:scale dtype=float32 shape=(2, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_attention1/gating_query:weights dtype=bfloat16 shape=(2, 64, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_attention1/k_projection:weights dtype=bfloat16 shape=(2, 4, 16, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_attention1/output_projection:weights dtype=bfloat16 shape=(2, 64, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_attention1/pair_bias_projection:weights dtype=bfloat16 shape=(2, 64, 4) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_attention1/q_projection:weights dtype=bfloat16 shape=(2, 4, 16, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_attention1/v_projection:weights dtype=bfloat16 shape=(2, 64, 4, 16) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_attention2/act_norm:offset dtype=float32 shape=(2, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_attention2/act_norm:scale dtype=float32 shape=(2, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_attention2/gating_query:weights dtype=bfloat16 shape=(2, 64, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_attention2/k_projection:weights dtype=bfloat16 shape=(2, 4, 16, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_attention2/output_projection:weights dtype=bfloat16 shape=(2, 64, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_attention2/pair_bias_projection:weights dtype=bfloat16 shape=(2, 64, 4) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_attention2/q_projection:weights dtype=bfloat16 shape=(2, 4, 16, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_attention2/v_projection:weights dtype=bfloat16 shape=(2, 64, 4, 16) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_transition/input_layer_norm:offset dtype=float32 shape=(2, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_transition/input_layer_norm:scale dtype=float32 shape=(2, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_transition/transition1:weights dtype=bfloat16 shape=(2, 64, 256) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/pair_transition/transition2:weights dtype=bfloat16 shape=(2, 128, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/triangle_multiplication_incoming/center_norm:offset dtype=float32 shape=(2, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/triangle_multiplication_incoming/center_norm:scale dtype=float32 shape=(2, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/triangle_multiplication_incoming/gate:weights dtype=bfloat16 shape=(2, 64, 128) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/triangle_multiplication_incoming/gating_linear:weights dtype=bfloat16 shape=(2, 64, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/triangle_multiplication_incoming/left_norm_input:offset dtype=float32 shape=(2, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/triangle_multiplication_incoming/left_norm_input:scale dtype=float32 shape=(2, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/triangle_multiplication_incoming/output_projection:weights dtype=bfloat16 shape=(2, 64, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/triangle_multiplication_incoming/projection:weights dtype=bfloat16 shape=(2, 64, 128) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/triangle_multiplication_outgoing/center_norm:offset dtype=float32 shape=(2, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/triangle_multiplication_outgoing/center_norm:scale dtype=float32 shape=(2, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/triangle_multiplication_outgoing/gate:weights dtype=bfloat16 shape=(2, 64, 128) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/triangle_multiplication_outgoing/gating_linear:weights dtype=bfloat16 shape=(2, 64, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/triangle_multiplication_outgoing/left_norm_input:offset dtype=float32 shape=(2, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/triangle_multiplication_outgoing/left_norm_input:scale dtype=float32 shape=(2, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/triangle_multiplication_outgoing/output_projection:weights dtype=bfloat16 shape=(2, 64, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/__layer_stack_no_per_layer/template_embedding_iteration/triangle_multiplication_outgoing/projection:weights dtype=bfloat16 shape=(2, 64, 128) name=diffuser/evoformer/template_embedding/single_template_embedding/output_layer_norm:offset dtype=float32 shape=(64,) name=diffuser/evoformer/template_embedding/single_template_embedding/output_layer_norm:scale dtype=float32 shape=(64,) name=diffuser/evoformer/template_embedding/single_template_embedding/query_embedding_norm:offset dtype=float32 shape=(128,) name=diffuser/evoformer/template_embedding/single_template_embedding/query_embedding_norm:scale dtype=float32 shape=(128,) name=diffuser/evoformer/template_embedding/single_template_embedding/template_pair_embedding_0:weights dtype=bfloat16 shape=(39, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/template_pair_embedding_1:weights dtype=bfloat16 shape=(64,) name=diffuser/evoformer/template_embedding/single_template_embedding/template_pair_embedding_2:weights dtype=bfloat16 shape=(31, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/template_pair_embedding_3:weights dtype=bfloat16 shape=(31, 64) name=diffuser/evoformer/template_embedding/single_template_embedding/template_pair_embedding_4:weights dtype=bfloat16 shape=(64,) name=diffuser/evoformer/template_embedding/single_template_embedding/template_pair_embedding_5:weights dtype=bfloat16 shape=(64,) name=diffuser/evoformer/template_embedding/single_template_embedding/template_pair_embedding_6:weights dtype=bfloat16 shape=(64,) name=diffuser/evoformer/template_embedding/single_template_embedding/template_pair_embedding_7:weights dtype=bfloat16 shape=(64,) name=diffuser/evoformer/template_embedding/single_template_embedding/template_pair_embedding_8:weights dtype=bfloat16 shape=(128, 64)