more fixes for distilled loras

Author: prishajain1

src/maxdiffusion/loaders/lora_conversion_utils.py

@@ -768,6 +768,24 @@ def translate_ltx2_nnx_path_to_diffusers_lora(nnx_path_str, scan_layers=False):
       "av_cross_attn_audio_v2a_gate.linear": "diffusion_model.av_ca_v2a_gate_adaln_single.linear",
       "av_cross_attn_audio_scale_shift.linear": "diffusion_model.av_ca_audio_scale_shift_adaln_single.linear",
       "av_cross_attn_video_scale_shift.linear": "diffusion_model.av_ca_video_scale_shift_adaln_single.linear",
+      
+      # Nested conditioning layers
+      "time_embed.emb.timestep_embedder.linear_1": "diffusion_model.adaln_single.emb.timestep_embedder.linear_1",
+      "time_embed.emb.timestep_embedder.linear_2": "diffusion_model.adaln_single.emb.timestep_embedder.linear_2",
+      "audio_time_embed.emb.timestep_embedder.linear_1": "diffusion_model.audio_adaln_single.emb.timestep_embedder.linear_1",
+      "audio_time_embed.emb.timestep_embedder.linear_2": "diffusion_model.audio_adaln_single.emb.timestep_embedder.linear_2",
+      "av_cross_attn_video_scale_shift.emb.timestep_embedder.linear_1": "diffusion_model.av_ca_video_scale_shift_adaln_single.emb.timestep_embedder.linear_1",
+      "av_cross_attn_video_scale_shift.emb.timestep_embedder.linear_2": "diffusion_model.av_ca_video_scale_shift_adaln_single.emb.timestep_embedder.linear_2",
+      "av_cross_attn_audio_scale_shift.emb.timestep_embedder.linear_1": "diffusion_model.av_ca_audio_scale_shift_adaln_single.emb.timestep_embedder.linear_1",
+      "av_cross_attn_audio_scale_shift.emb.timestep_embedder.linear_2": "diffusion_model.av_ca_audio_scale_shift_adaln_single.emb.timestep_embedder.linear_2",
+      "av_cross_attn_video_a2v_gate.emb.timestep_embedder.linear_1": "diffusion_model.av_ca_a2v_gate_adaln_single.emb.timestep_embedder.linear_1",
+      "av_cross_attn_video_a2v_gate.emb.timestep_embedder.linear_2": "diffusion_model.av_ca_a2v_gate_adaln_single.emb.timestep_embedder.linear_2",
+      "av_cross_attn_audio_v2a_gate.emb.timestep_embedder.linear_1": "diffusion_model.av_ca_v2a_gate_adaln_single.emb.timestep_embedder.linear_1",
+      "av_cross_attn_audio_v2a_gate.emb.timestep_embedder.linear_2": "diffusion_model.av_ca_v2a_gate_adaln_single.emb.timestep_embedder.linear_2",
+      "caption_projection.linear_1": "diffusion_model.caption_projection.linear_1",
+      "caption_projection.linear_2": "diffusion_model.caption_projection.linear_2",
+      "audio_caption_projection.linear_1": "diffusion_model.audio_caption_projection.linear_1",
+      "audio_caption_projection.linear_2": "diffusion_model.audio_caption_projection.linear_2",
   }
 
   if nnx_path_str in global_map: