jennyxue's picture
4bit build of HyperFlow H3 Flash
072644e verified
Raw History Blame Contribute Delete
21.7 kB
{
"_class_name": "HyperFlowH3FlashTransformer3DModel",
"_diffusers_version": "0.36.0.dev0",
"num_attention_heads": 56,
"attention_head_dim": 128,
"hidden_size": 5376,
"num_layers": 50,
"num_refiner_layers": 2,
"ffn_dim": 14336,
"in_channels": 24,
"audio_in_channels": 32,
"patch_size": [
1,
2,
2
],
"text_dim": 5120,
"freq_dim": 256,
"time_embed_hidden_dim": 5376,
"time_embed_dim": 16,
"rope_freq_dim": 16,
"rope_theta": 10000.0,
"norm_eps": 1e-05,
"qk_norm_eps": 1e-05,
"final_norm_eps": 1e-05,
"hyperflow_gate": 0.25,
"lowrank_ranks": {
"transformer_blocks.1.attn.to_k": 704,
"transformer_blocks.1.attn.to_out.0": 1792,
"transformer_blocks.1.attn.to_q": 768,
"transformer_blocks.1.attn.to_v": 896,
"transformer_blocks.1.ff.net.0.proj": 1664,
"transformer_blocks.1.ff.net.2": 2496,
"transformer_blocks.10.attn.to_k": 320,
"transformer_blocks.10.attn.to_out.0": 448,
"transformer_blocks.10.attn.to_q": 320,
"transformer_blocks.10.attn.to_v": 448,
"transformer_blocks.10.ff.net.0.proj": 384,
"transformer_blocks.10.ff.net.2": 1536,
"transformer_blocks.11.attn.to_k": 256,
"transformer_blocks.11.attn.to_out.0": 448,
"transformer_blocks.11.attn.to_q": 320,
"transformer_blocks.11.attn.to_v": 448,
"transformer_blocks.11.ff.net.0.proj": 384,
"transformer_blocks.11.ff.net.2": 512,
"transformer_blocks.12.attn.to_k": 192,
"transformer_blocks.12.attn.to_out.0": 256,
"transformer_blocks.12.attn.to_q": 192,
"transformer_blocks.12.attn.to_v": 320,
"transformer_blocks.12.ff.net.0.proj": 256,
"transformer_blocks.12.ff.net.2": 1088,
"transformer_blocks.13.attn.to_k": 576,
"transformer_blocks.13.attn.to_out.0": 704,
"transformer_blocks.13.attn.to_q": 576,
"transformer_blocks.13.attn.to_v": 832,
"transformer_blocks.13.ff.net.0.proj": 768,
"transformer_blocks.13.ff.net.2": 64,
"transformer_blocks.14.attn.to_k": 256,
"transformer_blocks.14.attn.to_out.0": 384,
"transformer_blocks.14.attn.to_q": 320,
"transformer_blocks.14.attn.to_v": 448,
"transformer_blocks.14.ff.net.0.proj": 320,
"transformer_blocks.14.ff.net.2": 1600,
"transformer_blocks.15.attn.to_k": 256,
"transformer_blocks.15.attn.to_out.0": 448,
"transformer_blocks.15.attn.to_q": 320,
"transformer_blocks.15.attn.to_v": 448,
"transformer_blocks.15.ff.net.0.proj": 320,
"transformer_blocks.15.ff.net.2": 1600,
"transformer_blocks.16.attn.to_k": 512,
"transformer_blocks.16.attn.to_out.0": 768,
"transformer_blocks.16.attn.to_q": 512,
"transformer_blocks.16.attn.to_v": 704,
"transformer_blocks.16.ff.net.0.proj": 704,
"transformer_blocks.16.ff.net.2": 2304,
"transformer_blocks.17.attn.to_k": 256,
"transformer_blocks.17.attn.to_out.0": 448,
"transformer_blocks.17.attn.to_q": 320,
"transformer_blocks.17.attn.to_v": 448,
"transformer_blocks.17.ff.net.0.proj": 384,
"transformer_blocks.17.ff.net.2": 1536,
"transformer_blocks.18.attn.to_k": 256,
"transformer_blocks.18.attn.to_out.0": 448,
"transformer_blocks.18.attn.to_q": 256,
"transformer_blocks.18.attn.to_v": 448,
"transformer_blocks.18.ff.net.0.proj": 448,
"transformer_blocks.18.ff.net.2": 1472,
"transformer_blocks.2.attn.to_k": 1600,
"transformer_blocks.2.attn.to_out.0": 2112,
"transformer_blocks.2.attn.to_q": 1600,
"transformer_blocks.2.attn.to_v": 1664,
"transformer_blocks.2.ff.net.0.proj": 3200,
"transformer_blocks.20.attn.to_k": 832,
"transformer_blocks.20.attn.to_out.0": 1216,
"transformer_blocks.20.attn.to_q": 832,
"transformer_blocks.20.attn.to_v": 1152,
"transformer_blocks.20.ff.net.0.proj": 1344,
"transformer_blocks.20.ff.net.2": 3072,
"transformer_blocks.23.attn.to_k": 128,
"transformer_blocks.23.attn.to_out.0": 320,
"transformer_blocks.23.attn.to_q": 128,
"transformer_blocks.23.attn.to_v": 256,
"transformer_blocks.23.ff.net.0.proj": 256,
"transformer_blocks.23.ff.net.2": 1152,
"transformer_blocks.4.attn.to_k": 192,
"transformer_blocks.4.attn.to_out.0": 256,
"transformer_blocks.4.attn.to_q": 192,
"transformer_blocks.4.attn.to_v": 320,
"transformer_blocks.4.ff.net.0.proj": 384,
"transformer_blocks.4.ff.net.2": 832,
"transformer_blocks.49.ff.net.2": 2944,
"transformer_blocks.7.attn.to_k": 192,
"transformer_blocks.7.attn.to_out.0": 320,
"transformer_blocks.7.attn.to_q": 256,
"transformer_blocks.7.attn.to_v": 384,
"transformer_blocks.7.ff.net.0.proj": 320,
"transformer_blocks.7.ff.net.2": 896,
"transformer_blocks.8.attn.to_k": 192,
"transformer_blocks.8.attn.to_out.0": 256,
"transformer_blocks.8.attn.to_q": 192,
"transformer_blocks.8.attn.to_v": 256,
"transformer_blocks.8.ff.net.0.proj": 256,
"transformer_blocks.8.ff.net.2": 1152,
"transformer_blocks.9.attn.to_k": 192,
"transformer_blocks.9.attn.to_out.0": 320,
"transformer_blocks.9.attn.to_q": 192,
"transformer_blocks.9.attn.to_v": 320,
"transformer_blocks.9.ff.net.0.proj": 192,
"transformer_blocks.9.ff.net.2": 1152
},
"dropped_blocks": [
6,
21,
22
],
"source": {
"repo": "VideoRebirthAI/hyperflow-h3-flash",
"revision": "c8f74e481b9f7743cd91d2d3c4bf108eeb723748",
"folder": "fl2va_int8_convrot",
"quantization": {
"method": "int8_convrot",
"group": 256
}
},
"mlx_quantization": {
"group_size": 64,
"bits": 4,
"linears": [
"token_refiner.refiner_blocks.0.attn.to_k",
"token_refiner.refiner_blocks.0.attn.to_out.0",
"token_refiner.refiner_blocks.0.attn.to_q",
"token_refiner.refiner_blocks.0.attn.to_v",
"token_refiner.refiner_blocks.0.ff.net.0.proj",
"token_refiner.refiner_blocks.0.ff.net.2",
"token_refiner.refiner_blocks.1.attn.to_k",
"token_refiner.refiner_blocks.1.attn.to_out.0",
"token_refiner.refiner_blocks.1.attn.to_q",
"token_refiner.refiner_blocks.1.attn.to_v",
"token_refiner.refiner_blocks.1.ff.net.0.proj",
"token_refiner.refiner_blocks.1.ff.net.2",
"transformer_blocks.0.attn.to_k",
"transformer_blocks.0.attn.to_out.0",
"transformer_blocks.0.attn.to_q",
"transformer_blocks.0.attn.to_v",
"transformer_blocks.0.ff.net.0.proj",
"transformer_blocks.0.ff.net.2",
"transformer_blocks.1.attn.to_k.down",
"transformer_blocks.1.attn.to_k.up",
"transformer_blocks.1.attn.to_out.0.down",
"transformer_blocks.1.attn.to_out.0.up",
"transformer_blocks.1.attn.to_q.down",
"transformer_blocks.1.attn.to_q.up",
"transformer_blocks.1.attn.to_v.down",
"transformer_blocks.1.attn.to_v.up",
"transformer_blocks.1.ff.net.0.proj.down",
"transformer_blocks.1.ff.net.0.proj.up",
"transformer_blocks.1.ff.net.2.down",
"transformer_blocks.1.ff.net.2.up",
"transformer_blocks.10.attn.to_k.down",
"transformer_blocks.10.attn.to_k.up",
"transformer_blocks.10.attn.to_out.0.down",
"transformer_blocks.10.attn.to_out.0.up",
"transformer_blocks.10.attn.to_q.down",
"transformer_blocks.10.attn.to_q.up",
"transformer_blocks.10.attn.to_v.down",
"transformer_blocks.10.attn.to_v.up",
"transformer_blocks.10.ff.net.0.proj.down",
"transformer_blocks.10.ff.net.0.proj.up",
"transformer_blocks.10.ff.net.2.down",
"transformer_blocks.10.ff.net.2.up",
"transformer_blocks.11.attn.to_k.down",
"transformer_blocks.11.attn.to_k.up",
"transformer_blocks.11.attn.to_out.0.down",
"transformer_blocks.11.attn.to_out.0.up",
"transformer_blocks.11.attn.to_q.down",
"transformer_blocks.11.attn.to_q.up",
"transformer_blocks.11.attn.to_v.down",
"transformer_blocks.11.attn.to_v.up",
"transformer_blocks.11.ff.net.0.proj.down",
"transformer_blocks.11.ff.net.0.proj.up",
"transformer_blocks.11.ff.net.2.down",
"transformer_blocks.11.ff.net.2.up",
"transformer_blocks.12.attn.to_k.down",
"transformer_blocks.12.attn.to_k.up",
"transformer_blocks.12.attn.to_out.0.down",
"transformer_blocks.12.attn.to_out.0.up",
"transformer_blocks.12.attn.to_q.down",
"transformer_blocks.12.attn.to_q.up",
"transformer_blocks.12.attn.to_v.down",
"transformer_blocks.12.attn.to_v.up",
"transformer_blocks.12.ff.net.0.proj.down",
"transformer_blocks.12.ff.net.0.proj.up",
"transformer_blocks.12.ff.net.2.down",
"transformer_blocks.12.ff.net.2.up",
"transformer_blocks.13.attn.to_k.down",
"transformer_blocks.13.attn.to_k.up",
"transformer_blocks.13.attn.to_out.0.down",
"transformer_blocks.13.attn.to_out.0.up",
"transformer_blocks.13.attn.to_q.down",
"transformer_blocks.13.attn.to_q.up",
"transformer_blocks.13.attn.to_v.down",
"transformer_blocks.13.attn.to_v.up",
"transformer_blocks.13.ff.net.0.proj.down",
"transformer_blocks.13.ff.net.0.proj.up",
"transformer_blocks.13.ff.net.2.down",
"transformer_blocks.13.ff.net.2.up",
"transformer_blocks.14.attn.to_k.down",
"transformer_blocks.14.attn.to_k.up",
"transformer_blocks.14.attn.to_out.0.down",
"transformer_blocks.14.attn.to_out.0.up",
"transformer_blocks.14.attn.to_q.down",
"transformer_blocks.14.attn.to_q.up",
"transformer_blocks.14.attn.to_v.down",
"transformer_blocks.14.attn.to_v.up",
"transformer_blocks.14.ff.net.0.proj.down",
"transformer_blocks.14.ff.net.0.proj.up",
"transformer_blocks.14.ff.net.2.down",
"transformer_blocks.14.ff.net.2.up",
"transformer_blocks.15.attn.to_k.down",
"transformer_blocks.15.attn.to_k.up",
"transformer_blocks.15.attn.to_out.0.down",
"transformer_blocks.15.attn.to_out.0.up",
"transformer_blocks.15.attn.to_q.down",
"transformer_blocks.15.attn.to_q.up",
"transformer_blocks.15.attn.to_v.down",
"transformer_blocks.15.attn.to_v.up",
"transformer_blocks.15.ff.net.0.proj.down",
"transformer_blocks.15.ff.net.0.proj.up",
"transformer_blocks.15.ff.net.2.down",
"transformer_blocks.15.ff.net.2.up",
"transformer_blocks.16.attn.to_k.down",
"transformer_blocks.16.attn.to_k.up",
"transformer_blocks.16.attn.to_out.0.down",
"transformer_blocks.16.attn.to_out.0.up",
"transformer_blocks.16.attn.to_q.down",
"transformer_blocks.16.attn.to_q.up",
"transformer_blocks.16.attn.to_v.down",
"transformer_blocks.16.attn.to_v.up",
"transformer_blocks.16.ff.net.0.proj.down",
"transformer_blocks.16.ff.net.0.proj.up",
"transformer_blocks.16.ff.net.2.down",
"transformer_blocks.16.ff.net.2.up",
"transformer_blocks.17.attn.to_k.down",
"transformer_blocks.17.attn.to_k.up",
"transformer_blocks.17.attn.to_out.0.down",
"transformer_blocks.17.attn.to_out.0.up",
"transformer_blocks.17.attn.to_q.down",
"transformer_blocks.17.attn.to_q.up",
"transformer_blocks.17.attn.to_v.down",
"transformer_blocks.17.attn.to_v.up",
"transformer_blocks.17.ff.net.0.proj.down",
"transformer_blocks.17.ff.net.0.proj.up",
"transformer_blocks.17.ff.net.2.down",
"transformer_blocks.17.ff.net.2.up",
"transformer_blocks.18.attn.to_k.down",
"transformer_blocks.18.attn.to_k.up",
"transformer_blocks.18.attn.to_out.0.down",
"transformer_blocks.18.attn.to_out.0.up",
"transformer_blocks.18.attn.to_q.down",
"transformer_blocks.18.attn.to_q.up",
"transformer_blocks.18.attn.to_v.down",
"transformer_blocks.18.attn.to_v.up",
"transformer_blocks.18.ff.net.0.proj.down",
"transformer_blocks.18.ff.net.0.proj.up",
"transformer_blocks.18.ff.net.2.down",
"transformer_blocks.18.ff.net.2.up",
"transformer_blocks.19.attn.to_k",
"transformer_blocks.19.attn.to_out.0",
"transformer_blocks.19.attn.to_q",
"transformer_blocks.19.attn.to_v",
"transformer_blocks.19.ff.net.0.proj",
"transformer_blocks.19.ff.net.2",
"transformer_blocks.2.attn.to_k.down",
"transformer_blocks.2.attn.to_k.up",
"transformer_blocks.2.attn.to_out.0.down",
"transformer_blocks.2.attn.to_out.0.up",
"transformer_blocks.2.attn.to_q.down",
"transformer_blocks.2.attn.to_q.up",
"transformer_blocks.2.attn.to_v.down",
"transformer_blocks.2.attn.to_v.up",
"transformer_blocks.2.ff.net.0.proj.down",
"transformer_blocks.2.ff.net.0.proj.up",
"transformer_blocks.2.ff.net.2",
"transformer_blocks.20.attn.to_k.down",
"transformer_blocks.20.attn.to_k.up",
"transformer_blocks.20.attn.to_out.0.down",
"transformer_blocks.20.attn.to_out.0.up",
"transformer_blocks.20.attn.to_q.down",
"transformer_blocks.20.attn.to_q.up",
"transformer_blocks.20.attn.to_v.down",
"transformer_blocks.20.attn.to_v.up",
"transformer_blocks.20.ff.net.0.proj.down",
"transformer_blocks.20.ff.net.0.proj.up",
"transformer_blocks.20.ff.net.2.down",
"transformer_blocks.20.ff.net.2.up",
"transformer_blocks.23.attn.to_k.down",
"transformer_blocks.23.attn.to_k.up",
"transformer_blocks.23.attn.to_out.0.down",
"transformer_blocks.23.attn.to_out.0.up",
"transformer_blocks.23.attn.to_q.down",
"transformer_blocks.23.attn.to_q.up",
"transformer_blocks.23.attn.to_v.down",
"transformer_blocks.23.attn.to_v.up",
"transformer_blocks.23.ff.net.0.proj.down",
"transformer_blocks.23.ff.net.0.proj.up",
"transformer_blocks.23.ff.net.2.down",
"transformer_blocks.23.ff.net.2.up",
"transformer_blocks.24.attn.to_k",
"transformer_blocks.24.attn.to_out.0",
"transformer_blocks.24.attn.to_q",
"transformer_blocks.24.attn.to_v",
"transformer_blocks.24.ff.net.0.proj",
"transformer_blocks.24.ff.net.2",
"transformer_blocks.25.attn.to_k",
"transformer_blocks.25.attn.to_out.0",
"transformer_blocks.25.attn.to_q",
"transformer_blocks.25.attn.to_v",
"transformer_blocks.25.ff.net.0.proj",
"transformer_blocks.25.ff.net.2",
"transformer_blocks.26.attn.to_k",
"transformer_blocks.26.attn.to_out.0",
"transformer_blocks.26.attn.to_q",
"transformer_blocks.26.attn.to_v",
"transformer_blocks.26.ff.net.0.proj",
"transformer_blocks.26.ff.net.2",
"transformer_blocks.27.attn.to_k",
"transformer_blocks.27.attn.to_out.0",
"transformer_blocks.27.attn.to_q",
"transformer_blocks.27.attn.to_v",
"transformer_blocks.27.ff.net.0.proj",
"transformer_blocks.27.ff.net.2",
"transformer_blocks.28.attn.to_k",
"transformer_blocks.28.attn.to_out.0",
"transformer_blocks.28.attn.to_q",
"transformer_blocks.28.attn.to_v",
"transformer_blocks.28.ff.net.0.proj",
"transformer_blocks.28.ff.net.2",
"transformer_blocks.29.attn.to_k",
"transformer_blocks.29.attn.to_out.0",
"transformer_blocks.29.attn.to_q",
"transformer_blocks.29.attn.to_v",
"transformer_blocks.29.ff.net.0.proj",
"transformer_blocks.29.ff.net.2",
"transformer_blocks.3.attn.to_k",
"transformer_blocks.3.attn.to_out.0",
"transformer_blocks.3.attn.to_q",
"transformer_blocks.3.attn.to_v",
"transformer_blocks.3.ff.net.0.proj",
"transformer_blocks.3.ff.net.2",
"transformer_blocks.30.attn.to_k",
"transformer_blocks.30.attn.to_out.0",
"transformer_blocks.30.attn.to_q",
"transformer_blocks.30.attn.to_v",
"transformer_blocks.30.ff.net.0.proj",
"transformer_blocks.30.ff.net.2",
"transformer_blocks.31.attn.to_k",
"transformer_blocks.31.attn.to_out.0",
"transformer_blocks.31.attn.to_q",
"transformer_blocks.31.attn.to_v",
"transformer_blocks.31.ff.net.0.proj",
"transformer_blocks.31.ff.net.2",
"transformer_blocks.32.attn.to_k",
"transformer_blocks.32.attn.to_out.0",
"transformer_blocks.32.attn.to_q",
"transformer_blocks.32.attn.to_v",
"transformer_blocks.32.ff.net.0.proj",
"transformer_blocks.32.ff.net.2",
"transformer_blocks.33.attn.to_k",
"transformer_blocks.33.attn.to_out.0",
"transformer_blocks.33.attn.to_q",
"transformer_blocks.33.attn.to_v",
"transformer_blocks.33.ff.net.0.proj",
"transformer_blocks.33.ff.net.2",
"transformer_blocks.34.attn.to_k",
"transformer_blocks.34.attn.to_out.0",
"transformer_blocks.34.attn.to_q",
"transformer_blocks.34.attn.to_v",
"transformer_blocks.34.ff.net.0.proj",
"transformer_blocks.34.ff.net.2",
"transformer_blocks.35.attn.to_k",
"transformer_blocks.35.attn.to_out.0",
"transformer_blocks.35.attn.to_q",
"transformer_blocks.35.attn.to_v",
"transformer_blocks.35.ff.net.0.proj",
"transformer_blocks.35.ff.net.2",
"transformer_blocks.36.attn.to_k",
"transformer_blocks.36.attn.to_out.0",
"transformer_blocks.36.attn.to_q",
"transformer_blocks.36.attn.to_v",
"transformer_blocks.36.ff.net.0.proj",
"transformer_blocks.36.ff.net.2",
"transformer_blocks.37.attn.to_k",
"transformer_blocks.37.attn.to_out.0",
"transformer_blocks.37.attn.to_q",
"transformer_blocks.37.attn.to_v",
"transformer_blocks.37.ff.net.0.proj",
"transformer_blocks.37.ff.net.2",
"transformer_blocks.38.attn.to_k",
"transformer_blocks.38.attn.to_out.0",
"transformer_blocks.38.attn.to_q",
"transformer_blocks.38.attn.to_v",
"transformer_blocks.38.ff.net.0.proj",
"transformer_blocks.38.ff.net.2",
"transformer_blocks.39.attn.to_k",
"transformer_blocks.39.attn.to_out.0",
"transformer_blocks.39.attn.to_q",
"transformer_blocks.39.attn.to_v",
"transformer_blocks.39.ff.net.0.proj",
"transformer_blocks.39.ff.net.2",
"transformer_blocks.4.attn.to_k.down",
"transformer_blocks.4.attn.to_k.up",
"transformer_blocks.4.attn.to_out.0.down",
"transformer_blocks.4.attn.to_out.0.up",
"transformer_blocks.4.attn.to_q.down",
"transformer_blocks.4.attn.to_q.up",
"transformer_blocks.4.attn.to_v.down",
"transformer_blocks.4.attn.to_v.up",
"transformer_blocks.4.ff.net.0.proj.down",
"transformer_blocks.4.ff.net.0.proj.up",
"transformer_blocks.4.ff.net.2.down",
"transformer_blocks.4.ff.net.2.up",
"transformer_blocks.40.attn.to_k",
"transformer_blocks.40.attn.to_out.0",
"transformer_blocks.40.attn.to_q",
"transformer_blocks.40.attn.to_v",
"transformer_blocks.40.ff.net.0.proj",
"transformer_blocks.40.ff.net.2",
"transformer_blocks.41.attn.to_k",
"transformer_blocks.41.attn.to_out.0",
"transformer_blocks.41.attn.to_q",
"transformer_blocks.41.attn.to_v",
"transformer_blocks.41.ff.net.0.proj",
"transformer_blocks.41.ff.net.2",
"transformer_blocks.42.attn.to_k",
"transformer_blocks.42.attn.to_out.0",
"transformer_blocks.42.attn.to_q",
"transformer_blocks.42.attn.to_v",
"transformer_blocks.42.ff.net.0.proj",
"transformer_blocks.42.ff.net.2",
"transformer_blocks.43.attn.to_k",
"transformer_blocks.43.attn.to_out.0",
"transformer_blocks.43.attn.to_q",
"transformer_blocks.43.attn.to_v",
"transformer_blocks.43.ff.net.0.proj",
"transformer_blocks.43.ff.net.2",
"transformer_blocks.44.attn.to_k",
"transformer_blocks.44.attn.to_out.0",
"transformer_blocks.44.attn.to_q",
"transformer_blocks.44.attn.to_v",
"transformer_blocks.44.ff.net.0.proj",
"transformer_blocks.44.ff.net.2",
"transformer_blocks.45.attn.to_k",
"transformer_blocks.45.attn.to_out.0",
"transformer_blocks.45.attn.to_q",
"transformer_blocks.45.attn.to_v",
"transformer_blocks.45.ff.net.0.proj",
"transformer_blocks.45.ff.net.2",
"transformer_blocks.46.attn.to_k",
"transformer_blocks.46.attn.to_out.0",
"transformer_blocks.46.attn.to_q",
"transformer_blocks.46.attn.to_v",
"transformer_blocks.46.ff.net.0.proj",
"transformer_blocks.46.ff.net.2",
"transformer_blocks.47.attn.to_k",
"transformer_blocks.47.attn.to_out.0",
"transformer_blocks.47.attn.to_q",
"transformer_blocks.47.attn.to_v",
"transformer_blocks.47.ff.net.0.proj",
"transformer_blocks.47.ff.net.2",
"transformer_blocks.48.attn.to_k",
"transformer_blocks.48.attn.to_out.0",
"transformer_blocks.48.attn.to_q",
"transformer_blocks.48.attn.to_v",
"transformer_blocks.48.ff.net.0.proj",
"transformer_blocks.48.ff.net.2",
"transformer_blocks.49.attn.to_k",
"transformer_blocks.49.attn.to_out.0",
"transformer_blocks.49.attn.to_q",
"transformer_blocks.49.attn.to_v",
"transformer_blocks.49.ff.net.0.proj",
"transformer_blocks.49.ff.net.2.down",
"transformer_blocks.49.ff.net.2.up",
"transformer_blocks.5.attn.to_k",
"transformer_blocks.5.attn.to_out.0",
"transformer_blocks.5.attn.to_q",
"transformer_blocks.5.attn.to_v",
"transformer_blocks.5.ff.net.0.proj",
"transformer_blocks.5.ff.net.2",
"transformer_blocks.7.attn.to_k.down",
"transformer_blocks.7.attn.to_k.up",
"transformer_blocks.7.attn.to_out.0.down",
"transformer_blocks.7.attn.to_out.0.up",
"transformer_blocks.7.attn.to_q.down",
"transformer_blocks.7.attn.to_q.up",
"transformer_blocks.7.attn.to_v.down",
"transformer_blocks.7.attn.to_v.up",
"transformer_blocks.7.ff.net.0.proj.down",
"transformer_blocks.7.ff.net.0.proj.up",
"transformer_blocks.7.ff.net.2.down",
"transformer_blocks.7.ff.net.2.up",
"transformer_blocks.8.attn.to_k.down",
"transformer_blocks.8.attn.to_k.up",
"transformer_blocks.8.attn.to_out.0.down",
"transformer_blocks.8.attn.to_out.0.up",
"transformer_blocks.8.attn.to_q.down",
"transformer_blocks.8.attn.to_q.up",
"transformer_blocks.8.attn.to_v.down",
"transformer_blocks.8.attn.to_v.up",
"transformer_blocks.8.ff.net.0.proj.down",
"transformer_blocks.8.ff.net.0.proj.up",
"transformer_blocks.8.ff.net.2.down",
"transformer_blocks.8.ff.net.2.up",
"transformer_blocks.9.attn.to_k.down",
"transformer_blocks.9.attn.to_k.up",
"transformer_blocks.9.attn.to_out.0.down",
"transformer_blocks.9.attn.to_out.0.up",
"transformer_blocks.9.attn.to_q.down",
"transformer_blocks.9.attn.to_q.up",
"transformer_blocks.9.attn.to_v.down",
"transformer_blocks.9.attn.to_v.up",
"transformer_blocks.9.ff.net.0.proj.down",
"transformer_blocks.9.ff.net.0.proj.up",
"transformer_blocks.9.ff.net.2.down",
"transformer_blocks.9.ff.net.2.up"
]
}
}