WaveCut commited on
Commit
6d40ed4
·
verified ·
1 Parent(s): b0eedea

Add Z-Image Turbo OrbitQuant W4A4 artifact

Browse files
README.md ADDED
@@ -0,0 +1,38 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: Tongyi-MAI/Z-Image-Turbo
3
+ tags:
4
+ - orbitquant
5
+ - quantized
6
+ - diffusers
7
+ ---
8
+
9
+ # Tongyi-MAI/Z-Image-Turbo OrbitQuant W4A4
10
+
11
+ This is an OrbitQuant artifact generated from the source model listed above.
12
+
13
+ ## Quantization
14
+
15
+ - Method: `orbitquant`
16
+ - Bits: `W4A4`
17
+ - Runtime mode: `dequant_bf16`
18
+ - Activation kernel backend: `auto`
19
+ - Target policy: `z_image`
20
+ - Rotation: `rpbh`
21
+ - Rotation seed: `0`
22
+ - Block size: `paper`
23
+ - Block size policy: `largest_power_of_two_dividing_dim`
24
+ - Codebook: `lloyd_max`
25
+ - Codebook version: `1`
26
+ - Calibration data: none
27
+ - Text encoders and VAE: left in source precision by default
28
+
29
+ ## Source
30
+
31
+ - Model: `Tongyi-MAI/Z-Image-Turbo`
32
+ - Revision: `f332072aa78be7aecdf3ee76d5c247082da564a6`
33
+ - Source license: `apache-2.0`
34
+ - OrbitQuant paper: https://arxiv.org/abs/2607.02461
35
+
36
+ ## Limitations
37
+
38
+ The initial runtime may dequantize packed weights before BF16 matmul. Disk artifacts are compact; current CUDA/MPS kernels optimize selected activation codebook lookup/rescale stages and packed weight dequantization; full fused low-bit kernels are separate work.
SHA256SUMS ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ 6a3b533aa63864e5eac0804a406ba824c8748c638bcc111e355c1f55ceb0d976 README.md
2
+ e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855 assets/.gitkeep
3
+ 90d54eb9870d76c7e992e1ca94341cc7a58c931a079162ef1774b7f2b571ac73 benchmark/orbitquant.metrics.csv
4
+ e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855 benchmark/orbitquant.metrics.jsonl
5
+ 90d54eb9870d76c7e992e1ca94341cc7a58c931a079162ef1774b7f2b571ac73 benchmark/original.metrics.csv
6
+ e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855 benchmark/original.metrics.jsonl
7
+ 1c99fc94d6fdd328753ec0cac0d2d93da0a477a39a458cd413903b11c79191ba benchmark/summary.json
8
+ 0181f43913c3c647ec1adeebbc78da67de41c46558f939e1ad330d01b56d9c9b model.safetensors
9
+ 7af2cec0bf4745140b0572ea3dfadd1f0d97dad6db16c5e0165cda60f3798e00 model_index.json
10
+ d30c534198062aa21847808e080a1043358025d436a8654ea2c66f20735dce68 orbitquant_codebooks.safetensors
11
+ 54a34286660324ec479f6770297ad13ae5ef16583d730a6782b214b90b0bdce2 orbitquant_manifest.json
12
+ e1724c9394bf83ea05da6b457b8072d2bcb3d5a533dafce916da8deb517d3825 orbitquant_rotations.safetensors
13
+ 65ec8afff39a6cf80feee2eac2f06fa66e51cd471a8382f4355a04fd74eda0af prompts.json
14
+ 3134bdd201c7006489051dfa6ae86ca250cf35d3921e6785fdd0675c9963db59 quantization_config.json
assets/.gitkeep ADDED
File without changes
benchmark/orbitquant.metrics.csv ADDED
@@ -0,0 +1 @@
 
 
1
+ metric,value
benchmark/orbitquant.metrics.jsonl ADDED
File without changes
benchmark/original.metrics.csv ADDED
@@ -0,0 +1 @@
 
 
1
+ metric,value
benchmark/original.metrics.jsonl ADDED
File without changes
benchmark/summary.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "status": "not_run",
3
+ "source_model_id": "Tongyi-MAI/Z-Image-Turbo",
4
+ "source_revision": "f332072aa78be7aecdf3ee76d5c247082da564a6",
5
+ "weight_bits": 4,
6
+ "activation_bits": 4,
7
+ "target_policy": "z_image",
8
+ "runtime_mode": "dequant_bf16",
9
+ "activation_kernel_backend": "auto"
10
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0181f43913c3c647ec1adeebbc78da67de41c46558f939e1ad330d01b56d9c9b
3
+ size 3102040000
model_index.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_class_name": "OrbitQuantComponentArtifact",
3
+ "artifact_format": "orbitquant-v1",
4
+ "quant_method": "orbitquant",
5
+ "source_model_id": "Tongyi-MAI/Z-Image-Turbo",
6
+ "source_revision": "f332072aa78be7aecdf3ee76d5c247082da564a6",
7
+ "source_license": "apache-2.0",
8
+ "component": "transformer",
9
+ "weight_name": "model.safetensors",
10
+ "quantization_config": "quantization_config.json",
11
+ "manifest": "orbitquant_manifest.json",
12
+ "codebooks": "orbitquant_codebooks.safetensors",
13
+ "rotations": "orbitquant_rotations.safetensors",
14
+ "weight_bits": 4,
15
+ "activation_bits": 4,
16
+ "target_policy": "z_image",
17
+ "runtime_mode": "dequant_bf16",
18
+ "activation_kernel_backend": "auto"
19
+ }
orbitquant_codebooks.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d30c534198062aa21847808e080a1043358025d436a8654ea2c66f20735dce68
3
+ size 584
orbitquant_manifest.json ADDED
@@ -0,0 +1,2735 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "artifact_format": "orbitquant-v1",
3
+ "source_model_id": "Tongyi-MAI/Z-Image-Turbo",
4
+ "source_revision": "f332072aa78be7aecdf3ee76d5c247082da564a6",
5
+ "source_license": "apache-2.0",
6
+ "quant_method": "orbitquant",
7
+ "paper": "https://arxiv.org/abs/2607.02461",
8
+ "weight_bits": 4,
9
+ "activation_bits": 4,
10
+ "rotation": "rpbh",
11
+ "rotation_seed": 0,
12
+ "block_size": "paper",
13
+ "block_size_policy": "largest_power_of_two_dividing_dim",
14
+ "codebook": "lloyd_max",
15
+ "codebook_version": 1,
16
+ "row_norm_dtype": "bfloat16",
17
+ "runtime_mode": "dequant_bf16",
18
+ "activation_kernel_backend": "auto",
19
+ "target_policy": "z_image",
20
+ "adaln_policy": "int4_rtn_group64_bf16_activation",
21
+ "quantized_modules": [
22
+ "noise_refiner.0.attention.to_q",
23
+ "noise_refiner.0.attention.to_k",
24
+ "noise_refiner.0.attention.to_v",
25
+ "noise_refiner.0.attention.to_out.0",
26
+ "noise_refiner.0.feed_forward.w1",
27
+ "noise_refiner.0.feed_forward.w2",
28
+ "noise_refiner.0.feed_forward.w3",
29
+ "noise_refiner.1.attention.to_q",
30
+ "noise_refiner.1.attention.to_k",
31
+ "noise_refiner.1.attention.to_v",
32
+ "noise_refiner.1.attention.to_out.0",
33
+ "noise_refiner.1.feed_forward.w1",
34
+ "noise_refiner.1.feed_forward.w2",
35
+ "noise_refiner.1.feed_forward.w3",
36
+ "context_refiner.0.attention.to_q",
37
+ "context_refiner.0.attention.to_k",
38
+ "context_refiner.0.attention.to_v",
39
+ "context_refiner.0.attention.to_out.0",
40
+ "context_refiner.0.feed_forward.w1",
41
+ "context_refiner.0.feed_forward.w2",
42
+ "context_refiner.0.feed_forward.w3",
43
+ "context_refiner.1.attention.to_q",
44
+ "context_refiner.1.attention.to_k",
45
+ "context_refiner.1.attention.to_v",
46
+ "context_refiner.1.attention.to_out.0",
47
+ "context_refiner.1.feed_forward.w1",
48
+ "context_refiner.1.feed_forward.w2",
49
+ "context_refiner.1.feed_forward.w3",
50
+ "layers.0.attention.to_q",
51
+ "layers.0.attention.to_k",
52
+ "layers.0.attention.to_v",
53
+ "layers.0.attention.to_out.0",
54
+ "layers.0.feed_forward.w1",
55
+ "layers.0.feed_forward.w2",
56
+ "layers.0.feed_forward.w3",
57
+ "layers.1.attention.to_q",
58
+ "layers.1.attention.to_k",
59
+ "layers.1.attention.to_v",
60
+ "layers.1.attention.to_out.0",
61
+ "layers.1.feed_forward.w1",
62
+ "layers.1.feed_forward.w2",
63
+ "layers.1.feed_forward.w3",
64
+ "layers.2.attention.to_q",
65
+ "layers.2.attention.to_k",
66
+ "layers.2.attention.to_v",
67
+ "layers.2.attention.to_out.0",
68
+ "layers.2.feed_forward.w1",
69
+ "layers.2.feed_forward.w2",
70
+ "layers.2.feed_forward.w3",
71
+ "layers.3.attention.to_q",
72
+ "layers.3.attention.to_k",
73
+ "layers.3.attention.to_v",
74
+ "layers.3.attention.to_out.0",
75
+ "layers.3.feed_forward.w1",
76
+ "layers.3.feed_forward.w2",
77
+ "layers.3.feed_forward.w3",
78
+ "layers.4.attention.to_q",
79
+ "layers.4.attention.to_k",
80
+ "layers.4.attention.to_v",
81
+ "layers.4.attention.to_out.0",
82
+ "layers.4.feed_forward.w1",
83
+ "layers.4.feed_forward.w2",
84
+ "layers.4.feed_forward.w3",
85
+ "layers.5.attention.to_q",
86
+ "layers.5.attention.to_k",
87
+ "layers.5.attention.to_v",
88
+ "layers.5.attention.to_out.0",
89
+ "layers.5.feed_forward.w1",
90
+ "layers.5.feed_forward.w2",
91
+ "layers.5.feed_forward.w3",
92
+ "layers.6.attention.to_q",
93
+ "layers.6.attention.to_k",
94
+ "layers.6.attention.to_v",
95
+ "layers.6.attention.to_out.0",
96
+ "layers.6.feed_forward.w1",
97
+ "layers.6.feed_forward.w2",
98
+ "layers.6.feed_forward.w3",
99
+ "layers.7.attention.to_q",
100
+ "layers.7.attention.to_k",
101
+ "layers.7.attention.to_v",
102
+ "layers.7.attention.to_out.0",
103
+ "layers.7.feed_forward.w1",
104
+ "layers.7.feed_forward.w2",
105
+ "layers.7.feed_forward.w3",
106
+ "layers.8.attention.to_q",
107
+ "layers.8.attention.to_k",
108
+ "layers.8.attention.to_v",
109
+ "layers.8.attention.to_out.0",
110
+ "layers.8.feed_forward.w1",
111
+ "layers.8.feed_forward.w2",
112
+ "layers.8.feed_forward.w3",
113
+ "layers.9.attention.to_q",
114
+ "layers.9.attention.to_k",
115
+ "layers.9.attention.to_v",
116
+ "layers.9.attention.to_out.0",
117
+ "layers.9.feed_forward.w1",
118
+ "layers.9.feed_forward.w2",
119
+ "layers.9.feed_forward.w3",
120
+ "layers.10.attention.to_q",
121
+ "layers.10.attention.to_k",
122
+ "layers.10.attention.to_v",
123
+ "layers.10.attention.to_out.0",
124
+ "layers.10.feed_forward.w1",
125
+ "layers.10.feed_forward.w2",
126
+ "layers.10.feed_forward.w3",
127
+ "layers.11.attention.to_q",
128
+ "layers.11.attention.to_k",
129
+ "layers.11.attention.to_v",
130
+ "layers.11.attention.to_out.0",
131
+ "layers.11.feed_forward.w1",
132
+ "layers.11.feed_forward.w2",
133
+ "layers.11.feed_forward.w3",
134
+ "layers.12.attention.to_q",
135
+ "layers.12.attention.to_k",
136
+ "layers.12.attention.to_v",
137
+ "layers.12.attention.to_out.0",
138
+ "layers.12.feed_forward.w1",
139
+ "layers.12.feed_forward.w2",
140
+ "layers.12.feed_forward.w3",
141
+ "layers.13.attention.to_q",
142
+ "layers.13.attention.to_k",
143
+ "layers.13.attention.to_v",
144
+ "layers.13.attention.to_out.0",
145
+ "layers.13.feed_forward.w1",
146
+ "layers.13.feed_forward.w2",
147
+ "layers.13.feed_forward.w3",
148
+ "layers.14.attention.to_q",
149
+ "layers.14.attention.to_k",
150
+ "layers.14.attention.to_v",
151
+ "layers.14.attention.to_out.0",
152
+ "layers.14.feed_forward.w1",
153
+ "layers.14.feed_forward.w2",
154
+ "layers.14.feed_forward.w3",
155
+ "layers.15.attention.to_q",
156
+ "layers.15.attention.to_k",
157
+ "layers.15.attention.to_v",
158
+ "layers.15.attention.to_out.0",
159
+ "layers.15.feed_forward.w1",
160
+ "layers.15.feed_forward.w2",
161
+ "layers.15.feed_forward.w3",
162
+ "layers.16.attention.to_q",
163
+ "layers.16.attention.to_k",
164
+ "layers.16.attention.to_v",
165
+ "layers.16.attention.to_out.0",
166
+ "layers.16.feed_forward.w1",
167
+ "layers.16.feed_forward.w2",
168
+ "layers.16.feed_forward.w3",
169
+ "layers.17.attention.to_q",
170
+ "layers.17.attention.to_k",
171
+ "layers.17.attention.to_v",
172
+ "layers.17.attention.to_out.0",
173
+ "layers.17.feed_forward.w1",
174
+ "layers.17.feed_forward.w2",
175
+ "layers.17.feed_forward.w3",
176
+ "layers.18.attention.to_q",
177
+ "layers.18.attention.to_k",
178
+ "layers.18.attention.to_v",
179
+ "layers.18.attention.to_out.0",
180
+ "layers.18.feed_forward.w1",
181
+ "layers.18.feed_forward.w2",
182
+ "layers.18.feed_forward.w3",
183
+ "layers.19.attention.to_q",
184
+ "layers.19.attention.to_k",
185
+ "layers.19.attention.to_v",
186
+ "layers.19.attention.to_out.0",
187
+ "layers.19.feed_forward.w1",
188
+ "layers.19.feed_forward.w2",
189
+ "layers.19.feed_forward.w3",
190
+ "layers.20.attention.to_q",
191
+ "layers.20.attention.to_k",
192
+ "layers.20.attention.to_v",
193
+ "layers.20.attention.to_out.0",
194
+ "layers.20.feed_forward.w1",
195
+ "layers.20.feed_forward.w2",
196
+ "layers.20.feed_forward.w3",
197
+ "layers.21.attention.to_q",
198
+ "layers.21.attention.to_k",
199
+ "layers.21.attention.to_v",
200
+ "layers.21.attention.to_out.0",
201
+ "layers.21.feed_forward.w1",
202
+ "layers.21.feed_forward.w2",
203
+ "layers.21.feed_forward.w3",
204
+ "layers.22.attention.to_q",
205
+ "layers.22.attention.to_k",
206
+ "layers.22.attention.to_v",
207
+ "layers.22.attention.to_out.0",
208
+ "layers.22.feed_forward.w1",
209
+ "layers.22.feed_forward.w2",
210
+ "layers.22.feed_forward.w3",
211
+ "layers.23.attention.to_q",
212
+ "layers.23.attention.to_k",
213
+ "layers.23.attention.to_v",
214
+ "layers.23.attention.to_out.0",
215
+ "layers.23.feed_forward.w1",
216
+ "layers.23.feed_forward.w2",
217
+ "layers.23.feed_forward.w3",
218
+ "layers.24.attention.to_q",
219
+ "layers.24.attention.to_k",
220
+ "layers.24.attention.to_v",
221
+ "layers.24.attention.to_out.0",
222
+ "layers.24.feed_forward.w1",
223
+ "layers.24.feed_forward.w2",
224
+ "layers.24.feed_forward.w3",
225
+ "layers.25.attention.to_q",
226
+ "layers.25.attention.to_k",
227
+ "layers.25.attention.to_v",
228
+ "layers.25.attention.to_out.0",
229
+ "layers.25.feed_forward.w1",
230
+ "layers.25.feed_forward.w2",
231
+ "layers.25.feed_forward.w3",
232
+ "layers.26.attention.to_q",
233
+ "layers.26.attention.to_k",
234
+ "layers.26.attention.to_v",
235
+ "layers.26.attention.to_out.0",
236
+ "layers.26.feed_forward.w1",
237
+ "layers.26.feed_forward.w2",
238
+ "layers.26.feed_forward.w3",
239
+ "layers.27.attention.to_q",
240
+ "layers.27.attention.to_k",
241
+ "layers.27.attention.to_v",
242
+ "layers.27.attention.to_out.0",
243
+ "layers.27.feed_forward.w1",
244
+ "layers.27.feed_forward.w2",
245
+ "layers.27.feed_forward.w3",
246
+ "layers.28.attention.to_q",
247
+ "layers.28.attention.to_k",
248
+ "layers.28.attention.to_v",
249
+ "layers.28.attention.to_out.0",
250
+ "layers.28.feed_forward.w1",
251
+ "layers.28.feed_forward.w2",
252
+ "layers.28.feed_forward.w3",
253
+ "layers.29.attention.to_q",
254
+ "layers.29.attention.to_k",
255
+ "layers.29.attention.to_v",
256
+ "layers.29.attention.to_out.0",
257
+ "layers.29.feed_forward.w1",
258
+ "layers.29.feed_forward.w2",
259
+ "layers.29.feed_forward.w3"
260
+ ],
261
+ "adaln_modules": [
262
+ "all_final_layer.2-1.adaLN_modulation.1",
263
+ "noise_refiner.0.adaLN_modulation.0",
264
+ "noise_refiner.1.adaLN_modulation.0",
265
+ "layers.0.adaLN_modulation.0",
266
+ "layers.1.adaLN_modulation.0",
267
+ "layers.2.adaLN_modulation.0",
268
+ "layers.3.adaLN_modulation.0",
269
+ "layers.4.adaLN_modulation.0",
270
+ "layers.5.adaLN_modulation.0",
271
+ "layers.6.adaLN_modulation.0",
272
+ "layers.7.adaLN_modulation.0",
273
+ "layers.8.adaLN_modulation.0",
274
+ "layers.9.adaLN_modulation.0",
275
+ "layers.10.adaLN_modulation.0",
276
+ "layers.11.adaLN_modulation.0",
277
+ "layers.12.adaLN_modulation.0",
278
+ "layers.13.adaLN_modulation.0",
279
+ "layers.14.adaLN_modulation.0",
280
+ "layers.15.adaLN_modulation.0",
281
+ "layers.16.adaLN_modulation.0",
282
+ "layers.17.adaLN_modulation.0",
283
+ "layers.18.adaLN_modulation.0",
284
+ "layers.19.adaLN_modulation.0",
285
+ "layers.20.adaLN_modulation.0",
286
+ "layers.21.adaLN_modulation.0",
287
+ "layers.22.adaLN_modulation.0",
288
+ "layers.23.adaLN_modulation.0",
289
+ "layers.24.adaLN_modulation.0",
290
+ "layers.25.adaLN_modulation.0",
291
+ "layers.26.adaLN_modulation.0",
292
+ "layers.27.adaLN_modulation.0",
293
+ "layers.28.adaLN_modulation.0",
294
+ "layers.29.adaLN_modulation.0"
295
+ ],
296
+ "skipped_modules": [
297
+ "all_x_embedder.2-1",
298
+ "all_final_layer.2-1.linear",
299
+ "t_embedder.mlp.0",
300
+ "t_embedder.mlp.2",
301
+ "cap_embedder.1"
302
+ ],
303
+ "module_shapes": {
304
+ "x_pad_token": [
305
+ 1,
306
+ 3840
307
+ ],
308
+ "cap_pad_token": [
309
+ 1,
310
+ 3840
311
+ ],
312
+ "all_x_embedder.2-1.weight": [
313
+ 3840,
314
+ 64
315
+ ],
316
+ "all_x_embedder.2-1.bias": [
317
+ 3840
318
+ ],
319
+ "all_final_layer.2-1.linear.weight": [
320
+ 64,
321
+ 3840
322
+ ],
323
+ "all_final_layer.2-1.linear.bias": [
324
+ 64
325
+ ],
326
+ "all_final_layer.2-1.adaLN_modulation.1.bias": [
327
+ 3840
328
+ ],
329
+ "all_final_layer.2-1.adaLN_modulation.1.packed_weight": [
330
+ 491520
331
+ ],
332
+ "all_final_layer.2-1.adaLN_modulation.1.scales": [
333
+ 3840,
334
+ 4
335
+ ],
336
+ "noise_refiner.0.attention.norm_q.weight": [
337
+ 128
338
+ ],
339
+ "noise_refiner.0.attention.norm_k.weight": [
340
+ 128
341
+ ],
342
+ "noise_refiner.0.attention.to_q.packed_weight_indices": [
343
+ 7372800
344
+ ],
345
+ "noise_refiner.0.attention.to_q.row_norms": [
346
+ 3840
347
+ ],
348
+ "noise_refiner.0.attention.to_k.packed_weight_indices": [
349
+ 7372800
350
+ ],
351
+ "noise_refiner.0.attention.to_k.row_norms": [
352
+ 3840
353
+ ],
354
+ "noise_refiner.0.attention.to_v.packed_weight_indices": [
355
+ 7372800
356
+ ],
357
+ "noise_refiner.0.attention.to_v.row_norms": [
358
+ 3840
359
+ ],
360
+ "noise_refiner.0.attention.to_out.0.packed_weight_indices": [
361
+ 7372800
362
+ ],
363
+ "noise_refiner.0.attention.to_out.0.row_norms": [
364
+ 3840
365
+ ],
366
+ "noise_refiner.0.feed_forward.w1.packed_weight_indices": [
367
+ 19660800
368
+ ],
369
+ "noise_refiner.0.feed_forward.w1.row_norms": [
370
+ 10240
371
+ ],
372
+ "noise_refiner.0.feed_forward.w2.packed_weight_indices": [
373
+ 19660800
374
+ ],
375
+ "noise_refiner.0.feed_forward.w2.row_norms": [
376
+ 3840
377
+ ],
378
+ "noise_refiner.0.feed_forward.w3.packed_weight_indices": [
379
+ 19660800
380
+ ],
381
+ "noise_refiner.0.feed_forward.w3.row_norms": [
382
+ 10240
383
+ ],
384
+ "noise_refiner.0.attention_norm1.weight": [
385
+ 3840
386
+ ],
387
+ "noise_refiner.0.ffn_norm1.weight": [
388
+ 3840
389
+ ],
390
+ "noise_refiner.0.attention_norm2.weight": [
391
+ 3840
392
+ ],
393
+ "noise_refiner.0.ffn_norm2.weight": [
394
+ 3840
395
+ ],
396
+ "noise_refiner.0.adaLN_modulation.0.bias": [
397
+ 15360
398
+ ],
399
+ "noise_refiner.0.adaLN_modulation.0.packed_weight": [
400
+ 1966080
401
+ ],
402
+ "noise_refiner.0.adaLN_modulation.0.scales": [
403
+ 15360,
404
+ 4
405
+ ],
406
+ "noise_refiner.1.attention.norm_q.weight": [
407
+ 128
408
+ ],
409
+ "noise_refiner.1.attention.norm_k.weight": [
410
+ 128
411
+ ],
412
+ "noise_refiner.1.attention.to_q.packed_weight_indices": [
413
+ 7372800
414
+ ],
415
+ "noise_refiner.1.attention.to_q.row_norms": [
416
+ 3840
417
+ ],
418
+ "noise_refiner.1.attention.to_k.packed_weight_indices": [
419
+ 7372800
420
+ ],
421
+ "noise_refiner.1.attention.to_k.row_norms": [
422
+ 3840
423
+ ],
424
+ "noise_refiner.1.attention.to_v.packed_weight_indices": [
425
+ 7372800
426
+ ],
427
+ "noise_refiner.1.attention.to_v.row_norms": [
428
+ 3840
429
+ ],
430
+ "noise_refiner.1.attention.to_out.0.packed_weight_indices": [
431
+ 7372800
432
+ ],
433
+ "noise_refiner.1.attention.to_out.0.row_norms": [
434
+ 3840
435
+ ],
436
+ "noise_refiner.1.feed_forward.w1.packed_weight_indices": [
437
+ 19660800
438
+ ],
439
+ "noise_refiner.1.feed_forward.w1.row_norms": [
440
+ 10240
441
+ ],
442
+ "noise_refiner.1.feed_forward.w2.packed_weight_indices": [
443
+ 19660800
444
+ ],
445
+ "noise_refiner.1.feed_forward.w2.row_norms": [
446
+ 3840
447
+ ],
448
+ "noise_refiner.1.feed_forward.w3.packed_weight_indices": [
449
+ 19660800
450
+ ],
451
+ "noise_refiner.1.feed_forward.w3.row_norms": [
452
+ 10240
453
+ ],
454
+ "noise_refiner.1.attention_norm1.weight": [
455
+ 3840
456
+ ],
457
+ "noise_refiner.1.ffn_norm1.weight": [
458
+ 3840
459
+ ],
460
+ "noise_refiner.1.attention_norm2.weight": [
461
+ 3840
462
+ ],
463
+ "noise_refiner.1.ffn_norm2.weight": [
464
+ 3840
465
+ ],
466
+ "noise_refiner.1.adaLN_modulation.0.bias": [
467
+ 15360
468
+ ],
469
+ "noise_refiner.1.adaLN_modulation.0.packed_weight": [
470
+ 1966080
471
+ ],
472
+ "noise_refiner.1.adaLN_modulation.0.scales": [
473
+ 15360,
474
+ 4
475
+ ],
476
+ "context_refiner.0.attention.norm_q.weight": [
477
+ 128
478
+ ],
479
+ "context_refiner.0.attention.norm_k.weight": [
480
+ 128
481
+ ],
482
+ "context_refiner.0.attention.to_q.packed_weight_indices": [
483
+ 7372800
484
+ ],
485
+ "context_refiner.0.attention.to_q.row_norms": [
486
+ 3840
487
+ ],
488
+ "context_refiner.0.attention.to_k.packed_weight_indices": [
489
+ 7372800
490
+ ],
491
+ "context_refiner.0.attention.to_k.row_norms": [
492
+ 3840
493
+ ],
494
+ "context_refiner.0.attention.to_v.packed_weight_indices": [
495
+ 7372800
496
+ ],
497
+ "context_refiner.0.attention.to_v.row_norms": [
498
+ 3840
499
+ ],
500
+ "context_refiner.0.attention.to_out.0.packed_weight_indices": [
501
+ 7372800
502
+ ],
503
+ "context_refiner.0.attention.to_out.0.row_norms": [
504
+ 3840
505
+ ],
506
+ "context_refiner.0.feed_forward.w1.packed_weight_indices": [
507
+ 19660800
508
+ ],
509
+ "context_refiner.0.feed_forward.w1.row_norms": [
510
+ 10240
511
+ ],
512
+ "context_refiner.0.feed_forward.w2.packed_weight_indices": [
513
+ 19660800
514
+ ],
515
+ "context_refiner.0.feed_forward.w2.row_norms": [
516
+ 3840
517
+ ],
518
+ "context_refiner.0.feed_forward.w3.packed_weight_indices": [
519
+ 19660800
520
+ ],
521
+ "context_refiner.0.feed_forward.w3.row_norms": [
522
+ 10240
523
+ ],
524
+ "context_refiner.0.attention_norm1.weight": [
525
+ 3840
526
+ ],
527
+ "context_refiner.0.ffn_norm1.weight": [
528
+ 3840
529
+ ],
530
+ "context_refiner.0.attention_norm2.weight": [
531
+ 3840
532
+ ],
533
+ "context_refiner.0.ffn_norm2.weight": [
534
+ 3840
535
+ ],
536
+ "context_refiner.1.attention.norm_q.weight": [
537
+ 128
538
+ ],
539
+ "context_refiner.1.attention.norm_k.weight": [
540
+ 128
541
+ ],
542
+ "context_refiner.1.attention.to_q.packed_weight_indices": [
543
+ 7372800
544
+ ],
545
+ "context_refiner.1.attention.to_q.row_norms": [
546
+ 3840
547
+ ],
548
+ "context_refiner.1.attention.to_k.packed_weight_indices": [
549
+ 7372800
550
+ ],
551
+ "context_refiner.1.attention.to_k.row_norms": [
552
+ 3840
553
+ ],
554
+ "context_refiner.1.attention.to_v.packed_weight_indices": [
555
+ 7372800
556
+ ],
557
+ "context_refiner.1.attention.to_v.row_norms": [
558
+ 3840
559
+ ],
560
+ "context_refiner.1.attention.to_out.0.packed_weight_indices": [
561
+ 7372800
562
+ ],
563
+ "context_refiner.1.attention.to_out.0.row_norms": [
564
+ 3840
565
+ ],
566
+ "context_refiner.1.feed_forward.w1.packed_weight_indices": [
567
+ 19660800
568
+ ],
569
+ "context_refiner.1.feed_forward.w1.row_norms": [
570
+ 10240
571
+ ],
572
+ "context_refiner.1.feed_forward.w2.packed_weight_indices": [
573
+ 19660800
574
+ ],
575
+ "context_refiner.1.feed_forward.w2.row_norms": [
576
+ 3840
577
+ ],
578
+ "context_refiner.1.feed_forward.w3.packed_weight_indices": [
579
+ 19660800
580
+ ],
581
+ "context_refiner.1.feed_forward.w3.row_norms": [
582
+ 10240
583
+ ],
584
+ "context_refiner.1.attention_norm1.weight": [
585
+ 3840
586
+ ],
587
+ "context_refiner.1.ffn_norm1.weight": [
588
+ 3840
589
+ ],
590
+ "context_refiner.1.attention_norm2.weight": [
591
+ 3840
592
+ ],
593
+ "context_refiner.1.ffn_norm2.weight": [
594
+ 3840
595
+ ],
596
+ "t_embedder.mlp.0.weight": [
597
+ 1024,
598
+ 256
599
+ ],
600
+ "t_embedder.mlp.0.bias": [
601
+ 1024
602
+ ],
603
+ "t_embedder.mlp.2.weight": [
604
+ 256,
605
+ 1024
606
+ ],
607
+ "t_embedder.mlp.2.bias": [
608
+ 256
609
+ ],
610
+ "cap_embedder.0.weight": [
611
+ 2560
612
+ ],
613
+ "cap_embedder.1.weight": [
614
+ 3840,
615
+ 2560
616
+ ],
617
+ "cap_embedder.1.bias": [
618
+ 3840
619
+ ],
620
+ "layers.0.attention.norm_q.weight": [
621
+ 128
622
+ ],
623
+ "layers.0.attention.norm_k.weight": [
624
+ 128
625
+ ],
626
+ "layers.0.attention.to_q.packed_weight_indices": [
627
+ 7372800
628
+ ],
629
+ "layers.0.attention.to_q.row_norms": [
630
+ 3840
631
+ ],
632
+ "layers.0.attention.to_k.packed_weight_indices": [
633
+ 7372800
634
+ ],
635
+ "layers.0.attention.to_k.row_norms": [
636
+ 3840
637
+ ],
638
+ "layers.0.attention.to_v.packed_weight_indices": [
639
+ 7372800
640
+ ],
641
+ "layers.0.attention.to_v.row_norms": [
642
+ 3840
643
+ ],
644
+ "layers.0.attention.to_out.0.packed_weight_indices": [
645
+ 7372800
646
+ ],
647
+ "layers.0.attention.to_out.0.row_norms": [
648
+ 3840
649
+ ],
650
+ "layers.0.feed_forward.w1.packed_weight_indices": [
651
+ 19660800
652
+ ],
653
+ "layers.0.feed_forward.w1.row_norms": [
654
+ 10240
655
+ ],
656
+ "layers.0.feed_forward.w2.packed_weight_indices": [
657
+ 19660800
658
+ ],
659
+ "layers.0.feed_forward.w2.row_norms": [
660
+ 3840
661
+ ],
662
+ "layers.0.feed_forward.w3.packed_weight_indices": [
663
+ 19660800
664
+ ],
665
+ "layers.0.feed_forward.w3.row_norms": [
666
+ 10240
667
+ ],
668
+ "layers.0.attention_norm1.weight": [
669
+ 3840
670
+ ],
671
+ "layers.0.ffn_norm1.weight": [
672
+ 3840
673
+ ],
674
+ "layers.0.attention_norm2.weight": [
675
+ 3840
676
+ ],
677
+ "layers.0.ffn_norm2.weight": [
678
+ 3840
679
+ ],
680
+ "layers.0.adaLN_modulation.0.bias": [
681
+ 15360
682
+ ],
683
+ "layers.0.adaLN_modulation.0.packed_weight": [
684
+ 1966080
685
+ ],
686
+ "layers.0.adaLN_modulation.0.scales": [
687
+ 15360,
688
+ 4
689
+ ],
690
+ "layers.1.attention.norm_q.weight": [
691
+ 128
692
+ ],
693
+ "layers.1.attention.norm_k.weight": [
694
+ 128
695
+ ],
696
+ "layers.1.attention.to_q.packed_weight_indices": [
697
+ 7372800
698
+ ],
699
+ "layers.1.attention.to_q.row_norms": [
700
+ 3840
701
+ ],
702
+ "layers.1.attention.to_k.packed_weight_indices": [
703
+ 7372800
704
+ ],
705
+ "layers.1.attention.to_k.row_norms": [
706
+ 3840
707
+ ],
708
+ "layers.1.attention.to_v.packed_weight_indices": [
709
+ 7372800
710
+ ],
711
+ "layers.1.attention.to_v.row_norms": [
712
+ 3840
713
+ ],
714
+ "layers.1.attention.to_out.0.packed_weight_indices": [
715
+ 7372800
716
+ ],
717
+ "layers.1.attention.to_out.0.row_norms": [
718
+ 3840
719
+ ],
720
+ "layers.1.feed_forward.w1.packed_weight_indices": [
721
+ 19660800
722
+ ],
723
+ "layers.1.feed_forward.w1.row_norms": [
724
+ 10240
725
+ ],
726
+ "layers.1.feed_forward.w2.packed_weight_indices": [
727
+ 19660800
728
+ ],
729
+ "layers.1.feed_forward.w2.row_norms": [
730
+ 3840
731
+ ],
732
+ "layers.1.feed_forward.w3.packed_weight_indices": [
733
+ 19660800
734
+ ],
735
+ "layers.1.feed_forward.w3.row_norms": [
736
+ 10240
737
+ ],
738
+ "layers.1.attention_norm1.weight": [
739
+ 3840
740
+ ],
741
+ "layers.1.ffn_norm1.weight": [
742
+ 3840
743
+ ],
744
+ "layers.1.attention_norm2.weight": [
745
+ 3840
746
+ ],
747
+ "layers.1.ffn_norm2.weight": [
748
+ 3840
749
+ ],
750
+ "layers.1.adaLN_modulation.0.bias": [
751
+ 15360
752
+ ],
753
+ "layers.1.adaLN_modulation.0.packed_weight": [
754
+ 1966080
755
+ ],
756
+ "layers.1.adaLN_modulation.0.scales": [
757
+ 15360,
758
+ 4
759
+ ],
760
+ "layers.2.attention.norm_q.weight": [
761
+ 128
762
+ ],
763
+ "layers.2.attention.norm_k.weight": [
764
+ 128
765
+ ],
766
+ "layers.2.attention.to_q.packed_weight_indices": [
767
+ 7372800
768
+ ],
769
+ "layers.2.attention.to_q.row_norms": [
770
+ 3840
771
+ ],
772
+ "layers.2.attention.to_k.packed_weight_indices": [
773
+ 7372800
774
+ ],
775
+ "layers.2.attention.to_k.row_norms": [
776
+ 3840
777
+ ],
778
+ "layers.2.attention.to_v.packed_weight_indices": [
779
+ 7372800
780
+ ],
781
+ "layers.2.attention.to_v.row_norms": [
782
+ 3840
783
+ ],
784
+ "layers.2.attention.to_out.0.packed_weight_indices": [
785
+ 7372800
786
+ ],
787
+ "layers.2.attention.to_out.0.row_norms": [
788
+ 3840
789
+ ],
790
+ "layers.2.feed_forward.w1.packed_weight_indices": [
791
+ 19660800
792
+ ],
793
+ "layers.2.feed_forward.w1.row_norms": [
794
+ 10240
795
+ ],
796
+ "layers.2.feed_forward.w2.packed_weight_indices": [
797
+ 19660800
798
+ ],
799
+ "layers.2.feed_forward.w2.row_norms": [
800
+ 3840
801
+ ],
802
+ "layers.2.feed_forward.w3.packed_weight_indices": [
803
+ 19660800
804
+ ],
805
+ "layers.2.feed_forward.w3.row_norms": [
806
+ 10240
807
+ ],
808
+ "layers.2.attention_norm1.weight": [
809
+ 3840
810
+ ],
811
+ "layers.2.ffn_norm1.weight": [
812
+ 3840
813
+ ],
814
+ "layers.2.attention_norm2.weight": [
815
+ 3840
816
+ ],
817
+ "layers.2.ffn_norm2.weight": [
818
+ 3840
819
+ ],
820
+ "layers.2.adaLN_modulation.0.bias": [
821
+ 15360
822
+ ],
823
+ "layers.2.adaLN_modulation.0.packed_weight": [
824
+ 1966080
825
+ ],
826
+ "layers.2.adaLN_modulation.0.scales": [
827
+ 15360,
828
+ 4
829
+ ],
830
+ "layers.3.attention.norm_q.weight": [
831
+ 128
832
+ ],
833
+ "layers.3.attention.norm_k.weight": [
834
+ 128
835
+ ],
836
+ "layers.3.attention.to_q.packed_weight_indices": [
837
+ 7372800
838
+ ],
839
+ "layers.3.attention.to_q.row_norms": [
840
+ 3840
841
+ ],
842
+ "layers.3.attention.to_k.packed_weight_indices": [
843
+ 7372800
844
+ ],
845
+ "layers.3.attention.to_k.row_norms": [
846
+ 3840
847
+ ],
848
+ "layers.3.attention.to_v.packed_weight_indices": [
849
+ 7372800
850
+ ],
851
+ "layers.3.attention.to_v.row_norms": [
852
+ 3840
853
+ ],
854
+ "layers.3.attention.to_out.0.packed_weight_indices": [
855
+ 7372800
856
+ ],
857
+ "layers.3.attention.to_out.0.row_norms": [
858
+ 3840
859
+ ],
860
+ "layers.3.feed_forward.w1.packed_weight_indices": [
861
+ 19660800
862
+ ],
863
+ "layers.3.feed_forward.w1.row_norms": [
864
+ 10240
865
+ ],
866
+ "layers.3.feed_forward.w2.packed_weight_indices": [
867
+ 19660800
868
+ ],
869
+ "layers.3.feed_forward.w2.row_norms": [
870
+ 3840
871
+ ],
872
+ "layers.3.feed_forward.w3.packed_weight_indices": [
873
+ 19660800
874
+ ],
875
+ "layers.3.feed_forward.w3.row_norms": [
876
+ 10240
877
+ ],
878
+ "layers.3.attention_norm1.weight": [
879
+ 3840
880
+ ],
881
+ "layers.3.ffn_norm1.weight": [
882
+ 3840
883
+ ],
884
+ "layers.3.attention_norm2.weight": [
885
+ 3840
886
+ ],
887
+ "layers.3.ffn_norm2.weight": [
888
+ 3840
889
+ ],
890
+ "layers.3.adaLN_modulation.0.bias": [
891
+ 15360
892
+ ],
893
+ "layers.3.adaLN_modulation.0.packed_weight": [
894
+ 1966080
895
+ ],
896
+ "layers.3.adaLN_modulation.0.scales": [
897
+ 15360,
898
+ 4
899
+ ],
900
+ "layers.4.attention.norm_q.weight": [
901
+ 128
902
+ ],
903
+ "layers.4.attention.norm_k.weight": [
904
+ 128
905
+ ],
906
+ "layers.4.attention.to_q.packed_weight_indices": [
907
+ 7372800
908
+ ],
909
+ "layers.4.attention.to_q.row_norms": [
910
+ 3840
911
+ ],
912
+ "layers.4.attention.to_k.packed_weight_indices": [
913
+ 7372800
914
+ ],
915
+ "layers.4.attention.to_k.row_norms": [
916
+ 3840
917
+ ],
918
+ "layers.4.attention.to_v.packed_weight_indices": [
919
+ 7372800
920
+ ],
921
+ "layers.4.attention.to_v.row_norms": [
922
+ 3840
923
+ ],
924
+ "layers.4.attention.to_out.0.packed_weight_indices": [
925
+ 7372800
926
+ ],
927
+ "layers.4.attention.to_out.0.row_norms": [
928
+ 3840
929
+ ],
930
+ "layers.4.feed_forward.w1.packed_weight_indices": [
931
+ 19660800
932
+ ],
933
+ "layers.4.feed_forward.w1.row_norms": [
934
+ 10240
935
+ ],
936
+ "layers.4.feed_forward.w2.packed_weight_indices": [
937
+ 19660800
938
+ ],
939
+ "layers.4.feed_forward.w2.row_norms": [
940
+ 3840
941
+ ],
942
+ "layers.4.feed_forward.w3.packed_weight_indices": [
943
+ 19660800
944
+ ],
945
+ "layers.4.feed_forward.w3.row_norms": [
946
+ 10240
947
+ ],
948
+ "layers.4.attention_norm1.weight": [
949
+ 3840
950
+ ],
951
+ "layers.4.ffn_norm1.weight": [
952
+ 3840
953
+ ],
954
+ "layers.4.attention_norm2.weight": [
955
+ 3840
956
+ ],
957
+ "layers.4.ffn_norm2.weight": [
958
+ 3840
959
+ ],
960
+ "layers.4.adaLN_modulation.0.bias": [
961
+ 15360
962
+ ],
963
+ "layers.4.adaLN_modulation.0.packed_weight": [
964
+ 1966080
965
+ ],
966
+ "layers.4.adaLN_modulation.0.scales": [
967
+ 15360,
968
+ 4
969
+ ],
970
+ "layers.5.attention.norm_q.weight": [
971
+ 128
972
+ ],
973
+ "layers.5.attention.norm_k.weight": [
974
+ 128
975
+ ],
976
+ "layers.5.attention.to_q.packed_weight_indices": [
977
+ 7372800
978
+ ],
979
+ "layers.5.attention.to_q.row_norms": [
980
+ 3840
981
+ ],
982
+ "layers.5.attention.to_k.packed_weight_indices": [
983
+ 7372800
984
+ ],
985
+ "layers.5.attention.to_k.row_norms": [
986
+ 3840
987
+ ],
988
+ "layers.5.attention.to_v.packed_weight_indices": [
989
+ 7372800
990
+ ],
991
+ "layers.5.attention.to_v.row_norms": [
992
+ 3840
993
+ ],
994
+ "layers.5.attention.to_out.0.packed_weight_indices": [
995
+ 7372800
996
+ ],
997
+ "layers.5.attention.to_out.0.row_norms": [
998
+ 3840
999
+ ],
1000
+ "layers.5.feed_forward.w1.packed_weight_indices": [
1001
+ 19660800
1002
+ ],
1003
+ "layers.5.feed_forward.w1.row_norms": [
1004
+ 10240
1005
+ ],
1006
+ "layers.5.feed_forward.w2.packed_weight_indices": [
1007
+ 19660800
1008
+ ],
1009
+ "layers.5.feed_forward.w2.row_norms": [
1010
+ 3840
1011
+ ],
1012
+ "layers.5.feed_forward.w3.packed_weight_indices": [
1013
+ 19660800
1014
+ ],
1015
+ "layers.5.feed_forward.w3.row_norms": [
1016
+ 10240
1017
+ ],
1018
+ "layers.5.attention_norm1.weight": [
1019
+ 3840
1020
+ ],
1021
+ "layers.5.ffn_norm1.weight": [
1022
+ 3840
1023
+ ],
1024
+ "layers.5.attention_norm2.weight": [
1025
+ 3840
1026
+ ],
1027
+ "layers.5.ffn_norm2.weight": [
1028
+ 3840
1029
+ ],
1030
+ "layers.5.adaLN_modulation.0.bias": [
1031
+ 15360
1032
+ ],
1033
+ "layers.5.adaLN_modulation.0.packed_weight": [
1034
+ 1966080
1035
+ ],
1036
+ "layers.5.adaLN_modulation.0.scales": [
1037
+ 15360,
1038
+ 4
1039
+ ],
1040
+ "layers.6.attention.norm_q.weight": [
1041
+ 128
1042
+ ],
1043
+ "layers.6.attention.norm_k.weight": [
1044
+ 128
1045
+ ],
1046
+ "layers.6.attention.to_q.packed_weight_indices": [
1047
+ 7372800
1048
+ ],
1049
+ "layers.6.attention.to_q.row_norms": [
1050
+ 3840
1051
+ ],
1052
+ "layers.6.attention.to_k.packed_weight_indices": [
1053
+ 7372800
1054
+ ],
1055
+ "layers.6.attention.to_k.row_norms": [
1056
+ 3840
1057
+ ],
1058
+ "layers.6.attention.to_v.packed_weight_indices": [
1059
+ 7372800
1060
+ ],
1061
+ "layers.6.attention.to_v.row_norms": [
1062
+ 3840
1063
+ ],
1064
+ "layers.6.attention.to_out.0.packed_weight_indices": [
1065
+ 7372800
1066
+ ],
1067
+ "layers.6.attention.to_out.0.row_norms": [
1068
+ 3840
1069
+ ],
1070
+ "layers.6.feed_forward.w1.packed_weight_indices": [
1071
+ 19660800
1072
+ ],
1073
+ "layers.6.feed_forward.w1.row_norms": [
1074
+ 10240
1075
+ ],
1076
+ "layers.6.feed_forward.w2.packed_weight_indices": [
1077
+ 19660800
1078
+ ],
1079
+ "layers.6.feed_forward.w2.row_norms": [
1080
+ 3840
1081
+ ],
1082
+ "layers.6.feed_forward.w3.packed_weight_indices": [
1083
+ 19660800
1084
+ ],
1085
+ "layers.6.feed_forward.w3.row_norms": [
1086
+ 10240
1087
+ ],
1088
+ "layers.6.attention_norm1.weight": [
1089
+ 3840
1090
+ ],
1091
+ "layers.6.ffn_norm1.weight": [
1092
+ 3840
1093
+ ],
1094
+ "layers.6.attention_norm2.weight": [
1095
+ 3840
1096
+ ],
1097
+ "layers.6.ffn_norm2.weight": [
1098
+ 3840
1099
+ ],
1100
+ "layers.6.adaLN_modulation.0.bias": [
1101
+ 15360
1102
+ ],
1103
+ "layers.6.adaLN_modulation.0.packed_weight": [
1104
+ 1966080
1105
+ ],
1106
+ "layers.6.adaLN_modulation.0.scales": [
1107
+ 15360,
1108
+ 4
1109
+ ],
1110
+ "layers.7.attention.norm_q.weight": [
1111
+ 128
1112
+ ],
1113
+ "layers.7.attention.norm_k.weight": [
1114
+ 128
1115
+ ],
1116
+ "layers.7.attention.to_q.packed_weight_indices": [
1117
+ 7372800
1118
+ ],
1119
+ "layers.7.attention.to_q.row_norms": [
1120
+ 3840
1121
+ ],
1122
+ "layers.7.attention.to_k.packed_weight_indices": [
1123
+ 7372800
1124
+ ],
1125
+ "layers.7.attention.to_k.row_norms": [
1126
+ 3840
1127
+ ],
1128
+ "layers.7.attention.to_v.packed_weight_indices": [
1129
+ 7372800
1130
+ ],
1131
+ "layers.7.attention.to_v.row_norms": [
1132
+ 3840
1133
+ ],
1134
+ "layers.7.attention.to_out.0.packed_weight_indices": [
1135
+ 7372800
1136
+ ],
1137
+ "layers.7.attention.to_out.0.row_norms": [
1138
+ 3840
1139
+ ],
1140
+ "layers.7.feed_forward.w1.packed_weight_indices": [
1141
+ 19660800
1142
+ ],
1143
+ "layers.7.feed_forward.w1.row_norms": [
1144
+ 10240
1145
+ ],
1146
+ "layers.7.feed_forward.w2.packed_weight_indices": [
1147
+ 19660800
1148
+ ],
1149
+ "layers.7.feed_forward.w2.row_norms": [
1150
+ 3840
1151
+ ],
1152
+ "layers.7.feed_forward.w3.packed_weight_indices": [
1153
+ 19660800
1154
+ ],
1155
+ "layers.7.feed_forward.w3.row_norms": [
1156
+ 10240
1157
+ ],
1158
+ "layers.7.attention_norm1.weight": [
1159
+ 3840
1160
+ ],
1161
+ "layers.7.ffn_norm1.weight": [
1162
+ 3840
1163
+ ],
1164
+ "layers.7.attention_norm2.weight": [
1165
+ 3840
1166
+ ],
1167
+ "layers.7.ffn_norm2.weight": [
1168
+ 3840
1169
+ ],
1170
+ "layers.7.adaLN_modulation.0.bias": [
1171
+ 15360
1172
+ ],
1173
+ "layers.7.adaLN_modulation.0.packed_weight": [
1174
+ 1966080
1175
+ ],
1176
+ "layers.7.adaLN_modulation.0.scales": [
1177
+ 15360,
1178
+ 4
1179
+ ],
1180
+ "layers.8.attention.norm_q.weight": [
1181
+ 128
1182
+ ],
1183
+ "layers.8.attention.norm_k.weight": [
1184
+ 128
1185
+ ],
1186
+ "layers.8.attention.to_q.packed_weight_indices": [
1187
+ 7372800
1188
+ ],
1189
+ "layers.8.attention.to_q.row_norms": [
1190
+ 3840
1191
+ ],
1192
+ "layers.8.attention.to_k.packed_weight_indices": [
1193
+ 7372800
1194
+ ],
1195
+ "layers.8.attention.to_k.row_norms": [
1196
+ 3840
1197
+ ],
1198
+ "layers.8.attention.to_v.packed_weight_indices": [
1199
+ 7372800
1200
+ ],
1201
+ "layers.8.attention.to_v.row_norms": [
1202
+ 3840
1203
+ ],
1204
+ "layers.8.attention.to_out.0.packed_weight_indices": [
1205
+ 7372800
1206
+ ],
1207
+ "layers.8.attention.to_out.0.row_norms": [
1208
+ 3840
1209
+ ],
1210
+ "layers.8.feed_forward.w1.packed_weight_indices": [
1211
+ 19660800
1212
+ ],
1213
+ "layers.8.feed_forward.w1.row_norms": [
1214
+ 10240
1215
+ ],
1216
+ "layers.8.feed_forward.w2.packed_weight_indices": [
1217
+ 19660800
1218
+ ],
1219
+ "layers.8.feed_forward.w2.row_norms": [
1220
+ 3840
1221
+ ],
1222
+ "layers.8.feed_forward.w3.packed_weight_indices": [
1223
+ 19660800
1224
+ ],
1225
+ "layers.8.feed_forward.w3.row_norms": [
1226
+ 10240
1227
+ ],
1228
+ "layers.8.attention_norm1.weight": [
1229
+ 3840
1230
+ ],
1231
+ "layers.8.ffn_norm1.weight": [
1232
+ 3840
1233
+ ],
1234
+ "layers.8.attention_norm2.weight": [
1235
+ 3840
1236
+ ],
1237
+ "layers.8.ffn_norm2.weight": [
1238
+ 3840
1239
+ ],
1240
+ "layers.8.adaLN_modulation.0.bias": [
1241
+ 15360
1242
+ ],
1243
+ "layers.8.adaLN_modulation.0.packed_weight": [
1244
+ 1966080
1245
+ ],
1246
+ "layers.8.adaLN_modulation.0.scales": [
1247
+ 15360,
1248
+ 4
1249
+ ],
1250
+ "layers.9.attention.norm_q.weight": [
1251
+ 128
1252
+ ],
1253
+ "layers.9.attention.norm_k.weight": [
1254
+ 128
1255
+ ],
1256
+ "layers.9.attention.to_q.packed_weight_indices": [
1257
+ 7372800
1258
+ ],
1259
+ "layers.9.attention.to_q.row_norms": [
1260
+ 3840
1261
+ ],
1262
+ "layers.9.attention.to_k.packed_weight_indices": [
1263
+ 7372800
1264
+ ],
1265
+ "layers.9.attention.to_k.row_norms": [
1266
+ 3840
1267
+ ],
1268
+ "layers.9.attention.to_v.packed_weight_indices": [
1269
+ 7372800
1270
+ ],
1271
+ "layers.9.attention.to_v.row_norms": [
1272
+ 3840
1273
+ ],
1274
+ "layers.9.attention.to_out.0.packed_weight_indices": [
1275
+ 7372800
1276
+ ],
1277
+ "layers.9.attention.to_out.0.row_norms": [
1278
+ 3840
1279
+ ],
1280
+ "layers.9.feed_forward.w1.packed_weight_indices": [
1281
+ 19660800
1282
+ ],
1283
+ "layers.9.feed_forward.w1.row_norms": [
1284
+ 10240
1285
+ ],
1286
+ "layers.9.feed_forward.w2.packed_weight_indices": [
1287
+ 19660800
1288
+ ],
1289
+ "layers.9.feed_forward.w2.row_norms": [
1290
+ 3840
1291
+ ],
1292
+ "layers.9.feed_forward.w3.packed_weight_indices": [
1293
+ 19660800
1294
+ ],
1295
+ "layers.9.feed_forward.w3.row_norms": [
1296
+ 10240
1297
+ ],
1298
+ "layers.9.attention_norm1.weight": [
1299
+ 3840
1300
+ ],
1301
+ "layers.9.ffn_norm1.weight": [
1302
+ 3840
1303
+ ],
1304
+ "layers.9.attention_norm2.weight": [
1305
+ 3840
1306
+ ],
1307
+ "layers.9.ffn_norm2.weight": [
1308
+ 3840
1309
+ ],
1310
+ "layers.9.adaLN_modulation.0.bias": [
1311
+ 15360
1312
+ ],
1313
+ "layers.9.adaLN_modulation.0.packed_weight": [
1314
+ 1966080
1315
+ ],
1316
+ "layers.9.adaLN_modulation.0.scales": [
1317
+ 15360,
1318
+ 4
1319
+ ],
1320
+ "layers.10.attention.norm_q.weight": [
1321
+ 128
1322
+ ],
1323
+ "layers.10.attention.norm_k.weight": [
1324
+ 128
1325
+ ],
1326
+ "layers.10.attention.to_q.packed_weight_indices": [
1327
+ 7372800
1328
+ ],
1329
+ "layers.10.attention.to_q.row_norms": [
1330
+ 3840
1331
+ ],
1332
+ "layers.10.attention.to_k.packed_weight_indices": [
1333
+ 7372800
1334
+ ],
1335
+ "layers.10.attention.to_k.row_norms": [
1336
+ 3840
1337
+ ],
1338
+ "layers.10.attention.to_v.packed_weight_indices": [
1339
+ 7372800
1340
+ ],
1341
+ "layers.10.attention.to_v.row_norms": [
1342
+ 3840
1343
+ ],
1344
+ "layers.10.attention.to_out.0.packed_weight_indices": [
1345
+ 7372800
1346
+ ],
1347
+ "layers.10.attention.to_out.0.row_norms": [
1348
+ 3840
1349
+ ],
1350
+ "layers.10.feed_forward.w1.packed_weight_indices": [
1351
+ 19660800
1352
+ ],
1353
+ "layers.10.feed_forward.w1.row_norms": [
1354
+ 10240
1355
+ ],
1356
+ "layers.10.feed_forward.w2.packed_weight_indices": [
1357
+ 19660800
1358
+ ],
1359
+ "layers.10.feed_forward.w2.row_norms": [
1360
+ 3840
1361
+ ],
1362
+ "layers.10.feed_forward.w3.packed_weight_indices": [
1363
+ 19660800
1364
+ ],
1365
+ "layers.10.feed_forward.w3.row_norms": [
1366
+ 10240
1367
+ ],
1368
+ "layers.10.attention_norm1.weight": [
1369
+ 3840
1370
+ ],
1371
+ "layers.10.ffn_norm1.weight": [
1372
+ 3840
1373
+ ],
1374
+ "layers.10.attention_norm2.weight": [
1375
+ 3840
1376
+ ],
1377
+ "layers.10.ffn_norm2.weight": [
1378
+ 3840
1379
+ ],
1380
+ "layers.10.adaLN_modulation.0.bias": [
1381
+ 15360
1382
+ ],
1383
+ "layers.10.adaLN_modulation.0.packed_weight": [
1384
+ 1966080
1385
+ ],
1386
+ "layers.10.adaLN_modulation.0.scales": [
1387
+ 15360,
1388
+ 4
1389
+ ],
1390
+ "layers.11.attention.norm_q.weight": [
1391
+ 128
1392
+ ],
1393
+ "layers.11.attention.norm_k.weight": [
1394
+ 128
1395
+ ],
1396
+ "layers.11.attention.to_q.packed_weight_indices": [
1397
+ 7372800
1398
+ ],
1399
+ "layers.11.attention.to_q.row_norms": [
1400
+ 3840
1401
+ ],
1402
+ "layers.11.attention.to_k.packed_weight_indices": [
1403
+ 7372800
1404
+ ],
1405
+ "layers.11.attention.to_k.row_norms": [
1406
+ 3840
1407
+ ],
1408
+ "layers.11.attention.to_v.packed_weight_indices": [
1409
+ 7372800
1410
+ ],
1411
+ "layers.11.attention.to_v.row_norms": [
1412
+ 3840
1413
+ ],
1414
+ "layers.11.attention.to_out.0.packed_weight_indices": [
1415
+ 7372800
1416
+ ],
1417
+ "layers.11.attention.to_out.0.row_norms": [
1418
+ 3840
1419
+ ],
1420
+ "layers.11.feed_forward.w1.packed_weight_indices": [
1421
+ 19660800
1422
+ ],
1423
+ "layers.11.feed_forward.w1.row_norms": [
1424
+ 10240
1425
+ ],
1426
+ "layers.11.feed_forward.w2.packed_weight_indices": [
1427
+ 19660800
1428
+ ],
1429
+ "layers.11.feed_forward.w2.row_norms": [
1430
+ 3840
1431
+ ],
1432
+ "layers.11.feed_forward.w3.packed_weight_indices": [
1433
+ 19660800
1434
+ ],
1435
+ "layers.11.feed_forward.w3.row_norms": [
1436
+ 10240
1437
+ ],
1438
+ "layers.11.attention_norm1.weight": [
1439
+ 3840
1440
+ ],
1441
+ "layers.11.ffn_norm1.weight": [
1442
+ 3840
1443
+ ],
1444
+ "layers.11.attention_norm2.weight": [
1445
+ 3840
1446
+ ],
1447
+ "layers.11.ffn_norm2.weight": [
1448
+ 3840
1449
+ ],
1450
+ "layers.11.adaLN_modulation.0.bias": [
1451
+ 15360
1452
+ ],
1453
+ "layers.11.adaLN_modulation.0.packed_weight": [
1454
+ 1966080
1455
+ ],
1456
+ "layers.11.adaLN_modulation.0.scales": [
1457
+ 15360,
1458
+ 4
1459
+ ],
1460
+ "layers.12.attention.norm_q.weight": [
1461
+ 128
1462
+ ],
1463
+ "layers.12.attention.norm_k.weight": [
1464
+ 128
1465
+ ],
1466
+ "layers.12.attention.to_q.packed_weight_indices": [
1467
+ 7372800
1468
+ ],
1469
+ "layers.12.attention.to_q.row_norms": [
1470
+ 3840
1471
+ ],
1472
+ "layers.12.attention.to_k.packed_weight_indices": [
1473
+ 7372800
1474
+ ],
1475
+ "layers.12.attention.to_k.row_norms": [
1476
+ 3840
1477
+ ],
1478
+ "layers.12.attention.to_v.packed_weight_indices": [
1479
+ 7372800
1480
+ ],
1481
+ "layers.12.attention.to_v.row_norms": [
1482
+ 3840
1483
+ ],
1484
+ "layers.12.attention.to_out.0.packed_weight_indices": [
1485
+ 7372800
1486
+ ],
1487
+ "layers.12.attention.to_out.0.row_norms": [
1488
+ 3840
1489
+ ],
1490
+ "layers.12.feed_forward.w1.packed_weight_indices": [
1491
+ 19660800
1492
+ ],
1493
+ "layers.12.feed_forward.w1.row_norms": [
1494
+ 10240
1495
+ ],
1496
+ "layers.12.feed_forward.w2.packed_weight_indices": [
1497
+ 19660800
1498
+ ],
1499
+ "layers.12.feed_forward.w2.row_norms": [
1500
+ 3840
1501
+ ],
1502
+ "layers.12.feed_forward.w3.packed_weight_indices": [
1503
+ 19660800
1504
+ ],
1505
+ "layers.12.feed_forward.w3.row_norms": [
1506
+ 10240
1507
+ ],
1508
+ "layers.12.attention_norm1.weight": [
1509
+ 3840
1510
+ ],
1511
+ "layers.12.ffn_norm1.weight": [
1512
+ 3840
1513
+ ],
1514
+ "layers.12.attention_norm2.weight": [
1515
+ 3840
1516
+ ],
1517
+ "layers.12.ffn_norm2.weight": [
1518
+ 3840
1519
+ ],
1520
+ "layers.12.adaLN_modulation.0.bias": [
1521
+ 15360
1522
+ ],
1523
+ "layers.12.adaLN_modulation.0.packed_weight": [
1524
+ 1966080
1525
+ ],
1526
+ "layers.12.adaLN_modulation.0.scales": [
1527
+ 15360,
1528
+ 4
1529
+ ],
1530
+ "layers.13.attention.norm_q.weight": [
1531
+ 128
1532
+ ],
1533
+ "layers.13.attention.norm_k.weight": [
1534
+ 128
1535
+ ],
1536
+ "layers.13.attention.to_q.packed_weight_indices": [
1537
+ 7372800
1538
+ ],
1539
+ "layers.13.attention.to_q.row_norms": [
1540
+ 3840
1541
+ ],
1542
+ "layers.13.attention.to_k.packed_weight_indices": [
1543
+ 7372800
1544
+ ],
1545
+ "layers.13.attention.to_k.row_norms": [
1546
+ 3840
1547
+ ],
1548
+ "layers.13.attention.to_v.packed_weight_indices": [
1549
+ 7372800
1550
+ ],
1551
+ "layers.13.attention.to_v.row_norms": [
1552
+ 3840
1553
+ ],
1554
+ "layers.13.attention.to_out.0.packed_weight_indices": [
1555
+ 7372800
1556
+ ],
1557
+ "layers.13.attention.to_out.0.row_norms": [
1558
+ 3840
1559
+ ],
1560
+ "layers.13.feed_forward.w1.packed_weight_indices": [
1561
+ 19660800
1562
+ ],
1563
+ "layers.13.feed_forward.w1.row_norms": [
1564
+ 10240
1565
+ ],
1566
+ "layers.13.feed_forward.w2.packed_weight_indices": [
1567
+ 19660800
1568
+ ],
1569
+ "layers.13.feed_forward.w2.row_norms": [
1570
+ 3840
1571
+ ],
1572
+ "layers.13.feed_forward.w3.packed_weight_indices": [
1573
+ 19660800
1574
+ ],
1575
+ "layers.13.feed_forward.w3.row_norms": [
1576
+ 10240
1577
+ ],
1578
+ "layers.13.attention_norm1.weight": [
1579
+ 3840
1580
+ ],
1581
+ "layers.13.ffn_norm1.weight": [
1582
+ 3840
1583
+ ],
1584
+ "layers.13.attention_norm2.weight": [
1585
+ 3840
1586
+ ],
1587
+ "layers.13.ffn_norm2.weight": [
1588
+ 3840
1589
+ ],
1590
+ "layers.13.adaLN_modulation.0.bias": [
1591
+ 15360
1592
+ ],
1593
+ "layers.13.adaLN_modulation.0.packed_weight": [
1594
+ 1966080
1595
+ ],
1596
+ "layers.13.adaLN_modulation.0.scales": [
1597
+ 15360,
1598
+ 4
1599
+ ],
1600
+ "layers.14.attention.norm_q.weight": [
1601
+ 128
1602
+ ],
1603
+ "layers.14.attention.norm_k.weight": [
1604
+ 128
1605
+ ],
1606
+ "layers.14.attention.to_q.packed_weight_indices": [
1607
+ 7372800
1608
+ ],
1609
+ "layers.14.attention.to_q.row_norms": [
1610
+ 3840
1611
+ ],
1612
+ "layers.14.attention.to_k.packed_weight_indices": [
1613
+ 7372800
1614
+ ],
1615
+ "layers.14.attention.to_k.row_norms": [
1616
+ 3840
1617
+ ],
1618
+ "layers.14.attention.to_v.packed_weight_indices": [
1619
+ 7372800
1620
+ ],
1621
+ "layers.14.attention.to_v.row_norms": [
1622
+ 3840
1623
+ ],
1624
+ "layers.14.attention.to_out.0.packed_weight_indices": [
1625
+ 7372800
1626
+ ],
1627
+ "layers.14.attention.to_out.0.row_norms": [
1628
+ 3840
1629
+ ],
1630
+ "layers.14.feed_forward.w1.packed_weight_indices": [
1631
+ 19660800
1632
+ ],
1633
+ "layers.14.feed_forward.w1.row_norms": [
1634
+ 10240
1635
+ ],
1636
+ "layers.14.feed_forward.w2.packed_weight_indices": [
1637
+ 19660800
1638
+ ],
1639
+ "layers.14.feed_forward.w2.row_norms": [
1640
+ 3840
1641
+ ],
1642
+ "layers.14.feed_forward.w3.packed_weight_indices": [
1643
+ 19660800
1644
+ ],
1645
+ "layers.14.feed_forward.w3.row_norms": [
1646
+ 10240
1647
+ ],
1648
+ "layers.14.attention_norm1.weight": [
1649
+ 3840
1650
+ ],
1651
+ "layers.14.ffn_norm1.weight": [
1652
+ 3840
1653
+ ],
1654
+ "layers.14.attention_norm2.weight": [
1655
+ 3840
1656
+ ],
1657
+ "layers.14.ffn_norm2.weight": [
1658
+ 3840
1659
+ ],
1660
+ "layers.14.adaLN_modulation.0.bias": [
1661
+ 15360
1662
+ ],
1663
+ "layers.14.adaLN_modulation.0.packed_weight": [
1664
+ 1966080
1665
+ ],
1666
+ "layers.14.adaLN_modulation.0.scales": [
1667
+ 15360,
1668
+ 4
1669
+ ],
1670
+ "layers.15.attention.norm_q.weight": [
1671
+ 128
1672
+ ],
1673
+ "layers.15.attention.norm_k.weight": [
1674
+ 128
1675
+ ],
1676
+ "layers.15.attention.to_q.packed_weight_indices": [
1677
+ 7372800
1678
+ ],
1679
+ "layers.15.attention.to_q.row_norms": [
1680
+ 3840
1681
+ ],
1682
+ "layers.15.attention.to_k.packed_weight_indices": [
1683
+ 7372800
1684
+ ],
1685
+ "layers.15.attention.to_k.row_norms": [
1686
+ 3840
1687
+ ],
1688
+ "layers.15.attention.to_v.packed_weight_indices": [
1689
+ 7372800
1690
+ ],
1691
+ "layers.15.attention.to_v.row_norms": [
1692
+ 3840
1693
+ ],
1694
+ "layers.15.attention.to_out.0.packed_weight_indices": [
1695
+ 7372800
1696
+ ],
1697
+ "layers.15.attention.to_out.0.row_norms": [
1698
+ 3840
1699
+ ],
1700
+ "layers.15.feed_forward.w1.packed_weight_indices": [
1701
+ 19660800
1702
+ ],
1703
+ "layers.15.feed_forward.w1.row_norms": [
1704
+ 10240
1705
+ ],
1706
+ "layers.15.feed_forward.w2.packed_weight_indices": [
1707
+ 19660800
1708
+ ],
1709
+ "layers.15.feed_forward.w2.row_norms": [
1710
+ 3840
1711
+ ],
1712
+ "layers.15.feed_forward.w3.packed_weight_indices": [
1713
+ 19660800
1714
+ ],
1715
+ "layers.15.feed_forward.w3.row_norms": [
1716
+ 10240
1717
+ ],
1718
+ "layers.15.attention_norm1.weight": [
1719
+ 3840
1720
+ ],
1721
+ "layers.15.ffn_norm1.weight": [
1722
+ 3840
1723
+ ],
1724
+ "layers.15.attention_norm2.weight": [
1725
+ 3840
1726
+ ],
1727
+ "layers.15.ffn_norm2.weight": [
1728
+ 3840
1729
+ ],
1730
+ "layers.15.adaLN_modulation.0.bias": [
1731
+ 15360
1732
+ ],
1733
+ "layers.15.adaLN_modulation.0.packed_weight": [
1734
+ 1966080
1735
+ ],
1736
+ "layers.15.adaLN_modulation.0.scales": [
1737
+ 15360,
1738
+ 4
1739
+ ],
1740
+ "layers.16.attention.norm_q.weight": [
1741
+ 128
1742
+ ],
1743
+ "layers.16.attention.norm_k.weight": [
1744
+ 128
1745
+ ],
1746
+ "layers.16.attention.to_q.packed_weight_indices": [
1747
+ 7372800
1748
+ ],
1749
+ "layers.16.attention.to_q.row_norms": [
1750
+ 3840
1751
+ ],
1752
+ "layers.16.attention.to_k.packed_weight_indices": [
1753
+ 7372800
1754
+ ],
1755
+ "layers.16.attention.to_k.row_norms": [
1756
+ 3840
1757
+ ],
1758
+ "layers.16.attention.to_v.packed_weight_indices": [
1759
+ 7372800
1760
+ ],
1761
+ "layers.16.attention.to_v.row_norms": [
1762
+ 3840
1763
+ ],
1764
+ "layers.16.attention.to_out.0.packed_weight_indices": [
1765
+ 7372800
1766
+ ],
1767
+ "layers.16.attention.to_out.0.row_norms": [
1768
+ 3840
1769
+ ],
1770
+ "layers.16.feed_forward.w1.packed_weight_indices": [
1771
+ 19660800
1772
+ ],
1773
+ "layers.16.feed_forward.w1.row_norms": [
1774
+ 10240
1775
+ ],
1776
+ "layers.16.feed_forward.w2.packed_weight_indices": [
1777
+ 19660800
1778
+ ],
1779
+ "layers.16.feed_forward.w2.row_norms": [
1780
+ 3840
1781
+ ],
1782
+ "layers.16.feed_forward.w3.packed_weight_indices": [
1783
+ 19660800
1784
+ ],
1785
+ "layers.16.feed_forward.w3.row_norms": [
1786
+ 10240
1787
+ ],
1788
+ "layers.16.attention_norm1.weight": [
1789
+ 3840
1790
+ ],
1791
+ "layers.16.ffn_norm1.weight": [
1792
+ 3840
1793
+ ],
1794
+ "layers.16.attention_norm2.weight": [
1795
+ 3840
1796
+ ],
1797
+ "layers.16.ffn_norm2.weight": [
1798
+ 3840
1799
+ ],
1800
+ "layers.16.adaLN_modulation.0.bias": [
1801
+ 15360
1802
+ ],
1803
+ "layers.16.adaLN_modulation.0.packed_weight": [
1804
+ 1966080
1805
+ ],
1806
+ "layers.16.adaLN_modulation.0.scales": [
1807
+ 15360,
1808
+ 4
1809
+ ],
1810
+ "layers.17.attention.norm_q.weight": [
1811
+ 128
1812
+ ],
1813
+ "layers.17.attention.norm_k.weight": [
1814
+ 128
1815
+ ],
1816
+ "layers.17.attention.to_q.packed_weight_indices": [
1817
+ 7372800
1818
+ ],
1819
+ "layers.17.attention.to_q.row_norms": [
1820
+ 3840
1821
+ ],
1822
+ "layers.17.attention.to_k.packed_weight_indices": [
1823
+ 7372800
1824
+ ],
1825
+ "layers.17.attention.to_k.row_norms": [
1826
+ 3840
1827
+ ],
1828
+ "layers.17.attention.to_v.packed_weight_indices": [
1829
+ 7372800
1830
+ ],
1831
+ "layers.17.attention.to_v.row_norms": [
1832
+ 3840
1833
+ ],
1834
+ "layers.17.attention.to_out.0.packed_weight_indices": [
1835
+ 7372800
1836
+ ],
1837
+ "layers.17.attention.to_out.0.row_norms": [
1838
+ 3840
1839
+ ],
1840
+ "layers.17.feed_forward.w1.packed_weight_indices": [
1841
+ 19660800
1842
+ ],
1843
+ "layers.17.feed_forward.w1.row_norms": [
1844
+ 10240
1845
+ ],
1846
+ "layers.17.feed_forward.w2.packed_weight_indices": [
1847
+ 19660800
1848
+ ],
1849
+ "layers.17.feed_forward.w2.row_norms": [
1850
+ 3840
1851
+ ],
1852
+ "layers.17.feed_forward.w3.packed_weight_indices": [
1853
+ 19660800
1854
+ ],
1855
+ "layers.17.feed_forward.w3.row_norms": [
1856
+ 10240
1857
+ ],
1858
+ "layers.17.attention_norm1.weight": [
1859
+ 3840
1860
+ ],
1861
+ "layers.17.ffn_norm1.weight": [
1862
+ 3840
1863
+ ],
1864
+ "layers.17.attention_norm2.weight": [
1865
+ 3840
1866
+ ],
1867
+ "layers.17.ffn_norm2.weight": [
1868
+ 3840
1869
+ ],
1870
+ "layers.17.adaLN_modulation.0.bias": [
1871
+ 15360
1872
+ ],
1873
+ "layers.17.adaLN_modulation.0.packed_weight": [
1874
+ 1966080
1875
+ ],
1876
+ "layers.17.adaLN_modulation.0.scales": [
1877
+ 15360,
1878
+ 4
1879
+ ],
1880
+ "layers.18.attention.norm_q.weight": [
1881
+ 128
1882
+ ],
1883
+ "layers.18.attention.norm_k.weight": [
1884
+ 128
1885
+ ],
1886
+ "layers.18.attention.to_q.packed_weight_indices": [
1887
+ 7372800
1888
+ ],
1889
+ "layers.18.attention.to_q.row_norms": [
1890
+ 3840
1891
+ ],
1892
+ "layers.18.attention.to_k.packed_weight_indices": [
1893
+ 7372800
1894
+ ],
1895
+ "layers.18.attention.to_k.row_norms": [
1896
+ 3840
1897
+ ],
1898
+ "layers.18.attention.to_v.packed_weight_indices": [
1899
+ 7372800
1900
+ ],
1901
+ "layers.18.attention.to_v.row_norms": [
1902
+ 3840
1903
+ ],
1904
+ "layers.18.attention.to_out.0.packed_weight_indices": [
1905
+ 7372800
1906
+ ],
1907
+ "layers.18.attention.to_out.0.row_norms": [
1908
+ 3840
1909
+ ],
1910
+ "layers.18.feed_forward.w1.packed_weight_indices": [
1911
+ 19660800
1912
+ ],
1913
+ "layers.18.feed_forward.w1.row_norms": [
1914
+ 10240
1915
+ ],
1916
+ "layers.18.feed_forward.w2.packed_weight_indices": [
1917
+ 19660800
1918
+ ],
1919
+ "layers.18.feed_forward.w2.row_norms": [
1920
+ 3840
1921
+ ],
1922
+ "layers.18.feed_forward.w3.packed_weight_indices": [
1923
+ 19660800
1924
+ ],
1925
+ "layers.18.feed_forward.w3.row_norms": [
1926
+ 10240
1927
+ ],
1928
+ "layers.18.attention_norm1.weight": [
1929
+ 3840
1930
+ ],
1931
+ "layers.18.ffn_norm1.weight": [
1932
+ 3840
1933
+ ],
1934
+ "layers.18.attention_norm2.weight": [
1935
+ 3840
1936
+ ],
1937
+ "layers.18.ffn_norm2.weight": [
1938
+ 3840
1939
+ ],
1940
+ "layers.18.adaLN_modulation.0.bias": [
1941
+ 15360
1942
+ ],
1943
+ "layers.18.adaLN_modulation.0.packed_weight": [
1944
+ 1966080
1945
+ ],
1946
+ "layers.18.adaLN_modulation.0.scales": [
1947
+ 15360,
1948
+ 4
1949
+ ],
1950
+ "layers.19.attention.norm_q.weight": [
1951
+ 128
1952
+ ],
1953
+ "layers.19.attention.norm_k.weight": [
1954
+ 128
1955
+ ],
1956
+ "layers.19.attention.to_q.packed_weight_indices": [
1957
+ 7372800
1958
+ ],
1959
+ "layers.19.attention.to_q.row_norms": [
1960
+ 3840
1961
+ ],
1962
+ "layers.19.attention.to_k.packed_weight_indices": [
1963
+ 7372800
1964
+ ],
1965
+ "layers.19.attention.to_k.row_norms": [
1966
+ 3840
1967
+ ],
1968
+ "layers.19.attention.to_v.packed_weight_indices": [
1969
+ 7372800
1970
+ ],
1971
+ "layers.19.attention.to_v.row_norms": [
1972
+ 3840
1973
+ ],
1974
+ "layers.19.attention.to_out.0.packed_weight_indices": [
1975
+ 7372800
1976
+ ],
1977
+ "layers.19.attention.to_out.0.row_norms": [
1978
+ 3840
1979
+ ],
1980
+ "layers.19.feed_forward.w1.packed_weight_indices": [
1981
+ 19660800
1982
+ ],
1983
+ "layers.19.feed_forward.w1.row_norms": [
1984
+ 10240
1985
+ ],
1986
+ "layers.19.feed_forward.w2.packed_weight_indices": [
1987
+ 19660800
1988
+ ],
1989
+ "layers.19.feed_forward.w2.row_norms": [
1990
+ 3840
1991
+ ],
1992
+ "layers.19.feed_forward.w3.packed_weight_indices": [
1993
+ 19660800
1994
+ ],
1995
+ "layers.19.feed_forward.w3.row_norms": [
1996
+ 10240
1997
+ ],
1998
+ "layers.19.attention_norm1.weight": [
1999
+ 3840
2000
+ ],
2001
+ "layers.19.ffn_norm1.weight": [
2002
+ 3840
2003
+ ],
2004
+ "layers.19.attention_norm2.weight": [
2005
+ 3840
2006
+ ],
2007
+ "layers.19.ffn_norm2.weight": [
2008
+ 3840
2009
+ ],
2010
+ "layers.19.adaLN_modulation.0.bias": [
2011
+ 15360
2012
+ ],
2013
+ "layers.19.adaLN_modulation.0.packed_weight": [
2014
+ 1966080
2015
+ ],
2016
+ "layers.19.adaLN_modulation.0.scales": [
2017
+ 15360,
2018
+ 4
2019
+ ],
2020
+ "layers.20.attention.norm_q.weight": [
2021
+ 128
2022
+ ],
2023
+ "layers.20.attention.norm_k.weight": [
2024
+ 128
2025
+ ],
2026
+ "layers.20.attention.to_q.packed_weight_indices": [
2027
+ 7372800
2028
+ ],
2029
+ "layers.20.attention.to_q.row_norms": [
2030
+ 3840
2031
+ ],
2032
+ "layers.20.attention.to_k.packed_weight_indices": [
2033
+ 7372800
2034
+ ],
2035
+ "layers.20.attention.to_k.row_norms": [
2036
+ 3840
2037
+ ],
2038
+ "layers.20.attention.to_v.packed_weight_indices": [
2039
+ 7372800
2040
+ ],
2041
+ "layers.20.attention.to_v.row_norms": [
2042
+ 3840
2043
+ ],
2044
+ "layers.20.attention.to_out.0.packed_weight_indices": [
2045
+ 7372800
2046
+ ],
2047
+ "layers.20.attention.to_out.0.row_norms": [
2048
+ 3840
2049
+ ],
2050
+ "layers.20.feed_forward.w1.packed_weight_indices": [
2051
+ 19660800
2052
+ ],
2053
+ "layers.20.feed_forward.w1.row_norms": [
2054
+ 10240
2055
+ ],
2056
+ "layers.20.feed_forward.w2.packed_weight_indices": [
2057
+ 19660800
2058
+ ],
2059
+ "layers.20.feed_forward.w2.row_norms": [
2060
+ 3840
2061
+ ],
2062
+ "layers.20.feed_forward.w3.packed_weight_indices": [
2063
+ 19660800
2064
+ ],
2065
+ "layers.20.feed_forward.w3.row_norms": [
2066
+ 10240
2067
+ ],
2068
+ "layers.20.attention_norm1.weight": [
2069
+ 3840
2070
+ ],
2071
+ "layers.20.ffn_norm1.weight": [
2072
+ 3840
2073
+ ],
2074
+ "layers.20.attention_norm2.weight": [
2075
+ 3840
2076
+ ],
2077
+ "layers.20.ffn_norm2.weight": [
2078
+ 3840
2079
+ ],
2080
+ "layers.20.adaLN_modulation.0.bias": [
2081
+ 15360
2082
+ ],
2083
+ "layers.20.adaLN_modulation.0.packed_weight": [
2084
+ 1966080
2085
+ ],
2086
+ "layers.20.adaLN_modulation.0.scales": [
2087
+ 15360,
2088
+ 4
2089
+ ],
2090
+ "layers.21.attention.norm_q.weight": [
2091
+ 128
2092
+ ],
2093
+ "layers.21.attention.norm_k.weight": [
2094
+ 128
2095
+ ],
2096
+ "layers.21.attention.to_q.packed_weight_indices": [
2097
+ 7372800
2098
+ ],
2099
+ "layers.21.attention.to_q.row_norms": [
2100
+ 3840
2101
+ ],
2102
+ "layers.21.attention.to_k.packed_weight_indices": [
2103
+ 7372800
2104
+ ],
2105
+ "layers.21.attention.to_k.row_norms": [
2106
+ 3840
2107
+ ],
2108
+ "layers.21.attention.to_v.packed_weight_indices": [
2109
+ 7372800
2110
+ ],
2111
+ "layers.21.attention.to_v.row_norms": [
2112
+ 3840
2113
+ ],
2114
+ "layers.21.attention.to_out.0.packed_weight_indices": [
2115
+ 7372800
2116
+ ],
2117
+ "layers.21.attention.to_out.0.row_norms": [
2118
+ 3840
2119
+ ],
2120
+ "layers.21.feed_forward.w1.packed_weight_indices": [
2121
+ 19660800
2122
+ ],
2123
+ "layers.21.feed_forward.w1.row_norms": [
2124
+ 10240
2125
+ ],
2126
+ "layers.21.feed_forward.w2.packed_weight_indices": [
2127
+ 19660800
2128
+ ],
2129
+ "layers.21.feed_forward.w2.row_norms": [
2130
+ 3840
2131
+ ],
2132
+ "layers.21.feed_forward.w3.packed_weight_indices": [
2133
+ 19660800
2134
+ ],
2135
+ "layers.21.feed_forward.w3.row_norms": [
2136
+ 10240
2137
+ ],
2138
+ "layers.21.attention_norm1.weight": [
2139
+ 3840
2140
+ ],
2141
+ "layers.21.ffn_norm1.weight": [
2142
+ 3840
2143
+ ],
2144
+ "layers.21.attention_norm2.weight": [
2145
+ 3840
2146
+ ],
2147
+ "layers.21.ffn_norm2.weight": [
2148
+ 3840
2149
+ ],
2150
+ "layers.21.adaLN_modulation.0.bias": [
2151
+ 15360
2152
+ ],
2153
+ "layers.21.adaLN_modulation.0.packed_weight": [
2154
+ 1966080
2155
+ ],
2156
+ "layers.21.adaLN_modulation.0.scales": [
2157
+ 15360,
2158
+ 4
2159
+ ],
2160
+ "layers.22.attention.norm_q.weight": [
2161
+ 128
2162
+ ],
2163
+ "layers.22.attention.norm_k.weight": [
2164
+ 128
2165
+ ],
2166
+ "layers.22.attention.to_q.packed_weight_indices": [
2167
+ 7372800
2168
+ ],
2169
+ "layers.22.attention.to_q.row_norms": [
2170
+ 3840
2171
+ ],
2172
+ "layers.22.attention.to_k.packed_weight_indices": [
2173
+ 7372800
2174
+ ],
2175
+ "layers.22.attention.to_k.row_norms": [
2176
+ 3840
2177
+ ],
2178
+ "layers.22.attention.to_v.packed_weight_indices": [
2179
+ 7372800
2180
+ ],
2181
+ "layers.22.attention.to_v.row_norms": [
2182
+ 3840
2183
+ ],
2184
+ "layers.22.attention.to_out.0.packed_weight_indices": [
2185
+ 7372800
2186
+ ],
2187
+ "layers.22.attention.to_out.0.row_norms": [
2188
+ 3840
2189
+ ],
2190
+ "layers.22.feed_forward.w1.packed_weight_indices": [
2191
+ 19660800
2192
+ ],
2193
+ "layers.22.feed_forward.w1.row_norms": [
2194
+ 10240
2195
+ ],
2196
+ "layers.22.feed_forward.w2.packed_weight_indices": [
2197
+ 19660800
2198
+ ],
2199
+ "layers.22.feed_forward.w2.row_norms": [
2200
+ 3840
2201
+ ],
2202
+ "layers.22.feed_forward.w3.packed_weight_indices": [
2203
+ 19660800
2204
+ ],
2205
+ "layers.22.feed_forward.w3.row_norms": [
2206
+ 10240
2207
+ ],
2208
+ "layers.22.attention_norm1.weight": [
2209
+ 3840
2210
+ ],
2211
+ "layers.22.ffn_norm1.weight": [
2212
+ 3840
2213
+ ],
2214
+ "layers.22.attention_norm2.weight": [
2215
+ 3840
2216
+ ],
2217
+ "layers.22.ffn_norm2.weight": [
2218
+ 3840
2219
+ ],
2220
+ "layers.22.adaLN_modulation.0.bias": [
2221
+ 15360
2222
+ ],
2223
+ "layers.22.adaLN_modulation.0.packed_weight": [
2224
+ 1966080
2225
+ ],
2226
+ "layers.22.adaLN_modulation.0.scales": [
2227
+ 15360,
2228
+ 4
2229
+ ],
2230
+ "layers.23.attention.norm_q.weight": [
2231
+ 128
2232
+ ],
2233
+ "layers.23.attention.norm_k.weight": [
2234
+ 128
2235
+ ],
2236
+ "layers.23.attention.to_q.packed_weight_indices": [
2237
+ 7372800
2238
+ ],
2239
+ "layers.23.attention.to_q.row_norms": [
2240
+ 3840
2241
+ ],
2242
+ "layers.23.attention.to_k.packed_weight_indices": [
2243
+ 7372800
2244
+ ],
2245
+ "layers.23.attention.to_k.row_norms": [
2246
+ 3840
2247
+ ],
2248
+ "layers.23.attention.to_v.packed_weight_indices": [
2249
+ 7372800
2250
+ ],
2251
+ "layers.23.attention.to_v.row_norms": [
2252
+ 3840
2253
+ ],
2254
+ "layers.23.attention.to_out.0.packed_weight_indices": [
2255
+ 7372800
2256
+ ],
2257
+ "layers.23.attention.to_out.0.row_norms": [
2258
+ 3840
2259
+ ],
2260
+ "layers.23.feed_forward.w1.packed_weight_indices": [
2261
+ 19660800
2262
+ ],
2263
+ "layers.23.feed_forward.w1.row_norms": [
2264
+ 10240
2265
+ ],
2266
+ "layers.23.feed_forward.w2.packed_weight_indices": [
2267
+ 19660800
2268
+ ],
2269
+ "layers.23.feed_forward.w2.row_norms": [
2270
+ 3840
2271
+ ],
2272
+ "layers.23.feed_forward.w3.packed_weight_indices": [
2273
+ 19660800
2274
+ ],
2275
+ "layers.23.feed_forward.w3.row_norms": [
2276
+ 10240
2277
+ ],
2278
+ "layers.23.attention_norm1.weight": [
2279
+ 3840
2280
+ ],
2281
+ "layers.23.ffn_norm1.weight": [
2282
+ 3840
2283
+ ],
2284
+ "layers.23.attention_norm2.weight": [
2285
+ 3840
2286
+ ],
2287
+ "layers.23.ffn_norm2.weight": [
2288
+ 3840
2289
+ ],
2290
+ "layers.23.adaLN_modulation.0.bias": [
2291
+ 15360
2292
+ ],
2293
+ "layers.23.adaLN_modulation.0.packed_weight": [
2294
+ 1966080
2295
+ ],
2296
+ "layers.23.adaLN_modulation.0.scales": [
2297
+ 15360,
2298
+ 4
2299
+ ],
2300
+ "layers.24.attention.norm_q.weight": [
2301
+ 128
2302
+ ],
2303
+ "layers.24.attention.norm_k.weight": [
2304
+ 128
2305
+ ],
2306
+ "layers.24.attention.to_q.packed_weight_indices": [
2307
+ 7372800
2308
+ ],
2309
+ "layers.24.attention.to_q.row_norms": [
2310
+ 3840
2311
+ ],
2312
+ "layers.24.attention.to_k.packed_weight_indices": [
2313
+ 7372800
2314
+ ],
2315
+ "layers.24.attention.to_k.row_norms": [
2316
+ 3840
2317
+ ],
2318
+ "layers.24.attention.to_v.packed_weight_indices": [
2319
+ 7372800
2320
+ ],
2321
+ "layers.24.attention.to_v.row_norms": [
2322
+ 3840
2323
+ ],
2324
+ "layers.24.attention.to_out.0.packed_weight_indices": [
2325
+ 7372800
2326
+ ],
2327
+ "layers.24.attention.to_out.0.row_norms": [
2328
+ 3840
2329
+ ],
2330
+ "layers.24.feed_forward.w1.packed_weight_indices": [
2331
+ 19660800
2332
+ ],
2333
+ "layers.24.feed_forward.w1.row_norms": [
2334
+ 10240
2335
+ ],
2336
+ "layers.24.feed_forward.w2.packed_weight_indices": [
2337
+ 19660800
2338
+ ],
2339
+ "layers.24.feed_forward.w2.row_norms": [
2340
+ 3840
2341
+ ],
2342
+ "layers.24.feed_forward.w3.packed_weight_indices": [
2343
+ 19660800
2344
+ ],
2345
+ "layers.24.feed_forward.w3.row_norms": [
2346
+ 10240
2347
+ ],
2348
+ "layers.24.attention_norm1.weight": [
2349
+ 3840
2350
+ ],
2351
+ "layers.24.ffn_norm1.weight": [
2352
+ 3840
2353
+ ],
2354
+ "layers.24.attention_norm2.weight": [
2355
+ 3840
2356
+ ],
2357
+ "layers.24.ffn_norm2.weight": [
2358
+ 3840
2359
+ ],
2360
+ "layers.24.adaLN_modulation.0.bias": [
2361
+ 15360
2362
+ ],
2363
+ "layers.24.adaLN_modulation.0.packed_weight": [
2364
+ 1966080
2365
+ ],
2366
+ "layers.24.adaLN_modulation.0.scales": [
2367
+ 15360,
2368
+ 4
2369
+ ],
2370
+ "layers.25.attention.norm_q.weight": [
2371
+ 128
2372
+ ],
2373
+ "layers.25.attention.norm_k.weight": [
2374
+ 128
2375
+ ],
2376
+ "layers.25.attention.to_q.packed_weight_indices": [
2377
+ 7372800
2378
+ ],
2379
+ "layers.25.attention.to_q.row_norms": [
2380
+ 3840
2381
+ ],
2382
+ "layers.25.attention.to_k.packed_weight_indices": [
2383
+ 7372800
2384
+ ],
2385
+ "layers.25.attention.to_k.row_norms": [
2386
+ 3840
2387
+ ],
2388
+ "layers.25.attention.to_v.packed_weight_indices": [
2389
+ 7372800
2390
+ ],
2391
+ "layers.25.attention.to_v.row_norms": [
2392
+ 3840
2393
+ ],
2394
+ "layers.25.attention.to_out.0.packed_weight_indices": [
2395
+ 7372800
2396
+ ],
2397
+ "layers.25.attention.to_out.0.row_norms": [
2398
+ 3840
2399
+ ],
2400
+ "layers.25.feed_forward.w1.packed_weight_indices": [
2401
+ 19660800
2402
+ ],
2403
+ "layers.25.feed_forward.w1.row_norms": [
2404
+ 10240
2405
+ ],
2406
+ "layers.25.feed_forward.w2.packed_weight_indices": [
2407
+ 19660800
2408
+ ],
2409
+ "layers.25.feed_forward.w2.row_norms": [
2410
+ 3840
2411
+ ],
2412
+ "layers.25.feed_forward.w3.packed_weight_indices": [
2413
+ 19660800
2414
+ ],
2415
+ "layers.25.feed_forward.w3.row_norms": [
2416
+ 10240
2417
+ ],
2418
+ "layers.25.attention_norm1.weight": [
2419
+ 3840
2420
+ ],
2421
+ "layers.25.ffn_norm1.weight": [
2422
+ 3840
2423
+ ],
2424
+ "layers.25.attention_norm2.weight": [
2425
+ 3840
2426
+ ],
2427
+ "layers.25.ffn_norm2.weight": [
2428
+ 3840
2429
+ ],
2430
+ "layers.25.adaLN_modulation.0.bias": [
2431
+ 15360
2432
+ ],
2433
+ "layers.25.adaLN_modulation.0.packed_weight": [
2434
+ 1966080
2435
+ ],
2436
+ "layers.25.adaLN_modulation.0.scales": [
2437
+ 15360,
2438
+ 4
2439
+ ],
2440
+ "layers.26.attention.norm_q.weight": [
2441
+ 128
2442
+ ],
2443
+ "layers.26.attention.norm_k.weight": [
2444
+ 128
2445
+ ],
2446
+ "layers.26.attention.to_q.packed_weight_indices": [
2447
+ 7372800
2448
+ ],
2449
+ "layers.26.attention.to_q.row_norms": [
2450
+ 3840
2451
+ ],
2452
+ "layers.26.attention.to_k.packed_weight_indices": [
2453
+ 7372800
2454
+ ],
2455
+ "layers.26.attention.to_k.row_norms": [
2456
+ 3840
2457
+ ],
2458
+ "layers.26.attention.to_v.packed_weight_indices": [
2459
+ 7372800
2460
+ ],
2461
+ "layers.26.attention.to_v.row_norms": [
2462
+ 3840
2463
+ ],
2464
+ "layers.26.attention.to_out.0.packed_weight_indices": [
2465
+ 7372800
2466
+ ],
2467
+ "layers.26.attention.to_out.0.row_norms": [
2468
+ 3840
2469
+ ],
2470
+ "layers.26.feed_forward.w1.packed_weight_indices": [
2471
+ 19660800
2472
+ ],
2473
+ "layers.26.feed_forward.w1.row_norms": [
2474
+ 10240
2475
+ ],
2476
+ "layers.26.feed_forward.w2.packed_weight_indices": [
2477
+ 19660800
2478
+ ],
2479
+ "layers.26.feed_forward.w2.row_norms": [
2480
+ 3840
2481
+ ],
2482
+ "layers.26.feed_forward.w3.packed_weight_indices": [
2483
+ 19660800
2484
+ ],
2485
+ "layers.26.feed_forward.w3.row_norms": [
2486
+ 10240
2487
+ ],
2488
+ "layers.26.attention_norm1.weight": [
2489
+ 3840
2490
+ ],
2491
+ "layers.26.ffn_norm1.weight": [
2492
+ 3840
2493
+ ],
2494
+ "layers.26.attention_norm2.weight": [
2495
+ 3840
2496
+ ],
2497
+ "layers.26.ffn_norm2.weight": [
2498
+ 3840
2499
+ ],
2500
+ "layers.26.adaLN_modulation.0.bias": [
2501
+ 15360
2502
+ ],
2503
+ "layers.26.adaLN_modulation.0.packed_weight": [
2504
+ 1966080
2505
+ ],
2506
+ "layers.26.adaLN_modulation.0.scales": [
2507
+ 15360,
2508
+ 4
2509
+ ],
2510
+ "layers.27.attention.norm_q.weight": [
2511
+ 128
2512
+ ],
2513
+ "layers.27.attention.norm_k.weight": [
2514
+ 128
2515
+ ],
2516
+ "layers.27.attention.to_q.packed_weight_indices": [
2517
+ 7372800
2518
+ ],
2519
+ "layers.27.attention.to_q.row_norms": [
2520
+ 3840
2521
+ ],
2522
+ "layers.27.attention.to_k.packed_weight_indices": [
2523
+ 7372800
2524
+ ],
2525
+ "layers.27.attention.to_k.row_norms": [
2526
+ 3840
2527
+ ],
2528
+ "layers.27.attention.to_v.packed_weight_indices": [
2529
+ 7372800
2530
+ ],
2531
+ "layers.27.attention.to_v.row_norms": [
2532
+ 3840
2533
+ ],
2534
+ "layers.27.attention.to_out.0.packed_weight_indices": [
2535
+ 7372800
2536
+ ],
2537
+ "layers.27.attention.to_out.0.row_norms": [
2538
+ 3840
2539
+ ],
2540
+ "layers.27.feed_forward.w1.packed_weight_indices": [
2541
+ 19660800
2542
+ ],
2543
+ "layers.27.feed_forward.w1.row_norms": [
2544
+ 10240
2545
+ ],
2546
+ "layers.27.feed_forward.w2.packed_weight_indices": [
2547
+ 19660800
2548
+ ],
2549
+ "layers.27.feed_forward.w2.row_norms": [
2550
+ 3840
2551
+ ],
2552
+ "layers.27.feed_forward.w3.packed_weight_indices": [
2553
+ 19660800
2554
+ ],
2555
+ "layers.27.feed_forward.w3.row_norms": [
2556
+ 10240
2557
+ ],
2558
+ "layers.27.attention_norm1.weight": [
2559
+ 3840
2560
+ ],
2561
+ "layers.27.ffn_norm1.weight": [
2562
+ 3840
2563
+ ],
2564
+ "layers.27.attention_norm2.weight": [
2565
+ 3840
2566
+ ],
2567
+ "layers.27.ffn_norm2.weight": [
2568
+ 3840
2569
+ ],
2570
+ "layers.27.adaLN_modulation.0.bias": [
2571
+ 15360
2572
+ ],
2573
+ "layers.27.adaLN_modulation.0.packed_weight": [
2574
+ 1966080
2575
+ ],
2576
+ "layers.27.adaLN_modulation.0.scales": [
2577
+ 15360,
2578
+ 4
2579
+ ],
2580
+ "layers.28.attention.norm_q.weight": [
2581
+ 128
2582
+ ],
2583
+ "layers.28.attention.norm_k.weight": [
2584
+ 128
2585
+ ],
2586
+ "layers.28.attention.to_q.packed_weight_indices": [
2587
+ 7372800
2588
+ ],
2589
+ "layers.28.attention.to_q.row_norms": [
2590
+ 3840
2591
+ ],
2592
+ "layers.28.attention.to_k.packed_weight_indices": [
2593
+ 7372800
2594
+ ],
2595
+ "layers.28.attention.to_k.row_norms": [
2596
+ 3840
2597
+ ],
2598
+ "layers.28.attention.to_v.packed_weight_indices": [
2599
+ 7372800
2600
+ ],
2601
+ "layers.28.attention.to_v.row_norms": [
2602
+ 3840
2603
+ ],
2604
+ "layers.28.attention.to_out.0.packed_weight_indices": [
2605
+ 7372800
2606
+ ],
2607
+ "layers.28.attention.to_out.0.row_norms": [
2608
+ 3840
2609
+ ],
2610
+ "layers.28.feed_forward.w1.packed_weight_indices": [
2611
+ 19660800
2612
+ ],
2613
+ "layers.28.feed_forward.w1.row_norms": [
2614
+ 10240
2615
+ ],
2616
+ "layers.28.feed_forward.w2.packed_weight_indices": [
2617
+ 19660800
2618
+ ],
2619
+ "layers.28.feed_forward.w2.row_norms": [
2620
+ 3840
2621
+ ],
2622
+ "layers.28.feed_forward.w3.packed_weight_indices": [
2623
+ 19660800
2624
+ ],
2625
+ "layers.28.feed_forward.w3.row_norms": [
2626
+ 10240
2627
+ ],
2628
+ "layers.28.attention_norm1.weight": [
2629
+ 3840
2630
+ ],
2631
+ "layers.28.ffn_norm1.weight": [
2632
+ 3840
2633
+ ],
2634
+ "layers.28.attention_norm2.weight": [
2635
+ 3840
2636
+ ],
2637
+ "layers.28.ffn_norm2.weight": [
2638
+ 3840
2639
+ ],
2640
+ "layers.28.adaLN_modulation.0.bias": [
2641
+ 15360
2642
+ ],
2643
+ "layers.28.adaLN_modulation.0.packed_weight": [
2644
+ 1966080
2645
+ ],
2646
+ "layers.28.adaLN_modulation.0.scales": [
2647
+ 15360,
2648
+ 4
2649
+ ],
2650
+ "layers.29.attention.norm_q.weight": [
2651
+ 128
2652
+ ],
2653
+ "layers.29.attention.norm_k.weight": [
2654
+ 128
2655
+ ],
2656
+ "layers.29.attention.to_q.packed_weight_indices": [
2657
+ 7372800
2658
+ ],
2659
+ "layers.29.attention.to_q.row_norms": [
2660
+ 3840
2661
+ ],
2662
+ "layers.29.attention.to_k.packed_weight_indices": [
2663
+ 7372800
2664
+ ],
2665
+ "layers.29.attention.to_k.row_norms": [
2666
+ 3840
2667
+ ],
2668
+ "layers.29.attention.to_v.packed_weight_indices": [
2669
+ 7372800
2670
+ ],
2671
+ "layers.29.attention.to_v.row_norms": [
2672
+ 3840
2673
+ ],
2674
+ "layers.29.attention.to_out.0.packed_weight_indices": [
2675
+ 7372800
2676
+ ],
2677
+ "layers.29.attention.to_out.0.row_norms": [
2678
+ 3840
2679
+ ],
2680
+ "layers.29.feed_forward.w1.packed_weight_indices": [
2681
+ 19660800
2682
+ ],
2683
+ "layers.29.feed_forward.w1.row_norms": [
2684
+ 10240
2685
+ ],
2686
+ "layers.29.feed_forward.w2.packed_weight_indices": [
2687
+ 19660800
2688
+ ],
2689
+ "layers.29.feed_forward.w2.row_norms": [
2690
+ 3840
2691
+ ],
2692
+ "layers.29.feed_forward.w3.packed_weight_indices": [
2693
+ 19660800
2694
+ ],
2695
+ "layers.29.feed_forward.w3.row_norms": [
2696
+ 10240
2697
+ ],
2698
+ "layers.29.attention_norm1.weight": [
2699
+ 3840
2700
+ ],
2701
+ "layers.29.ffn_norm1.weight": [
2702
+ 3840
2703
+ ],
2704
+ "layers.29.attention_norm2.weight": [
2705
+ 3840
2706
+ ],
2707
+ "layers.29.ffn_norm2.weight": [
2708
+ 3840
2709
+ ],
2710
+ "layers.29.adaLN_modulation.0.bias": [
2711
+ 15360
2712
+ ],
2713
+ "layers.29.adaLN_modulation.0.packed_weight": [
2714
+ 1966080
2715
+ ],
2716
+ "layers.29.adaLN_modulation.0.scales": [
2717
+ 15360,
2718
+ 4
2719
+ ]
2720
+ },
2721
+ "checksums": {
2722
+ "model.safetensors": "0181f43913c3c647ec1adeebbc78da67de41c46558f939e1ad330d01b56d9c9b",
2723
+ "orbitquant_codebooks.safetensors": "d30c534198062aa21847808e080a1043358025d436a8654ea2c66f20735dce68",
2724
+ "orbitquant_rotations.safetensors": "e1724c9394bf83ea05da6b457b8072d2bcb3d5a533dafce916da8deb517d3825",
2725
+ "quantization_config.json": "3134bdd201c7006489051dfa6ae86ca250cf35d3921e6785fdd0675c9963db59",
2726
+ "model_index.json": "7af2cec0bf4745140b0572ea3dfadd1f0d97dad6db16c5e0165cda60f3798e00",
2727
+ "prompts.json": "65ec8afff39a6cf80feee2eac2f06fa66e51cd471a8382f4355a04fd74eda0af",
2728
+ "benchmark/summary.json": "1c99fc94d6fdd328753ec0cac0d2d93da0a477a39a458cd413903b11c79191ba",
2729
+ "benchmark/original.metrics.jsonl": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
2730
+ "benchmark/orbitquant.metrics.jsonl": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
2731
+ "benchmark/original.metrics.csv": "90d54eb9870d76c7e992e1ca94341cc7a58c931a079162ef1774b7f2b571ac73",
2732
+ "benchmark/orbitquant.metrics.csv": "90d54eb9870d76c7e992e1ca94341cc7a58c931a079162ef1774b7f2b571ac73",
2733
+ "assets/.gitkeep": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
2734
+ }
2735
+ }
orbitquant_rotations.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e1724c9394bf83ea05da6b457b8072d2bcb3d5a533dafce916da8deb517d3825
3
+ size 240176
prompts.json ADDED
@@ -0,0 +1,57 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "prompt_pack": "image_visual_v1",
3
+ "media_type": "image",
4
+ "target_policy": "z_image",
5
+ "prompts": [
6
+ {
7
+ "id": "simple-object",
8
+ "category": "simple_object",
9
+ "prompt": "A red ceramic mug on a wooden desk, soft daylight, shallow depth of field"
10
+ },
11
+ {
12
+ "id": "two-object-composition",
13
+ "category": "two_object_composition",
14
+ "prompt": "Two robots playing chess in a quiet library, one silver and one matte black"
15
+ },
16
+ {
17
+ "id": "counting",
18
+ "category": "counting",
19
+ "prompt": "Five glass marbles arranged in a straight line on white paper"
20
+ },
21
+ {
22
+ "id": "color-binding",
23
+ "category": "color_binding",
24
+ "prompt": "A blue cube beside a yellow sphere and a green pyramid on a gray table"
25
+ },
26
+ {
27
+ "id": "spatial-relationship",
28
+ "category": "spatial_relationship",
29
+ "prompt": "A blue cube to the left of a yellow sphere, both centered in the frame"
30
+ },
31
+ {
32
+ "id": "long-prompt",
33
+ "category": "long_prompt",
34
+ "prompt": "A precise editorial photo of a compact robotics workbench with labeled tools, a small oscilloscope, coiled cables, a half-assembled drone, and a warm desk lamp; the scene should remain readable without clutter"
35
+ },
36
+ {
37
+ "id": "english-text-rendering",
38
+ "category": "text_rendering",
39
+ "prompt": "A clean street sign with the exact text \"ORBIT QUANT\""
40
+ },
41
+ {
42
+ "id": "cyrillic-text-rendering",
43
+ "category": "text_rendering",
44
+ "prompt": "A handwritten note in Cyrillic that says \"\u041a\u0412\u0410\u041d\u0422\u041e\u0412\u0410\u041d\u0418\u0415\""
45
+ },
46
+ {
47
+ "id": "style-heavy",
48
+ "category": "style",
49
+ "prompt": "A crowded train platform at sunset, documentary photo style, high dynamic range, natural grain, realistic faces, detailed architecture"
50
+ },
51
+ {
52
+ "id": "occlusion-reflection",
53
+ "category": "occlusion_reflection",
54
+ "prompt": "A reflective chrome teapot partly hidden behind orange flowers, with the room visible as a coherent reflection on the metal"
55
+ }
56
+ ]
57
+ }
quantization_config.json ADDED
@@ -0,0 +1,22 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "quant_method": "orbitquant",
3
+ "weight_bits": 4,
4
+ "activation_bits": 4,
5
+ "rotation": "rpbh",
6
+ "rotation_seed": 0,
7
+ "block_size": "paper",
8
+ "codebook": "lloyd_max",
9
+ "codebook_dtype": "float32",
10
+ "row_norm_dtype": "bfloat16",
11
+ "activation_norm_dtype": "float32",
12
+ "activation_eps": 1e-12,
13
+ "weight_pack_dtype": "uint8",
14
+ "target_policy": "z_image",
15
+ "adaln_policy": "int4_rtn",
16
+ "adaln_group_size": 64,
17
+ "modules_to_not_convert": [],
18
+ "modules_dtype_dict": {},
19
+ "artifact_format_version": 1,
20
+ "runtime_mode": "dequant_bf16",
21
+ "activation_kernel_backend": "auto"
22
+ }