Diffusers
Safetensors
OrbitQuantComponentArtifact
orbitquant
quantized
diffusion-transformer
8-bit precision
Instructions to use WaveCut/Z-Image-Turbo-OrbitQuant-W2A4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use WaveCut/Z-Image-Turbo-OrbitQuant-W2A4 with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("WaveCut/Z-Image-Turbo-OrbitQuant-W2A4", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
Replace artifact with paper-correct codebook v2 and native proof
Browse files- README.md +2 -2
- SHA256SUMS +8 -8
- assets/image_generation_comparison_matrix.webp +2 -2
- benchmark/summary.json +9 -24
- model.safetensors +1 -1
- model_index.json +2 -1
- orbitquant_codebooks.safetensors +1 -1
- orbitquant_manifest.json +8 -8
- quantization_config.json +5 -4
README.md
CHANGED
|
@@ -90,7 +90,7 @@ The compact benchmark summary records native BF16-vs-OrbitQuant evidence for the
|
|
| 90 |
- Method: `orbitquant`
|
| 91 |
- Bits: `W2A4`
|
| 92 |
- Runtime mode: `auto_fused`
|
| 93 |
-
- Activation kernel backend: `
|
| 94 |
- Activation normalization epsilon: `1e-10`
|
| 95 |
- Quantization device: `cuda`
|
| 96 |
- Weight quantization backend: `triton_cuda`
|
|
@@ -103,7 +103,7 @@ The compact benchmark summary records native BF16-vs-OrbitQuant evidence for the
|
|
| 103 |
- Block size: `paper`
|
| 104 |
- Block size policy: `largest_power_of_two_dividing_dim`
|
| 105 |
- Codebook: `lloyd_max`
|
| 106 |
-
- Codebook version: `
|
| 107 |
- Quantized transformer modules: `238`
|
| 108 |
- AdaLN INT4 modules: `32`
|
| 109 |
- Skipped modules: `6`
|
|
|
|
| 90 |
- Method: `orbitquant`
|
| 91 |
- Bits: `W2A4`
|
| 92 |
- Runtime mode: `auto_fused`
|
| 93 |
+
- Activation kernel backend: `auto`
|
| 94 |
- Activation normalization epsilon: `1e-10`
|
| 95 |
- Quantization device: `cuda`
|
| 96 |
- Weight quantization backend: `triton_cuda`
|
|
|
|
| 103 |
- Block size: `paper`
|
| 104 |
- Block size policy: `largest_power_of_two_dividing_dim`
|
| 105 |
- Codebook: `lloyd_max`
|
| 106 |
+
- Codebook version: `2`
|
| 107 |
- Quantized transformer modules: `238`
|
| 108 |
- AdaLN INT4 modules: `32`
|
| 109 |
- Skipped modules: `6`
|
SHA256SUMS
CHANGED
|
@@ -1,10 +1,10 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
e1724c9394bf83ea05da6b457b8072d2bcb3d5a533dafce916da8deb517d3825 orbitquant_rotations.safetensors
|
| 9 |
65ec8afff39a6cf80feee2eac2f06fa66e51cd471a8382f4355a04fd74eda0af prompts.json
|
| 10 |
-
|
|
|
|
| 1 |
+
1a52efbd377f8254884be98aa7ab1ef92d4fe599dac50fcb412946c4a365add9 README.md
|
| 2 |
+
812f3828b3bcd19c9879f903c6b768ad2a7d4293016518fda916ab6dba663aa3 assets/image_generation_comparison_matrix.webp
|
| 3 |
+
618770414076f33ca4c7aebe0c173c7e419dcf4c76b4082fb1a8af88bfc54fda benchmark/summary.json
|
| 4 |
+
aa8e04810221025dc262f66632975bdbf8466b341ad19db61ba4ad4f174f402f model.safetensors
|
| 5 |
+
33663f26f6c41269f71b4300ba1e92d57e426e52128854b079c25fec0e9080c3 model_index.json
|
| 6 |
+
9ee09b3fcc9615269db665df2a76b9a94099aea67fb5c9ce1523f5c53f0e7723 orbitquant_codebooks.safetensors
|
| 7 |
+
643f1d8f45a489dde2521adad560803bc4949ea9f0f9e732deb8e3999bc575f3 orbitquant_manifest.json
|
| 8 |
e1724c9394bf83ea05da6b457b8072d2bcb3d5a533dafce916da8deb517d3825 orbitquant_rotations.safetensors
|
| 9 |
65ec8afff39a6cf80feee2eac2f06fa66e51cd471a8382f4355a04fd74eda0af prompts.json
|
| 10 |
+
6cac0dabb20f4fa9cceda63aa86ba4ca5df66902aeec52b98ab133f0c381bc71 quantization_config.json
|
assets/image_generation_comparison_matrix.webp
CHANGED
|
Git LFS Details
|
|
Git LFS Details
|
benchmark/summary.json
CHANGED
|
@@ -1,42 +1,27 @@
|
|
| 1 |
{
|
| 2 |
-
"status": "
|
| 3 |
"source_model_id": "Tongyi-MAI/Z-Image-Turbo",
|
| 4 |
"source_revision": "f332072aa78be7aecdf3ee76d5c247082da564a6",
|
| 5 |
"weight_bits": 2,
|
| 6 |
"activation_bits": 4,
|
|
|
|
| 7 |
"target_policy": "z_image",
|
| 8 |
"runtime_mode": "auto_fused",
|
| 9 |
-
"activation_kernel_backend": "
|
| 10 |
"activation_eps": 1e-10,
|
| 11 |
"quantization_device": "cuda",
|
| 12 |
"weight_quantization_backend": "triton_cuda",
|
| 13 |
"quantization_staging_mode": "streaming",
|
| 14 |
"synchronize_per_module": false,
|
| 15 |
-
"quantization_elapsed_seconds":
|
| 16 |
-
"orbitquant_seconds":
|
| 17 |
-
"adaln_seconds": 0.
|
| 18 |
-
"device_transfer_seconds": 1.
|
| 19 |
"module_device_transfer_count": 270,
|
| 20 |
-
"metrics": {
|
| 21 |
-
"orbitquant": {
|
| 22 |
-
"records": 1,
|
| 23 |
-
"latest_metrics": {
|
| 24 |
-
"generated_samples": 1,
|
| 25 |
-
"wall_time_seconds": 35.89397249696776,
|
| 26 |
-
"peak_vram_bytes": 13017512448
|
| 27 |
-
}
|
| 28 |
-
},
|
| 29 |
-
"original": {
|
| 30 |
-
"records": 1,
|
| 31 |
-
"latest_metrics": {
|
| 32 |
-
"generated_samples": 1,
|
| 33 |
-
"wall_time_seconds": 5.200292671099305,
|
| 34 |
-
"peak_vram_bytes": 23267081728
|
| 35 |
-
}
|
| 36 |
-
}
|
| 37 |
-
},
|
| 38 |
"native_smoke": {
|
| 39 |
"proof_format": "orbitquant-native-smoke-v1",
|
|
|
|
| 40 |
"comparison_asset_path": "assets/image_generation_comparison_matrix.webp",
|
| 41 |
"paired_prompt_seed_count": 1,
|
| 42 |
"paired_prompt_seed_keys": [
|
|
|
|
| 1 |
{
|
| 2 |
+
"status": "not_run",
|
| 3 |
"source_model_id": "Tongyi-MAI/Z-Image-Turbo",
|
| 4 |
"source_revision": "f332072aa78be7aecdf3ee76d5c247082da564a6",
|
| 5 |
"weight_bits": 2,
|
| 6 |
"activation_bits": 4,
|
| 7 |
+
"codebook_version": 2,
|
| 8 |
"target_policy": "z_image",
|
| 9 |
"runtime_mode": "auto_fused",
|
| 10 |
+
"activation_kernel_backend": "auto",
|
| 11 |
"activation_eps": 1e-10,
|
| 12 |
"quantization_device": "cuda",
|
| 13 |
"weight_quantization_backend": "triton_cuda",
|
| 14 |
"quantization_staging_mode": "streaming",
|
| 15 |
"synchronize_per_module": false,
|
| 16 |
+
"quantization_elapsed_seconds": 4.982273037079722,
|
| 17 |
+
"orbitquant_seconds": 3.547152883838862,
|
| 18 |
+
"adaln_seconds": 0.024637432768940926,
|
| 19 |
+
"device_transfer_seconds": 1.3974662092514336,
|
| 20 |
"module_device_transfer_count": 270,
|
| 21 |
+
"metrics": {},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
"native_smoke": {
|
| 23 |
"proof_format": "orbitquant-native-smoke-v1",
|
| 24 |
+
"proof_source": "local_compare_native_bundle",
|
| 25 |
"comparison_asset_path": "assets/image_generation_comparison_matrix.webp",
|
| 26 |
"paired_prompt_seed_count": 1,
|
| 27 |
"paired_prompt_seed_keys": [
|
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 1599432256
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:aa8e04810221025dc262f66632975bdbf8466b341ad19db61ba4ad4f174f402f
|
| 3 |
size 1599432256
|
model_index.json
CHANGED
|
@@ -13,9 +13,10 @@
|
|
| 13 |
"rotations": "orbitquant_rotations.safetensors",
|
| 14 |
"weight_bits": 2,
|
| 15 |
"activation_bits": 4,
|
|
|
|
| 16 |
"target_policy": "z_image",
|
| 17 |
"runtime_mode": "auto_fused",
|
| 18 |
-
"activation_kernel_backend": "
|
| 19 |
"activation_eps": 1e-10,
|
| 20 |
"quantization_device": "cuda",
|
| 21 |
"weight_quantization_backend": "triton_cuda",
|
|
|
|
| 13 |
"rotations": "orbitquant_rotations.safetensors",
|
| 14 |
"weight_bits": 2,
|
| 15 |
"activation_bits": 4,
|
| 16 |
+
"codebook_version": 2,
|
| 17 |
"target_policy": "z_image",
|
| 18 |
"runtime_mode": "auto_fused",
|
| 19 |
+
"activation_kernel_backend": "auto",
|
| 20 |
"activation_eps": 1e-10,
|
| 21 |
"quantization_device": "cuda",
|
| 22 |
"weight_quantization_backend": "triton_cuda",
|
orbitquant_codebooks.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 952
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9ee09b3fcc9615269db665df2a76b9a94099aea67fb5c9ce1523f5c53f0e7723
|
| 3 |
size 952
|
orbitquant_manifest.json
CHANGED
|
@@ -12,10 +12,10 @@
|
|
| 12 |
"block_size": "paper",
|
| 13 |
"block_size_policy": "largest_power_of_two_dividing_dim",
|
| 14 |
"codebook": "lloyd_max",
|
| 15 |
-
"codebook_version":
|
| 16 |
"row_norm_dtype": "bfloat16",
|
| 17 |
"runtime_mode": "auto_fused",
|
| 18 |
-
"activation_kernel_backend": "
|
| 19 |
"activation_eps": 1e-10,
|
| 20 |
"adaln_group_size": 64,
|
| 21 |
"quantization_device": "cuda",
|
|
@@ -2721,13 +2721,13 @@
|
|
| 2721 |
]
|
| 2722 |
},
|
| 2723 |
"checksums": {
|
| 2724 |
-
"assets/image_generation_comparison_matrix.webp": "
|
| 2725 |
-
"benchmark/summary.json": "
|
| 2726 |
-
"model.safetensors": "
|
| 2727 |
-
"model_index.json": "
|
| 2728 |
-
"orbitquant_codebooks.safetensors": "
|
| 2729 |
"orbitquant_rotations.safetensors": "e1724c9394bf83ea05da6b457b8072d2bcb3d5a533dafce916da8deb517d3825",
|
| 2730 |
"prompts.json": "65ec8afff39a6cf80feee2eac2f06fa66e51cd471a8382f4355a04fd74eda0af",
|
| 2731 |
-
"quantization_config.json": "
|
| 2732 |
}
|
| 2733 |
}
|
|
|
|
| 12 |
"block_size": "paper",
|
| 13 |
"block_size_policy": "largest_power_of_two_dividing_dim",
|
| 14 |
"codebook": "lloyd_max",
|
| 15 |
+
"codebook_version": 2,
|
| 16 |
"row_norm_dtype": "bfloat16",
|
| 17 |
"runtime_mode": "auto_fused",
|
| 18 |
+
"activation_kernel_backend": "auto",
|
| 19 |
"activation_eps": 1e-10,
|
| 20 |
"adaln_group_size": 64,
|
| 21 |
"quantization_device": "cuda",
|
|
|
|
| 2721 |
]
|
| 2722 |
},
|
| 2723 |
"checksums": {
|
| 2724 |
+
"assets/image_generation_comparison_matrix.webp": "812f3828b3bcd19c9879f903c6b768ad2a7d4293016518fda916ab6dba663aa3",
|
| 2725 |
+
"benchmark/summary.json": "618770414076f33ca4c7aebe0c173c7e419dcf4c76b4082fb1a8af88bfc54fda",
|
| 2726 |
+
"model.safetensors": "aa8e04810221025dc262f66632975bdbf8466b341ad19db61ba4ad4f174f402f",
|
| 2727 |
+
"model_index.json": "33663f26f6c41269f71b4300ba1e92d57e426e52128854b079c25fec0e9080c3",
|
| 2728 |
+
"orbitquant_codebooks.safetensors": "9ee09b3fcc9615269db665df2a76b9a94099aea67fb5c9ce1523f5c53f0e7723",
|
| 2729 |
"orbitquant_rotations.safetensors": "e1724c9394bf83ea05da6b457b8072d2bcb3d5a533dafce916da8deb517d3825",
|
| 2730 |
"prompts.json": "65ec8afff39a6cf80feee2eac2f06fa66e51cd471a8382f4355a04fd74eda0af",
|
| 2731 |
+
"quantization_config.json": "6cac0dabb20f4fa9cceda63aa86ba4ca5df66902aeec52b98ab133f0c381bc71"
|
| 2732 |
}
|
| 2733 |
}
|
quantization_config.json
CHANGED
|
@@ -6,6 +6,7 @@
|
|
| 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",
|
|
@@ -18,9 +19,9 @@
|
|
| 18 |
"modules_dtype_dict": {},
|
| 19 |
"artifact_format_version": 1,
|
| 20 |
"runtime_mode": "auto_fused",
|
| 21 |
-
"activation_kernel_backend": "
|
| 22 |
-
"packed_matmul_block_m":
|
| 23 |
"packed_matmul_block_n": 64,
|
| 24 |
-
"packed_matmul_block_k":
|
| 25 |
-
"packed_matmul_num_warps":
|
| 26 |
}
|
|
|
|
| 6 |
"rotation_seed": 0,
|
| 7 |
"block_size": "paper",
|
| 8 |
"codebook": "lloyd_max",
|
| 9 |
+
"codebook_version": 2,
|
| 10 |
"codebook_dtype": "float32",
|
| 11 |
"row_norm_dtype": "bfloat16",
|
| 12 |
"activation_norm_dtype": "float32",
|
|
|
|
| 19 |
"modules_dtype_dict": {},
|
| 20 |
"artifact_format_version": 1,
|
| 21 |
"runtime_mode": "auto_fused",
|
| 22 |
+
"activation_kernel_backend": "auto",
|
| 23 |
+
"packed_matmul_block_m": 64,
|
| 24 |
"packed_matmul_block_n": 64,
|
| 25 |
+
"packed_matmul_block_k": 128,
|
| 26 |
+
"packed_matmul_num_warps": 4
|
| 27 |
}
|