Add files using upload-large-folder tool

.gitattributes

@@ -33,3 +33,16 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+pe_tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+comparison/00_00-cyrillic-poster.webp filter=lfs diff=lfs merge=lfs -text
+comparison/01_01-long-text-bakery-ad.webp filter=lfs diff=lfs merge=lfs -text
+comparison/02_02-technical-diagram.webp filter=lfs diff=lfs merge=lfs -text
+comparison/03_03-four-panel-comic.webp filter=lfs diff=lfs merge=lfs -text
+comparison/04_04-public-domain-painter-fusion.webp filter=lfs diff=lfs merge=lfs -text
+comparison/06_06-glass-still-life.webp filter=lfs diff=lfs merge=lfs -text
+comparison/07_07-botanical-field-guide.webp filter=lfs diff=lfs merge=lfs -text
+comparison/08_08-restaurant-menu-board.webp filter=lfs diff=lfs merge=lfs -text
+comparison/09_09-isometric-city-map.webp filter=lfs diff=lfs merge=lfs -text
+comparison/original_vs_sdnq_uint4_static_matrix.webp filter=lfs diff=lfs merge=lfs -text
+comparison/original_vs_sdnq_qmm_vs_aifarm_current_matrix.webp filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,99 @@
+---
+license: apache-2.0
+base_model: baidu/ERNIE-Image-Turbo
+pipeline_tag: text-to-image
+library_name: diffusers
+tags:
+- text-to-image
+- diffusers
+- safetensors
+- ernie-image
+- sdnq
+- quantized
+- uint4
+- static
+---
+
+# ERNIE-Image-Turbo SDNQ UINT4 Static
+
+This is a 4-bit SDNQ static quantization of [baidu/ERNIE-Image-Turbo](https://huggingface.co/baidu/ERNIE-Image-Turbo).
+It keeps the original Diffusers `ErnieImagePipeline` layout and was produced with `sdnq` using `weights_dtype=uint4`,
+static weight quantization, `dequantize_fp32=false`, and no dynamic threshold.
+
+The original model card describes ERNIE-Image-Turbo as an 8B distilled text-to-image model for fast 8-step generation,
+with strong text rendering, structured image generation, and complex prompt following. This quantized release is intended
+to reduce local storage and VRAM needs while preserving those practical behaviors.
+
+## Recipe
+
+- Base model: `baidu/ERNIE-Image-Turbo`
+- Quantizer: `sdnq` {'weights_dtype': 'uint4', 'group_size': 0, 'use_svd': False, 'svd_rank': 32, 'svd_steps': 8, 'dynamic_loss_threshold': None, 'use_quantized_matmul': False, 'dequantize_fp32': False}
+- Validation loader: SDNQ quantized matmul enabled for `pe`, `text_encoder`, and `transformer`
+- Output artifact size: `10.05 GB`
+- GPU used for quantization and validation: NVIDIA RTX A6000
+- Validation prompts: 10 fixed prompt/seed pairs covering Cyrillic text, long English text, diagrams, UI, comics, reflections, maps, and painterly styles.
+
+## Measured Results
+
+| Model | Load s | Load peak VRAM MiB | Cold inference s | Cold peak VRAM MiB | Hot mean s/img | Hot peak VRAM MiB |
+|---|---:|---:|---:|---:|---:|---:|
+| Original BF16 | 84.64 | 29614 | 19.96 | 34848 | 21.26 | 35126 |
+| SDNQ UINT4 static + quantized matmul | 62.36 | 9776 | 37.53 | 12870 | 39.90 | 12974 |
+
+## Visual Comparison
+
+[![Original vs SDNQ UINT4 static comparison matrix](comparison/original_vs_sdnq_uint4_static_matrix.webp)](comparison/original_vs_sdnq_uint4_static_matrix.webp)
+
+Individual prompt pairs are stored in `comparison/`, and full metrics are stored in `metrics/`.
+
+## Usage
+
+Install current Diffusers and SDNQ, then load the pipeline normally:
+
+```python
+import torch
+import sdnq  # registers the SDNQ quantizer
+from diffusers import ErnieImagePipeline
+from sdnq.loader import apply_sdnq_options_to_model
+
+pipe = ErnieImagePipeline.from_pretrained(
+    "WaveCut/ERNIE-Image-Turbo-SDNQ-uint4-static",
+    torch_dtype=torch.bfloat16,
+).to("cuda")
+
+for name in ("pe", "text_encoder", "transformer"):
+    component = getattr(pipe, name, None)
+    if component is not None and hasattr(component, "quantization_config"):
+        setattr(pipe, name, apply_sdnq_options_to_model(component, use_quantized_matmul=True))
+
+image = pipe(
+    prompt="A clean poster with exact readable text: HELLO ERNIE",
+    width=1024,
+    height=1024,
+    num_inference_steps=8,
+    guidance_scale=1.0,
+    use_pe=True,
+).images[0]
+image.save("output.png")
+```
+
+## Prompt Set
+
+| # | Prompt ID | Size | Seed | Focus |
+|---:|---|---:|---:|---|
+| 00 | `00-cyrillic-poster` | 1024x1024 | 41001 | Cyrillic event poster |
+| 01 | `01-long-text-bakery-ad` | 896x1200 | 41002 | Long text product ad |
+| 02 | `02-technical-diagram` | 1200x896 | 41003 | Technical diagram |
+| 03 | `03-four-panel-comic` | 1024x1024 | 41004 | Four-panel comic |
+| 04 | `04-public-domain-painter-fusion` | 1024x1024 | 41005 | Painterly style fusion |
+| 05 | `05-dashboard-ui` | 1376x768 | 41006 | Dense UI dashboard |
+| 06 | `06-glass-still-life` | 1024x1024 | 41007 | Glass and reflections |
+| 07 | `07-botanical-field-guide` | 896x1200 | 41008 | Field guide plate |
+| 08 | `08-restaurant-menu-board` | 1024x1024 | 41009 | Menu board text |
+| 09 | `09-isometric-city-map` | 1200x896 | 41010 | Isometric map |
+
+## Notes
+
+- The comparison uses the same prompts, dimensions, seeds, 8 inference steps, guidance scale 1.0, `use_pe=True`, and SDNQ quantized matmul for the quantized model.
+- The model was visually checked against the original comparison matrix for layout preservation, prompt adherence, text rendering, and obvious quantization artifacts.
+- This is an independent quantized artifact; see the original Baidu model card for upstream model details, benchmarks, and license terms.

comparison/manifest.json

@@ -0,0 +1,49 @@
+{
+  "matrix": "comparison/original_vs_sdnq_uint4_static_matrix.webp",
+  "pairs": [
+    {
+      "prompt_id": "00-cyrillic-poster",
+      "pair": "comparison/00_00-cyrillic-poster.webp"
+    },
+    {
+      "prompt_id": "01-long-text-bakery-ad",
+      "pair": "comparison/01_01-long-text-bakery-ad.webp"
+    },
+    {
+      "prompt_id": "02-technical-diagram",
+      "pair": "comparison/02_02-technical-diagram.webp"
+    },
+    {
+      "prompt_id": "03-four-panel-comic",
+      "pair": "comparison/03_03-four-panel-comic.webp"
+    },
+    {
+      "prompt_id": "04-public-domain-painter-fusion",
+      "pair": "comparison/04_04-public-domain-painter-fusion.webp"
+    },
+    {
+      "prompt_id": "05-dashboard-ui",
+      "pair": "comparison/05_05-dashboard-ui.webp"
+    },
+    {
+      "prompt_id": "06-glass-still-life",
+      "pair": "comparison/06_06-glass-still-life.webp"
+    },
+    {
+      "prompt_id": "07-botanical-field-guide",
+      "pair": "comparison/07_07-botanical-field-guide.webp"
+    },
+    {
+      "prompt_id": "08-restaurant-menu-board",
+      "pair": "comparison/08_08-restaurant-menu-board.webp"
+    },
+    {
+      "prompt_id": "09-isometric-city-map",
+      "pair": "comparison/09_09-isometric-city-map.webp"
+    }
+  ],
+  "original_label": "original",
+  "quant_label": "sdnq_uint4_static_qmm",
+  "quant_title": "SDNQ UINT4 static + qmm",
+  "format": "WebP quality 97"
+}

comparison/three_way_manifest.json

@@ -0,0 +1,10 @@
+{
+  "matrix": "comparison/original_vs_sdnq_qmm_vs_aifarm_current_matrix.webp",
+  "columns": [
+    "Original BF16",
+    "ERNIE SDNQ UINT4 qmm",
+    "AI Farm FLUX.2 klein"
+  ],
+  "format": "WebP quality 97",
+  "uploaded_to_hf_but_not_linked_from_readme": true
+}

metrics/aifarm_current_metrics.json

@@ -0,0 +1,151 @@
+{
+  "label": "aifarm_current",
+  "model": "WaveCut/FLUX.2-klein-9B-SDNQ-SNOfS-lora05-heretic-te-uint4-static",
+  "settings": {
+    "num_inference_steps": 8,
+    "guidance_scale": 1.0,
+    "seed": "passed to draw-api"
+  },
+  "generations": [
+    {
+      "name": "generate_00-cyrillic-poster",
+      "seconds": 22.099228713996126,
+      "elapsed_ms_reported": 21650,
+      "job_id": "aifarm-ernie-compare-00-cyrillic-poster-1779059445702",
+      "request_id": "ernie-compare-00-cyrillic-poster",
+      "model_id": "WaveCut/FLUX.2-klein-9B-SDNQ-SNOfS-lora05-heretic-te-uint4-static",
+      "prompt_id": "00-cyrillic-poster",
+      "title": "Cyrillic event poster",
+      "seed": 41001,
+      "width": 1024,
+      "height": 1024,
+      "image": "images/00_00-cyrillic-poster.jpg"
+    },
+    {
+      "name": "generate_01-long-text-bakery-ad",
+      "seconds": 45.140852112992434,
+      "elapsed_ms_reported": 30304,
+      "job_id": "aifarm-ernie-compare-01-long-text-bakery-ad-1779059467802",
+      "request_id": "ernie-compare-01-long-text-bakery-ad",
+      "model_id": "WaveCut/FLUX.2-klein-9B-SDNQ-SNOfS-lora05-heretic-te-uint4-static",
+      "prompt_id": "01-long-text-bakery-ad",
+      "title": "Long text product ad",
+      "seed": 41002,
+      "width": 896,
+      "height": 1200,
+      "image": "images/01_01-long-text-bakery-ad.jpg"
+    },
+    {
+      "name": "generate_02-technical-diagram",
+      "seconds": 28.08245640298992,
+      "elapsed_ms_reported": 27132,
+      "job_id": "aifarm-ernie-compare-02-technical-diagram-1779059512943",
+      "request_id": "ernie-compare-02-technical-diagram",
+      "model_id": "WaveCut/FLUX.2-klein-9B-SDNQ-SNOfS-lora05-heretic-te-uint4-static",
+      "prompt_id": "02-technical-diagram",
+      "title": "Technical diagram",
+      "seed": 41003,
+      "width": 1200,
+      "height": 896,
+      "image": "images/02_02-technical-diagram.jpg"
+    },
+    {
+      "name": "generate_03-four-panel-comic",
+      "seconds": 36.10489075900114,
+      "elapsed_ms_reported": 21818,
+      "job_id": "aifarm-ernie-compare-03-four-panel-comic-1779059541026",
+      "request_id": "ernie-compare-03-four-panel-comic",
+      "model_id": "WaveCut/FLUX.2-klein-9B-SDNQ-SNOfS-lora05-heretic-te-uint4-static",
+      "prompt_id": "03-four-panel-comic",
+      "title": "Four-panel comic",
+      "seed": 41004,
+      "width": 1024,
+      "height": 1024,
+      "image": "images/03_03-four-panel-comic.jpg"
+    },
+    {
+      "name": "generate_04-public-domain-painter-fusion",
+      "seconds": 22.075391891994514,
+      "elapsed_ms_reported": 21789,
+      "job_id": "aifarm-ernie-compare-04-public-domain-painter-fusion-1779059577131",
+      "request_id": "ernie-compare-04-public-domain-painter-fusion",
+      "model_id": "WaveCut/FLUX.2-klein-9B-SDNQ-SNOfS-lora05-heretic-te-uint4-static",
+      "prompt_id": "04-public-domain-painter-fusion",
+      "title": "Painterly style fusion",
+      "seed": 41005,
+      "width": 1024,
+      "height": 1024,
+      "image": "images/04_04-public-domain-painter-fusion.jpg"
+    },
+    {
+      "name": "generate_05-dashboard-ui",
+      "seconds": 46.1179654699954,
+      "elapsed_ms_reported": 30210,
+      "job_id": "aifarm-ernie-compare-05-dashboard-ui-1779059599207",
+      "request_id": "ernie-compare-05-dashboard-ui",
+      "model_id": "WaveCut/FLUX.2-klein-9B-SDNQ-SNOfS-lora05-heretic-te-uint4-static",
+      "prompt_id": "05-dashboard-ui",
+      "title": "Dense UI dashboard",
+      "seed": 41006,
+      "width": 1376,
+      "height": 768,
+      "image": "images/05_05-dashboard-ui.jpg"
+    },
+    {
+      "name": "generate_06-glass-still-life",
+      "seconds": 36.09438178999699,
+      "elapsed_ms_reported": 21733,
+      "job_id": "aifarm-ernie-compare-06-glass-still-life-1779059645325",
+      "request_id": "ernie-compare-06-glass-still-life",
+      "model_id": "WaveCut/FLUX.2-klein-9B-SDNQ-SNOfS-lora05-heretic-te-uint4-static",
+      "prompt_id": "06-glass-still-life",
+      "title": "Glass and reflections",
+      "seed": 41007,
+      "width": 1024,
+      "height": 1024,
+      "image": "images/06_06-glass-still-life.jpg"
+    },
+    {
+      "name": "generate_07-botanical-field-guide",
+      "seconds": 39.10773252599756,
+      "elapsed_ms_reported": 23354,
+      "job_id": "aifarm-ernie-compare-07-botanical-field-guide-1779059681420",
+      "request_id": "ernie-compare-07-botanical-field-guide",
+      "model_id": "WaveCut/FLUX.2-klein-9B-SDNQ-SNOfS-lora05-heretic-te-uint4-static",
+      "prompt_id": "07-botanical-field-guide",
+      "title": "Field guide plate",
+      "seed": 41008,
+      "width": 896,
+      "height": 1200,
+      "image": "images/07_07-botanical-field-guide.jpg"
+    },
+    {
+      "name": "generate_08-restaurant-menu-board",
+      "seconds": 36.10639868999715,
+      "elapsed_ms_reported": 21702,
+      "job_id": "aifarm-ernie-compare-08-restaurant-menu-board-1779059720528",
+      "request_id": "ernie-compare-08-restaurant-menu-board",
+      "model_id": "WaveCut/FLUX.2-klein-9B-SDNQ-SNOfS-lora05-heretic-te-uint4-static",
+      "prompt_id": "08-restaurant-menu-board",
+      "title": "Menu board text",
+      "seed": 41009,
+      "width": 1024,
+      "height": 1024,
+      "image": "images/08_08-restaurant-menu-board.jpg"
+    },
+    {
+      "name": "generate_09-isometric-city-map",
+      "seconds": 24.073563486992498,
+      "elapsed_ms_reported": 23341,
+      "job_id": "aifarm-ernie-compare-09-isometric-city-map-1779059756635",
+      "request_id": "ernie-compare-09-isometric-city-map",
+      "model_id": "WaveCut/FLUX.2-klein-9B-SDNQ-SNOfS-lora05-heretic-te-uint4-static",
+      "prompt_id": "09-isometric-city-map",
+      "title": "Isometric map",
+      "seed": 41010,
+      "width": 1200,
+      "height": 896,
+      "image": "images/09_09-isometric-city-map.jpg"
+    }
+  ]
+}

metrics/original_metrics.json

@@ -0,0 +1,170 @@
+{
+  "label": "original",
+  "model": "baidu/ERNIE-Image-Turbo",
+  "settings": {
+    "num_inference_steps": 8,
+    "guidance_scale": 1.0,
+    "use_pe": true
+  },
+  "load": {
+    "name": "load_pipeline",
+    "seconds": 84.6412352710031,
+    "gpu_start_mib": 4,
+    "gpu_end_mib": 29698,
+    "gpu_peak_mib": 29614,
+    "torch_peak_allocated_mib": 29379,
+    "torch_peak_reserved_mib": 29428
+  },
+  "generations": [
+    {
+      "name": "generate_00-cyrillic-poster",
+      "seconds": 19.96482683205977,
+      "gpu_start_mib": 29698,
+      "gpu_end_mib": 34848,
+      "gpu_peak_mib": 34848,
+      "torch_peak_allocated_mib": 31830,
+      "torch_peak_reserved_mib": 34516,
+      "prompt_id": "00-cyrillic-poster",
+      "title": "Cyrillic event poster",
+      "seed": 41001,
+      "width": 1024,
+      "height": 1024,
+      "image": "original/images/00_00-cyrillic-poster.png"
+    },
+    {
+      "name": "generate_01-long-text-bakery-ad",
+      "seconds": 20.59528775140643,
+      "gpu_start_mib": 29760,
+      "gpu_end_mib": 35018,
+      "gpu_peak_mib": 35018,
+      "torch_peak_allocated_mib": 31897,
+      "torch_peak_reserved_mib": 34686,
+      "prompt_id": "01-long-text-bakery-ad",
+      "title": "Long text product ad",
+      "seed": 41002,
+      "width": 896,
+      "height": 1200,
+      "image": "original/images/01_01-long-text-bakery-ad.png"
+    },
+    {
+      "name": "generate_02-technical-diagram",
+      "seconds": 20.183351018000394,
+      "gpu_start_mib": 29760,
+      "gpu_end_mib": 35012,
+      "gpu_peak_mib": 35012,
+      "torch_peak_allocated_mib": 31897,
+      "torch_peak_reserved_mib": 34680,
+      "prompt_id": "02-technical-diagram",
+      "title": "Technical diagram",
+      "seed": 41003,
+      "width": 1200,
+      "height": 896,
+      "image": "original/images/02_02-technical-diagram.png"
+    },
+    {
+      "name": "generate_03-four-panel-comic",
+      "seconds": 24.42651346232742,
+      "gpu_start_mib": 29760,
+      "gpu_end_mib": 35004,
+      "gpu_peak_mib": 35004,
+      "torch_peak_allocated_mib": 31833,
+      "torch_peak_reserved_mib": 34672,
+      "prompt_id": "03-four-panel-comic",
+      "title": "Four-panel comic",
+      "seed": 41004,
+      "width": 1024,
+      "height": 1024,
+      "image": "original/images/03_03-four-panel-comic.png"
+    },
+    {
+      "name": "generate_04-public-domain-painter-fusion",
+      "seconds": 21.53745141904801,
+      "gpu_start_mib": 29760,
+      "gpu_end_mib": 34948,
+      "gpu_peak_mib": 34948,
+      "torch_peak_allocated_mib": 31831,
+      "torch_peak_reserved_mib": 34616,
+      "prompt_id": "04-public-domain-painter-fusion",
+      "title": "Painterly style fusion",
+      "seed": 41005,
+      "width": 1024,
+      "height": 1024,
+      "image": "original/images/04_04-public-domain-painter-fusion.png"
+    },
+    {
+      "name": "generate_05-dashboard-ui",
+      "seconds": 22.292551767081022,
+      "gpu_start_mib": 29760,
+      "gpu_end_mib": 35000,
+      "gpu_peak_mib": 35000,
+      "torch_peak_allocated_mib": 31854,
+      "torch_peak_reserved_mib": 34668,
+      "prompt_id": "05-dashboard-ui",
+      "title": "Dense UI dashboard",
+      "seed": 41006,
+      "width": 1376,
+      "height": 768,
+      "image": "original/images/05_05-dashboard-ui.png"
+    },
+    {
+      "name": "generate_06-glass-still-life",
+      "seconds": 19.866105827037245,
+      "gpu_start_mib": 29760,
+      "gpu_end_mib": 34874,
+      "gpu_peak_mib": 34874,
+      "torch_peak_allocated_mib": 31831,
+      "torch_peak_reserved_mib": 34542,
+      "prompt_id": "06-glass-still-life",
+      "title": "Glass and reflections",
+      "seed": 41007,
+      "width": 1024,
+      "height": 1024,
+      "image": "original/images/06_06-glass-still-life.png"
+    },
+    {
+      "name": "generate_07-botanical-field-guide",
+      "seconds": 23.37823957297951,
+      "gpu_start_mib": 29760,
+      "gpu_end_mib": 35126,
+      "gpu_peak_mib": 35126,
+      "torch_peak_allocated_mib": 31899,
+      "torch_peak_reserved_mib": 34794,
+      "prompt_id": "07-botanical-field-guide",
+      "title": "Field guide plate",
+      "seed": 41008,
+      "width": 896,
+      "height": 1200,
+      "image": "original/images/07_07-botanical-field-guide.png"
+    },
+    {
+      "name": "generate_08-restaurant-menu-board",
+      "seconds": 18.75183128984645,
+      "gpu_start_mib": 29760,
+      "gpu_end_mib": 34876,
+      "gpu_peak_mib": 34876,
+      "torch_peak_allocated_mib": 31830,
+      "torch_peak_reserved_mib": 34544,
+      "prompt_id": "08-restaurant-menu-board",
+      "title": "Menu board text",
+      "seed": 41009,
+      "width": 1024,
+      "height": 1024,
+      "image": "original/images/08_08-restaurant-menu-board.png"
+    },
+    {
+      "name": "generate_09-isometric-city-map",
+      "seconds": 20.34645133698359,
+      "gpu_start_mib": 29760,
+      "gpu_end_mib": 35010,
+      "gpu_peak_mib": 35010,
+      "torch_peak_allocated_mib": 31897,
+      "torch_peak_reserved_mib": 34678,
+      "prompt_id": "09-isometric-city-map",
+      "title": "Isometric map",
+      "seed": 41010,
+      "width": 1200,
+      "height": 896,
+      "image": "original/images/09_09-isometric-city-map.png"
+    }
+  ]
+}

metrics/prompts.json

@@ -0,0 +1,82 @@
+[
+  {
+    "id": "00-cyrillic-poster",
+    "seed": 41001,
+    "width": 1024,
+    "height": 1024,
+    "title": "Cyrillic event poster",
+    "prompt": "A clean modern cultural festival poster on matte paper, centered typographic hierarchy, large exact Russian headline text: \"НОЧЬ МУЗЕЕВ 2026\". Below it, smaller exact text: \"Лекции, музыка, световые инсталляции, вход свободный\". Include a map-like floor plan, ticket stubs, QR-code-like square, subtle cyan and red ink over warm white, crisp readable Cyrillic, Swiss poster grid, realistic print texture, no extra words."
+  },
+  {
+    "id": "01-long-text-bakery-ad",
+    "seed": 41002,
+    "width": 896,
+    "height": 1200,
+    "title": "Long text product ad",
+    "prompt": "A premium vertical magazine advertisement for a fictional bakery called \"Quantum Bakery\". Render a glossy croissant, a small espresso cup, and a paper bag. Include exact readable English text: \"48 LAYERS OF BUTTER, 12 HOURS OF PATIENCE, ONE QUIET MORNING\". Add a secondary line: \"OPEN 07:30-19:00, RUE DES ORCHIDS\". Elegant editorial photography, shallow depth of field, precise typography integrated into the scene, soft shadows, no misspellings."
+  },
+  {
+    "id": "02-technical-diagram",
+    "seed": 41003,
+    "width": 1200,
+    "height": 896,
+    "title": "Technical diagram",
+    "prompt": "A clear technical cutaway diagram of a lunar greenhouse module, white background, thin black vector lines, arrows, numbered callouts, and readable labels. Include exact labels: \"AIRLOCK\", \"WATER LOOP\", \"ROOT MIST\", \"LED CANOPY\", \"CO2 SENSOR\", and Russian label \"ЗОНА РОСТА\". Show pumps, transparent pipes, plant trays, pressure shell, and an inset timeline with 4 stages. High information density but clean layout, like a NASA engineering poster."
+  },
+  {
+    "id": "03-four-panel-comic",
+    "seed": 41004,
+    "width": 1024,
+    "height": 1024,
+    "title": "Four-panel comic",
+    "prompt": "A four-panel comic page with consistent characters and panel borders. Panel 1: a young engineer finds a tiny glowing compass on a rainy tram platform. Panel 2: the compass projects a map of the city above her hands. Panel 3: she repairs a broken street clock using the compass light. Panel 4: dawn breaks and the tram sign reads exact Russian text: \"СЛЕДУЮЩАЯ ОСТАНОВКА — ДОМ\". Warm cinematic color, expressive faces, clean speech bubbles, readable lettering, no random extra panels."
+  },
+  {
+    "id": "04-public-domain-painter-fusion",
+    "seed": 41005,
+    "width": 1024,
+    "height": 1024,
+    "title": "Painterly style fusion",
+    "prompt": "A large oil painting of a stormy botanical library, inspired by public-domain art language: luminous Turner-like atmosphere, Hokusai-style curling waves outside the windows, and Alphonse Mucha ornamental plant borders. A librarian in a dark green coat protects floating herbarium pages. Rich brushwork, layered glazes, coherent anatomy, dramatic light, museum-grade texture, no text."
+  },
+  {
+    "id": "05-dashboard-ui",
+    "seed": 41006,
+    "width": 1376,
+    "height": 768,
+    "title": "Dense UI dashboard",
+    "prompt": "A realistic screenshot of a professional operations dashboard for a city bike network, dark-on-light utilitarian UI, dense but organized. Include a map, 5 KPI cards, a sortable table, tiny trend charts, and readable labels: \"Fleet Health\", \"Open Repairs\", \"Battery Risk\", \"Station 12\", \"ETA 08:45\". No marketing hero layout, no decorative blobs, crisp interface typography, believable data."
+  },
+  {
+    "id": "06-glass-still-life",
+    "seed": 41007,
+    "width": 1024,
+    "height": 1024,
+    "title": "Glass and reflections",
+    "prompt": "A photorealistic studio still life with difficult transparent materials: two nested glass bowls, a chrome spoon, water droplets, sliced blood oranges, a folded blue silk scarf, and a small handwritten card that says exactly \"REFRACTION TEST\". Complex caustics, believable reflections, ray-traced look, high dynamic range, sharp focus."
+  },
+  {
+    "id": "07-botanical-field-guide",
+    "seed": 41008,
+    "width": 896,
+    "height": 1200,
+    "title": "Field guide plate",
+    "prompt": "A vintage botanical field guide plate showing six fictional alpine plants, each with roots, leaves, flowers, scale bars, and small Latin-style labels. Include exact title text: \"PLATE VII — ALPINE LUMEN FLORA\". Fine ink engraving, watercolor washes, aged paper, scientific layout, tiny annotations, no malformed plant parts."
+  },
+  {
+    "id": "08-restaurant-menu-board",
+    "seed": 41009,
+    "width": 1024,
+    "height": 1024,
+    "title": "Menu board text",
+    "prompt": "A cozy bistro interior with a chalkboard menu behind the counter. The chalkboard must contain readable Russian menu text: \"СУП ДНЯ — ТЫКВЕННЫЙ\", \"ПИРОГ С ЯБЛОКАМИ\", \"ЧАЙ С ОБЛЕПИХОЙ\", \"КОФЕ 180\". Warm tungsten light, realistic people in the background, chalk dust texture, legible Cyrillic, no extra gibberish."
+  },
+  {
+    "id": "09-isometric-city-map",
+    "seed": 41010,
+    "width": 1200,
+    "height": 896,
+    "title": "Isometric map",
+    "prompt": "An isometric illustrated city map for a compact waterfront district, with tiny buildings, bridges, tram tracks, parks, and a ferry terminal. Include readable place labels: \"North Pier\", \"Library Square\", \"Market Hall\", \"Старый мост\", and \"Line B\". Precise geometry, playful but clean cartographic style, many small details, consistent perspective, no floating text."
+  }
+]

metrics/quantization_metrics.json

@@ -0,0 +1,22 @@
+{
+  "name": "quantize_and_save",
+  "seconds": 71.19859568402171,
+  "gpu_start_mib": 270,
+  "gpu_end_mib": 312,
+  "gpu_peak_mib": 1276,
+  "torch_peak_allocated_mib": 870,
+  "torch_peak_reserved_mib": 964,
+  "base_model": "baidu/ERNIE-Image-Turbo",
+  "output": "artifacts/ERNIE-Image-Turbo-SDNQ-uint4-static",
+  "output_size_bytes": 10048077414,
+  "recipe": {
+    "weights_dtype": "uint4",
+    "group_size": 0,
+    "use_svd": false,
+    "svd_rank": 32,
+    "svd_steps": 8,
+    "dynamic_loss_threshold": null,
+    "use_quantized_matmul": false,
+    "dequantize_fp32": false
+  }
+}

metrics/sdnq_uint4_static_metrics.json

@@ -0,0 +1,171 @@
+{
+  "label": "sdnq_uint4_static_qmm",
+  "model": "artifacts/ERNIE-Image-Turbo-SDNQ-uint4-static",
+  "settings": {
+    "num_inference_steps": 8,
+    "guidance_scale": 1.0,
+    "use_pe": true,
+    "use_quantized_matmul": true
+  },
+  "load": {
+    "name": "load_pipeline",
+    "seconds": 62.3583417320624,
+    "gpu_start_mib": 4,
+    "gpu_end_mib": 9840,
+    "gpu_peak_mib": 9776,
+    "torch_peak_allocated_mib": 9555,
+    "torch_peak_reserved_mib": 9572
+  },
+  "generations": [
+    {
+      "name": "generate_00-cyrillic-poster",
+      "seconds": 37.53265648102388,
+      "gpu_start_mib": 9840,
+      "gpu_end_mib": 12870,
+      "gpu_peak_mib": 12870,
+      "torch_peak_allocated_mib": 12005,
+      "torch_peak_reserved_mib": 12538,
+      "prompt_id": "00-cyrillic-poster",
+      "title": "Cyrillic event poster",
+      "seed": 41001,
+      "width": 1024,
+      "height": 1024,
+      "image": "sdnq_uint4_static_qmm/images/00_00-cyrillic-poster.png"
+    },
+    {
+      "name": "generate_01-long-text-bakery-ad",
+      "seconds": 31.206937978044152,
+      "gpu_start_mib": 9906,
+      "gpu_end_mib": 12974,
+      "gpu_peak_mib": 12974,
+      "torch_peak_allocated_mib": 12067,
+      "torch_peak_reserved_mib": 12640,
+      "prompt_id": "01-long-text-bakery-ad",
+      "title": "Long text product ad",
+      "seed": 41002,
+      "width": 896,
+      "height": 1200,
+      "image": "sdnq_uint4_static_qmm/images/01_01-long-text-bakery-ad.png"
+    },
+    {
+      "name": "generate_02-technical-diagram",
+      "seconds": 27.878881530370563,
+      "gpu_start_mib": 9906,
+      "gpu_end_mib": 12952,
+      "gpu_peak_mib": 12952,
+      "torch_peak_allocated_mib": 12066,
+      "torch_peak_reserved_mib": 12618,
+      "prompt_id": "02-technical-diagram",
+      "title": "Technical diagram",
+      "seed": 41003,
+      "width": 1200,
+      "height": 896,
+      "image": "sdnq_uint4_static_qmm/images/02_02-technical-diagram.png"
+    },
+    {
+      "name": "generate_03-four-panel-comic",
+      "seconds": 50.17965244408697,
+      "gpu_start_mib": 9906,
+      "gpu_end_mib": 12874,
+      "gpu_peak_mib": 12874,
+      "torch_peak_allocated_mib": 12010,
+      "torch_peak_reserved_mib": 12540,
+      "prompt_id": "03-four-panel-comic",
+      "title": "Four-panel comic",
+      "seed": 41004,
+      "width": 1024,
+      "height": 1024,
+      "image": "sdnq_uint4_static_qmm/images/03_03-four-panel-comic.png"
+    },
+    {
+      "name": "generate_04-public-domain-painter-fusion",
+      "seconds": 39.54749615164474,
+      "gpu_start_mib": 9906,
+      "gpu_end_mib": 12872,
+      "gpu_peak_mib": 12872,
+      "torch_peak_allocated_mib": 12007,
+      "torch_peak_reserved_mib": 12538,
+      "prompt_id": "04-public-domain-painter-fusion",
+      "title": "Painterly style fusion",
+      "seed": 41005,
+      "width": 1024,
+      "height": 1024,
+      "image": "sdnq_uint4_static_qmm/images/04_04-public-domain-painter-fusion.png"
+    },
+    {
+      "name": "generate_05-dashboard-ui",
+      "seconds": 52.02436265302822,
+      "gpu_start_mib": 9908,
+      "gpu_end_mib": 12912,
+      "gpu_peak_mib": 12912,
+      "torch_peak_allocated_mib": 12029,
+      "torch_peak_reserved_mib": 12578,
+      "prompt_id": "05-dashboard-ui",
+      "title": "Dense UI dashboard",
+      "seed": 41006,
+      "width": 1376,
+      "height": 768,
+      "image": "sdnq_uint4_static_qmm/images/05_05-dashboard-ui.png"
+    },
+    {
+      "name": "generate_06-glass-still-life",
+      "seconds": 41.64155141590163,
+      "gpu_start_mib": 9906,
+      "gpu_end_mib": 12872,
+      "gpu_peak_mib": 12872,
+      "torch_peak_allocated_mib": 12007,
+      "torch_peak_reserved_mib": 12538,
+      "prompt_id": "06-glass-still-life",
+      "title": "Glass and reflections",
+      "seed": 41007,
+      "width": 1024,
+      "height": 1024,
+      "image": "sdnq_uint4_static_qmm/images/06_06-glass-still-life.png"
+    },
+    {
+      "name": "generate_07-botanical-field-guide",
+      "seconds": 38.91597466357052,
+      "gpu_start_mib": 9908,
+      "gpu_end_mib": 12952,
+      "gpu_peak_mib": 12952,
+      "torch_peak_allocated_mib": 12069,
+      "torch_peak_reserved_mib": 12618,
+      "prompt_id": "07-botanical-field-guide",
+      "title": "Field guide plate",
+      "seed": 41008,
+      "width": 896,
+      "height": 1200,
+      "image": "sdnq_uint4_static_qmm/images/07_07-botanical-field-guide.png"
+    },
+    {
+      "name": "generate_08-restaurant-menu-board",
+      "seconds": 28.993361878208816,
+      "gpu_start_mib": 9906,
+      "gpu_end_mib": 12872,
+      "gpu_peak_mib": 12872,
+      "torch_peak_allocated_mib": 12005,
+      "torch_peak_reserved_mib": 12538,
+      "prompt_id": "08-restaurant-menu-board",
+      "title": "Menu board text",
+      "seed": 41009,
+      "width": 1024,
+      "height": 1024,
+      "image": "sdnq_uint4_static_qmm/images/08_08-restaurant-menu-board.png"
+    },
+    {
+      "name": "generate_09-isometric-city-map",
+      "seconds": 48.68713657883927,
+      "gpu_start_mib": 9908,
+      "gpu_end_mib": 12952,
+      "gpu_peak_mib": 12952,
+      "torch_peak_allocated_mib": 12072,
+      "torch_peak_reserved_mib": 12618,
+      "prompt_id": "09-isometric-city-map",
+      "title": "Isometric map",
+      "seed": 41010,
+      "width": 1200,
+      "height": 896,
+      "image": "sdnq_uint4_static_qmm/images/09_09-isometric-city-map.png"
+    }
+  ]
+}

model_index.json

@@ -0,0 +1,33 @@
+{
+  "_class_name": "ErnieImagePipeline",
+  "_diffusers_version": "0.39.0.dev0",
+  "_name_or_path": "baidu/ERNIE-Image-Turbo",
+  "pe": [
+    "transformers",
+    "Ministral3ForCausalLM"
+  ],
+  "pe_tokenizer": [
+    "transformers",
+    "TokenizersBackend"
+  ],
+  "scheduler": [
+    "diffusers",
+    "FlowMatchEulerDiscreteScheduler"
+  ],
+  "text_encoder": [
+    "transformers",
+    "Mistral3Model"
+  ],
+  "tokenizer": [
+    "transformers",
+    "TokenizersBackend"
+  ],
+  "transformer": [
+    "diffusers",
+    "ErnieImageTransformer2DModel"
+  ],
+  "vae": [
+    "diffusers",
+    "AutoencoderKLFlux2"
+  ]
+}

pe/config.json

@@ -0,0 +1,151 @@
+{
+  "architectures": [
+    "Ministral3ForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "dtype": "bfloat16",
+  "eos_token_id": 2,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 3072,
+  "initializer_range": 0.02,
+  "intermediate_size": 9216,
+  "max_position_embeddings": 262144,
+  "model_type": "ministral3",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 26,
+  "num_key_value_heads": 8,
+  "pad_token_id": 11,
+  "quantization_config": {
+    "add_skip_keys": false,
+    "dequantize_fp32": false,
+    "dynamic_loss_threshold": null,
+    "group_size": 0,
+    "is_integer": true,
+    "is_training": false,
+    "modules_dtype_dict": {},
+    "modules_quant_config": {},
+    "modules_to_not_convert": [
+      ".img_out",
+      ".proj_out",
+      ".emb_in",
+      ".final_layer",
+      "patch_embed",
+      ".time_embed",
+      "multi_modal_projector",
+      ".condition_embedder",
+      ".t_embedder",
+      "lm_head.weight",
+      "wte",
+      "model.embed_tokens",
+      "lm_head",
+      ".txt_out",
+      "time_text_embed",
+      ".context_embedder",
+      ".txt_in",
+      ".emb_out",
+      ".norm_out",
+      ".img_in",
+      ".vid_in",
+      ".x_embedder",
+      "patch_embedding",
+      "patch_emb",
+      ".vid_out",
+      ".y_embedder",
+      "model.embed_tokens.weight",
+      "lm_head.weight",
+      "model.embed_tokens.weight",
+      "model.layers.0.input_layernorm.weight",
+      "model.layers.0.post_attention_layernorm.weight",
+      "model.layers.1.input_layernorm.weight",
+      "model.layers.1.post_attention_layernorm.weight",
+      "model.layers.2.input_layernorm.weight",
+      "model.layers.2.post_attention_layernorm.weight",
+      "model.layers.3.input_layernorm.weight",
+      "model.layers.3.post_attention_layernorm.weight",
+      "model.layers.4.input_layernorm.weight",
+      "model.layers.4.post_attention_layernorm.weight",
+      "model.layers.5.input_layernorm.weight",
+      "model.layers.5.post_attention_layernorm.weight",
+      "model.layers.6.input_layernorm.weight",
+      "model.layers.6.post_attention_layernorm.weight",
+      "model.layers.7.input_layernorm.weight",
+      "model.layers.7.post_attention_layernorm.weight",
+      "model.layers.8.input_layernorm.weight",
+      "model.layers.8.post_attention_layernorm.weight",
+      "model.layers.9.input_layernorm.weight",
+      "model.layers.9.post_attention_layernorm.weight",
+      "model.layers.10.input_layernorm.weight",
+      "model.layers.10.post_attention_layernorm.weight",
+      "model.layers.11.input_layernorm.weight",
+      "model.layers.11.post_attention_layernorm.weight",
+      "model.layers.12.input_layernorm.weight",
+      "model.layers.12.post_attention_layernorm.weight",
+      "model.layers.13.input_layernorm.weight",
+      "model.layers.13.post_attention_layernorm.weight",
+      "model.layers.14.input_layernorm.weight",
+      "model.layers.14.post_attention_layernorm.weight",
+      "model.layers.15.input_layernorm.weight",
+      "model.layers.15.post_attention_layernorm.weight",
+      "model.layers.16.input_layernorm.weight",
+      "model.layers.16.post_attention_layernorm.weight",
+      "model.layers.17.input_layernorm.weight",
+      "model.layers.17.post_attention_layernorm.weight",
+      "model.layers.18.input_layernorm.weight",
+      "model.layers.18.post_attention_layernorm.weight",
+      "model.layers.19.input_layernorm.weight",
+      "model.layers.19.post_attention_layernorm.weight",
+      "model.layers.20.input_layernorm.weight",
+      "model.layers.20.post_attention_layernorm.weight",
+      "model.layers.21.input_layernorm.weight",
+      "model.layers.21.post_attention_layernorm.weight",
+      "model.layers.22.input_layernorm.weight",
+      "model.layers.22.post_attention_layernorm.weight",
+      "model.layers.23.input_layernorm.weight",
+      "model.layers.23.post_attention_layernorm.weight",
+      "model.layers.24.input_layernorm.weight",
+      "model.layers.24.post_attention_layernorm.weight",
+      "model.layers.25.input_layernorm.weight",
+      "model.layers.25.post_attention_layernorm.weight",
+      "model.norm.weight"
+    ],
+    "modules_to_not_use_matmul": [],
+    "non_blocking": false,
+    "quant_conv": false,
+    "quant_embedding": false,
+    "quant_method": "sdnq",
+    "quantization_device": null,
+    "quantized_matmul_dtype": null,
+    "return_device": null,
+    "sdnq_version": "0.1.9",
+    "svd_rank": 32,
+    "svd_steps": 8,
+    "use_dynamic_quantization": false,
+    "use_grad_ckpt": true,
+    "use_quantized_matmul": false,
+    "use_quantized_matmul_conv": false,
+    "use_static_quantization": true,
+    "use_stochastic_rounding": false,
+    "use_svd": false,
+    "weights_dtype": "uint4"
+  },
+  "rms_norm_eps": 1e-05,
+  "rope_parameters": {
+    "beta_fast": 32.0,
+    "beta_slow": 1.0,
+    "factor": 16.0,
+    "llama_4_scaling_beta": 0.1,
+    "mscale": 1.0,
+    "mscale_all_dim": 1.0,
+    "original_max_position_embeddings": 16384,
+    "rope_theta": 1000000.0,
+    "rope_type": "yarn",
+    "type": "yarn"
+  },
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "transformers_version": "5.8.0.dev0",
+  "use_cache": true,
+  "vocab_size": 131072
+}

pe/generation_config.json

@@ -0,0 +1,8 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "pad_token_id": 11,
+  "transformers_version": "5.8.0.dev0",
+  "use_cache": true
+}

pe/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b11d70dfd43b69ce8d57a51d467726f776d79bfc41a029e2bf5a9e0eae8c8a1
+size 2507932808

pe/quantization_config.json

@@ -0,0 +1,113 @@
+{
+  "add_skip_keys": false,
+  "dequantize_fp32": false,
+  "dynamic_loss_threshold": null,
+  "group_size": 0,
+  "is_integer": true,
+  "is_training": false,
+  "modules_dtype_dict": {},
+  "modules_quant_config": {},
+  "modules_to_not_convert": [
+    ".img_out",
+    ".proj_out",
+    ".emb_in",
+    ".final_layer",
+    "patch_embed",
+    ".time_embed",
+    "multi_modal_projector",
+    ".condition_embedder",
+    ".t_embedder",
+    "lm_head.weight",
+    "wte",
+    "model.embed_tokens",
+    "lm_head",
+    ".txt_out",
+    "time_text_embed",
+    ".context_embedder",
+    ".txt_in",
+    ".emb_out",
+    ".norm_out",
+    ".img_in",
+    ".vid_in",
+    ".x_embedder",
+    "patch_embedding",
+    "patch_emb",
+    ".vid_out",
+    ".y_embedder",
+    "model.embed_tokens.weight",
+    "lm_head.weight",
+    "model.embed_tokens.weight",
+    "model.layers.0.input_layernorm.weight",
+    "model.layers.0.post_attention_layernorm.weight",
+    "model.layers.1.input_layernorm.weight",
+    "model.layers.1.post_attention_layernorm.weight",
+    "model.layers.2.input_layernorm.weight",
+    "model.layers.2.post_attention_layernorm.weight",
+    "model.layers.3.input_layernorm.weight",
+    "model.layers.3.post_attention_layernorm.weight",
+    "model.layers.4.input_layernorm.weight",
+    "model.layers.4.post_attention_layernorm.weight",
+    "model.layers.5.input_layernorm.weight",
+    "model.layers.5.post_attention_layernorm.weight",
+    "model.layers.6.input_layernorm.weight",
+    "model.layers.6.post_attention_layernorm.weight",
+    "model.layers.7.input_layernorm.weight",
+    "model.layers.7.post_attention_layernorm.weight",
+    "model.layers.8.input_layernorm.weight",
+    "model.layers.8.post_attention_layernorm.weight",
+    "model.layers.9.input_layernorm.weight",
+    "model.layers.9.post_attention_layernorm.weight",
+    "model.layers.10.input_layernorm.weight",
+    "model.layers.10.post_attention_layernorm.weight",
+    "model.layers.11.input_layernorm.weight",
+    "model.layers.11.post_attention_layernorm.weight",
+    "model.layers.12.input_layernorm.weight",
+    "model.layers.12.post_attention_layernorm.weight",
+    "model.layers.13.input_layernorm.weight",
+    "model.layers.13.post_attention_layernorm.weight",
+    "model.layers.14.input_layernorm.weight",
+    "model.layers.14.post_attention_layernorm.weight",
+    "model.layers.15.input_layernorm.weight",
+    "model.layers.15.post_attention_layernorm.weight",
+    "model.layers.16.input_layernorm.weight",
+    "model.layers.16.post_attention_layernorm.weight",
+    "model.layers.17.input_layernorm.weight",
+    "model.layers.17.post_attention_layernorm.weight",
+    "model.layers.18.input_layernorm.weight",
+    "model.layers.18.post_attention_layernorm.weight",
+    "model.layers.19.input_layernorm.weight",
+    "model.layers.19.post_attention_layernorm.weight",
+    "model.layers.20.input_layernorm.weight",
+    "model.layers.20.post_attention_layernorm.weight",
+    "model.layers.21.input_layernorm.weight",
+    "model.layers.21.post_attention_layernorm.weight",
+    "model.layers.22.input_layernorm.weight",
+    "model.layers.22.post_attention_layernorm.weight",
+    "model.layers.23.input_layernorm.weight",
+    "model.layers.23.post_attention_layernorm.weight",
+    "model.layers.24.input_layernorm.weight",
+    "model.layers.24.post_attention_layernorm.weight",
+    "model.layers.25.input_layernorm.weight",
+    "model.layers.25.post_attention_layernorm.weight",
+    "model.norm.weight"
+  ],
+  "modules_to_not_use_matmul": [],
+  "non_blocking": false,
+  "quant_conv": false,
+  "quant_embedding": false,
+  "quant_method": "sdnq",
+  "quantization_device": null,
+  "quantized_matmul_dtype": null,
+  "return_device": null,
+  "sdnq_version": "0.1.9",
+  "svd_rank": 32,
+  "svd_steps": 8,
+  "use_dynamic_quantization": false,
+  "use_grad_ckpt": true,
+  "use_quantized_matmul": false,
+  "use_quantized_matmul_conv": false,
+  "use_static_quantization": true,
+  "use_stochastic_rounding": false,
+  "use_svd": false,
+  "weights_dtype": "uint4"
+}

pe_tokenizer/chat_template.jinja

@@ -0,0 +1,8 @@
+{{- bos_token }}[SYSTEM_PROMPT]你是一个专业的文生图 Prompt 增强助手。你将收到用户的简短图片描述及目标生成分辨率，请据此扩写为一段内容丰富、细节充分的视觉描述，以帮助文生图模型生成高质量的图片。仅输出增强后的描述，不要包含任何解释或前缀。[/SYSTEM_PROMPT]
+{%- for message in messages %}
+{%- if message['role'] == 'user' %}
+{{- '[INST]' + message['content'] + '[/INST]' }}
+{%- elif message['role'] == 'assistant' %}
+{%- generation %}{{ message['content'] }}{{ eos_token }}{% endgeneration %}
+{%- endif %}
+{%- endfor %}

pe_tokenizer/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:286acad9b0e27fce778ac429763536accf618ccb6ed72963b6f94685e531c5c7
+size 17077402

pe_tokenizer/tokenizer_config.json

@@ -0,0 +1,1014 @@
+{
+  "backend": "tokenizers",
+  "bos_token": "<s>",
+  "eos_token": "</s>",
+  "extra_special_tokens": [
+    "<unk>",
+    "<s>",
+    "</s>",
+    "[INST]",
+    "[/INST]",
+    "[AVAILABLE_TOOLS]",
+    "[/AVAILABLE_TOOLS]",
+    "[TOOL_RESULTS]",
+    "[/TOOL_RESULTS]",
+    "[TOOL_CALLS]",
+    "[IMG]",
+    "<pad>",
+    "[IMG_BREAK]",
+    "[IMG_END]",
+    "[PREFIX]",
+    "[MIDDLE]",
+    "[SUFFIX]",
+    "[SYSTEM_PROMPT]",
+    "[/SYSTEM_PROMPT]",
+    "[TOOL_CONTENT]",
+    "<SPECIAL_20>",
+    "<SPECIAL_21>",
+    "<SPECIAL_22>",
+    "<SPECIAL_23>",
+    "[AUDIO]",
+    "[BEGIN_AUDIO]",
+    "<SPECIAL_26>",
+    "<SPECIAL_27>",
+    "<SPECIAL_28>",
+    "<SPECIAL_29>",
+    "<SPECIAL_30>",
+    "<SPECIAL_31>",
+    "[ARGS]",
+    "[CALL_ID]",
+    "[THINK]",
+    "[/THINK]",
+    "<SPECIAL_36>",
+    "<SPECIAL_37>",
+    "<SPECIAL_38>",
+    "<SPECIAL_39>",
+    "<SPECIAL_40>",
+    "<SPECIAL_41>",
+    "<SPECIAL_42>",
+    "<SPECIAL_43>",
+    "<SPECIAL_44>",
+    "<SPECIAL_45>",
+    "<SPECIAL_46>",
+    "<SPECIAL_47>",
+    "<SPECIAL_48>",
+    "<SPECIAL_49>",
+    "<SPECIAL_50>",
+    "<SPECIAL_51>",
+    "<SPECIAL_52>",
+    "<SPECIAL_53>",
+    "<SPECIAL_54>",
+    "<SPECIAL_55>",
+    "<SPECIAL_56>",
+    "<SPECIAL_57>",
+    "<SPECIAL_58>",
+    "<SPECIAL_59>",
+    "<SPECIAL_60>",
+    "<SPECIAL_61>",
+    "<SPECIAL_62>",
+    "<SPECIAL_63>",
+    "<SPECIAL_64>",
+    "<SPECIAL_65>",
+    "<SPECIAL_66>",
+    "<SPECIAL_67>",
+    "<SPECIAL_68>",
+    "<SPECIAL_69>",
+    "<SPECIAL_70>",
+    "<SPECIAL_71>",
+    "<SPECIAL_72>",
+    "<SPECIAL_73>",
+    "<SPECIAL_74>",
+    "<SPECIAL_75>",
+    "<SPECIAL_76>",
+    "<SPECIAL_77>",
+    "<SPECIAL_78>",
+    "<SPECIAL_79>",
+    "<SPECIAL_80>",
+    "<SPECIAL_81>",
+    "<SPECIAL_82>",
+    "<SPECIAL_83>",
+    "<SPECIAL_84>",
+    "<SPECIAL_85>",
+    "<SPECIAL_86>",
+    "<SPECIAL_87>",
+    "<SPECIAL_88>",
+    "<SPECIAL_89>",
+    "<SPECIAL_90>",
+    "<SPECIAL_91>",
+    "<SPECIAL_92>",
+    "<SPECIAL_93>",
+    "<SPECIAL_94>",
+    "<SPECIAL_95>",
+    "<SPECIAL_96>",
+    "<SPECIAL_97>",
+    "<SPECIAL_98>",
+    "<SPECIAL_99>",
+    "<SPECIAL_100>",
+    "<SPECIAL_101>",
+    "<SPECIAL_102>",
+    "<SPECIAL_103>",
+    "<SPECIAL_104>",
+    "<SPECIAL_105>",
+    "<SPECIAL_106>",
+    "<SPECIAL_107>",
+    "<SPECIAL_108>",
+    "<SPECIAL_109>",
+    "<SPECIAL_110>",
+    "<SPECIAL_111>",
+    "<SPECIAL_112>",
+    "<SPECIAL_113>",
+    "<SPECIAL_114>",
+    "<SPECIAL_115>",
+    "<SPECIAL_116>",
+    "<SPECIAL_117>",
+    "<SPECIAL_118>",
+    "<SPECIAL_119>",
+    "<SPECIAL_120>",
+    "<SPECIAL_121>",
+    "<SPECIAL_122>",
+    "<SPECIAL_123>",
+    "<SPECIAL_124>",
+    "<SPECIAL_125>",
+    "<SPECIAL_126>",
+    "<SPECIAL_127>",
+    "<SPECIAL_128>",
+    "<SPECIAL_129>",
+    "<SPECIAL_130>",
+    "<SPECIAL_131>",
+    "<SPECIAL_132>",
+    "<SPECIAL_133>",
+    "<SPECIAL_134>",
+    "<SPECIAL_135>",
+    "<SPECIAL_136>",
+    "<SPECIAL_137>",
+    "<SPECIAL_138>",
+    "<SPECIAL_139>",
+    "<SPECIAL_140>",
+    "<SPECIAL_141>",
+    "<SPECIAL_142>",
+    "<SPECIAL_143>",
+    "<SPECIAL_144>",
+    "<SPECIAL_145>",
+    "<SPECIAL_146>",
+    "<SPECIAL_147>",
+    "<SPECIAL_148>",
+    "<SPECIAL_149>",
+    "<SPECIAL_150>",
+    "<SPECIAL_151>",
+    "<SPECIAL_152>",
+    "<SPECIAL_153>",
+    "<SPECIAL_154>",
+    "<SPECIAL_155>",
+    "<SPECIAL_156>",
+    "<SPECIAL_157>",
+    "<SPECIAL_158>",
+    "<SPECIAL_159>",
+    "<SPECIAL_160>",
+    "<SPECIAL_161>",
+    "<SPECIAL_162>",
+    "<SPECIAL_163>",
+    "<SPECIAL_164>",
+    "<SPECIAL_165>",
+    "<SPECIAL_166>",
+    "<SPECIAL_167>",
+    "<SPECIAL_168>",
+    "<SPECIAL_169>",
+    "<SPECIAL_170>",
+    "<SPECIAL_171>",
+    "<SPECIAL_172>",
+    "<SPECIAL_173>",
+    "<SPECIAL_174>",
+    "<SPECIAL_175>",
+    "<SPECIAL_176>",
+    "<SPECIAL_177>",
+    "<SPECIAL_178>",
+    "<SPECIAL_179>",
+    "<SPECIAL_180>",
+    "<SPECIAL_181>",
+    "<SPECIAL_182>",
+    "<SPECIAL_183>",
+    "<SPECIAL_184>",
+    "<SPECIAL_185>",
+    "<SPECIAL_186>",
+    "<SPECIAL_187>",
+    "<SPECIAL_188>",
+    "<SPECIAL_189>",
+    "<SPECIAL_190>",
+    "<SPECIAL_191>",
+    "<SPECIAL_192>",
+    "<SPECIAL_193>",
+    "<SPECIAL_194>",
+    "<SPECIAL_195>",
+    "<SPECIAL_196>",
+    "<SPECIAL_197>",
+    "<SPECIAL_198>",
+    "<SPECIAL_199>",
+    "<SPECIAL_200>",
+    "<SPECIAL_201>",
+    "<SPECIAL_202>",
+    "<SPECIAL_203>",
+    "<SPECIAL_204>",
+    "<SPECIAL_205>",
+    "<SPECIAL_206>",
+    "<SPECIAL_207>",
+    "<SPECIAL_208>",
+    "<SPECIAL_209>",
+    "<SPECIAL_210>",
+    "<SPECIAL_211>",
+    "<SPECIAL_212>",
+    "<SPECIAL_213>",
+    "<SPECIAL_214>",
+    "<SPECIAL_215>",
+    "<SPECIAL_216>",
+    "<SPECIAL_217>",
+    "<SPECIAL_218>",
+    "<SPECIAL_219>",
+    "<SPECIAL_220>",
+    "<SPECIAL_221>",
+    "<SPECIAL_222>",
+    "<SPECIAL_223>",
+    "<SPECIAL_224>",
+    "<SPECIAL_225>",
+    "<SPECIAL_226>",
+    "<SPECIAL_227>",
+    "<SPECIAL_228>",
+    "<SPECIAL_229>",
+    "<SPECIAL_230>",
+    "<SPECIAL_231>",
+    "<SPECIAL_232>",
+    "<SPECIAL_233>",
+    "<SPECIAL_234>",
+    "<SPECIAL_235>",
+    "<SPECIAL_236>",
+    "<SPECIAL_237>",
+    "<SPECIAL_238>",
+    "<SPECIAL_239>",
+    "<SPECIAL_240>",
+    "<SPECIAL_241>",
+    "<SPECIAL_242>",
+    "<SPECIAL_243>",
+    "<SPECIAL_244>",
+    "<SPECIAL_245>",
+    "<SPECIAL_246>",
+    "<SPECIAL_247>",
+    "<SPECIAL_248>",
+    "<SPECIAL_249>",
+    "<SPECIAL_250>",
+    "<SPECIAL_251>",
+    "<SPECIAL_252>",
+    "<SPECIAL_253>",
+    "<SPECIAL_254>",
+    "<SPECIAL_255>",
+    "<SPECIAL_256>",
+    "<SPECIAL_257>",
+    "<SPECIAL_258>",
+    "<SPECIAL_259>",
+    "<SPECIAL_260>",
+    "<SPECIAL_261>",
+    "<SPECIAL_262>",
+    "<SPECIAL_263>",
+    "<SPECIAL_264>",
+    "<SPECIAL_265>",
+    "<SPECIAL_266>",
+    "<SPECIAL_267>",
+    "<SPECIAL_268>",
+    "<SPECIAL_269>",
+    "<SPECIAL_270>",
+    "<SPECIAL_271>",
+    "<SPECIAL_272>",
+    "<SPECIAL_273>",
+    "<SPECIAL_274>",
+    "<SPECIAL_275>",
+    "<SPECIAL_276>",
+    "<SPECIAL_277>",
+    "<SPECIAL_278>",
+    "<SPECIAL_279>",
+    "<SPECIAL_280>",
+    "<SPECIAL_281>",
+    "<SPECIAL_282>",
+    "<SPECIAL_283>",
+    "<SPECIAL_284>",
+    "<SPECIAL_285>",
+    "<SPECIAL_286>",
+    "<SPECIAL_287>",
+    "<SPECIAL_288>",
+    "<SPECIAL_289>",
+    "<SPECIAL_290>",
+    "<SPECIAL_291>",
+    "<SPECIAL_292>",
+    "<SPECIAL_293>",
+    "<SPECIAL_294>",
+    "<SPECIAL_295>",
+    "<SPECIAL_296>",
+    "<SPECIAL_297>",
+    "<SPECIAL_298>",
+    "<SPECIAL_299>",
+    "<SPECIAL_300>",
+    "<SPECIAL_301>",
+    "<SPECIAL_302>",
+    "<SPECIAL_303>",
+    "<SPECIAL_304>",
+    "<SPECIAL_305>",
+    "<SPECIAL_306>",
+    "<SPECIAL_307>",
+    "<SPECIAL_308>",
+    "<SPECIAL_309>",
+    "<SPECIAL_310>",
+    "<SPECIAL_311>",
+    "<SPECIAL_312>",
+    "<SPECIAL_313>",
+    "<SPECIAL_314>",
+    "<SPECIAL_315>",
+    "<SPECIAL_316>",
+    "<SPECIAL_317>",
+    "<SPECIAL_318>",
+    "<SPECIAL_319>",
+    "<SPECIAL_320>",
+    "<SPECIAL_321>",
+    "<SPECIAL_322>",
+    "<SPECIAL_323>",
+    "<SPECIAL_324>",
+    "<SPECIAL_325>",
+    "<SPECIAL_326>",
+    "<SPECIAL_327>",
+    "<SPECIAL_328>",
+    "<SPECIAL_329>",
+    "<SPECIAL_330>",
+    "<SPECIAL_331>",
+    "<SPECIAL_332>",
+    "<SPECIAL_333>",
+    "<SPECIAL_334>",
+    "<SPECIAL_335>",
+    "<SPECIAL_336>",
+    "<SPECIAL_337>",
+    "<SPECIAL_338>",
+    "<SPECIAL_339>",
+    "<SPECIAL_340>",
+    "<SPECIAL_341>",
+    "<SPECIAL_342>",
+    "<SPECIAL_343>",
+    "<SPECIAL_344>",
+    "<SPECIAL_345>",
+    "<SPECIAL_346>",
+    "<SPECIAL_347>",
+    "<SPECIAL_348>",
+    "<SPECIAL_349>",
+    "<SPECIAL_350>",
+    "<SPECIAL_351>",
+    "<SPECIAL_352>",
+    "<SPECIAL_353>",
+    "<SPECIAL_354>",
+    "<SPECIAL_355>",
+    "<SPECIAL_356>",
+    "<SPECIAL_357>",
+    "<SPECIAL_358>",
+    "<SPECIAL_359>",
+    "<SPECIAL_360>",
+    "<SPECIAL_361>",
+    "<SPECIAL_362>",
+    "<SPECIAL_363>",
+    "<SPECIAL_364>",
+    "<SPECIAL_365>",
+    "<SPECIAL_366>",
+    "<SPECIAL_367>",
+    "<SPECIAL_368>",
+    "<SPECIAL_369>",
+    "<SPECIAL_370>",
+    "<SPECIAL_371>",
+    "<SPECIAL_372>",
+    "<SPECIAL_373>",
+    "<SPECIAL_374>",
+    "<SPECIAL_375>",
+    "<SPECIAL_376>",
+    "<SPECIAL_377>",
+    "<SPECIAL_378>",
+    "<SPECIAL_379>",
+    "<SPECIAL_380>",
+    "<SPECIAL_381>",
+    "<SPECIAL_382>",
+    "<SPECIAL_383>",
+    "<SPECIAL_384>",
+    "<SPECIAL_385>",
+    "<SPECIAL_386>",
+    "<SPECIAL_387>",
+    "<SPECIAL_388>",
+    "<SPECIAL_389>",
+    "<SPECIAL_390>",
+    "<SPECIAL_391>",
+    "<SPECIAL_392>",
+    "<SPECIAL_393>",
+    "<SPECIAL_394>",
+    "<SPECIAL_395>",
+    "<SPECIAL_396>",
+    "<SPECIAL_397>",
+    "<SPECIAL_398>",
+    "<SPECIAL_399>",
+    "<SPECIAL_400>",
+    "<SPECIAL_401>",
+    "<SPECIAL_402>",
+    "<SPECIAL_403>",
+    "<SPECIAL_404>",
+    "<SPECIAL_405>",
+    "<SPECIAL_406>",
+    "<SPECIAL_407>",
+    "<SPECIAL_408>",
+    "<SPECIAL_409>",
+    "<SPECIAL_410>",
+    "<SPECIAL_411>",
+    "<SPECIAL_412>",
+    "<SPECIAL_413>",
+    "<SPECIAL_414>",
+    "<SPECIAL_415>",
+    "<SPECIAL_416>",
+    "<SPECIAL_417>",
+    "<SPECIAL_418>",
+    "<SPECIAL_419>",
+    "<SPECIAL_420>",
+    "<SPECIAL_421>",
+    "<SPECIAL_422>",
+    "<SPECIAL_423>",
+    "<SPECIAL_424>",
+    "<SPECIAL_425>",
+    "<SPECIAL_426>",
+    "<SPECIAL_427>",
+    "<SPECIAL_428>",
+    "<SPECIAL_429>",
+    "<SPECIAL_430>",
+    "<SPECIAL_431>",
+    "<SPECIAL_432>",
+    "<SPECIAL_433>",
+    "<SPECIAL_434>",
+    "<SPECIAL_435>",
+    "<SPECIAL_436>",
+    "<SPECIAL_437>",
+    "<SPECIAL_438>",
+    "<SPECIAL_439>",
+    "<SPECIAL_440>",
+    "<SPECIAL_441>",
+    "<SPECIAL_442>",
+    "<SPECIAL_443>",
+    "<SPECIAL_444>",
+    "<SPECIAL_445>",
+    "<SPECIAL_446>",
+    "<SPECIAL_447>",
+    "<SPECIAL_448>",
+    "<SPECIAL_449>",
+    "<SPECIAL_450>",
+    "<SPECIAL_451>",
+    "<SPECIAL_452>",
+    "<SPECIAL_453>",
+    "<SPECIAL_454>",
+    "<SPECIAL_455>",
+    "<SPECIAL_456>",
+    "<SPECIAL_457>",
+    "<SPECIAL_458>",
+    "<SPECIAL_459>",
+    "<SPECIAL_460>",
+    "<SPECIAL_461>",
+    "<SPECIAL_462>",
+    "<SPECIAL_463>",
+    "<SPECIAL_464>",
+    "<SPECIAL_465>",
+    "<SPECIAL_466>",
+    "<SPECIAL_467>",
+    "<SPECIAL_468>",
+    "<SPECIAL_469>",
+    "<SPECIAL_470>",
+    "<SPECIAL_471>",
+    "<SPECIAL_472>",
+    "<SPECIAL_473>",
+    "<SPECIAL_474>",
+    "<SPECIAL_475>",
+    "<SPECIAL_476>",
+    "<SPECIAL_477>",
+    "<SPECIAL_478>",
+    "<SPECIAL_479>",
+    "<SPECIAL_480>",
+    "<SPECIAL_481>",
+    "<SPECIAL_482>",
+    "<SPECIAL_483>",
+    "<SPECIAL_484>",
+    "<SPECIAL_485>",
+    "<SPECIAL_486>",
+    "<SPECIAL_487>",
+    "<SPECIAL_488>",
+    "<SPECIAL_489>",
+    "<SPECIAL_490>",
+    "<SPECIAL_491>",
+    "<SPECIAL_492>",
+    "<SPECIAL_493>",
+    "<SPECIAL_494>",
+    "<SPECIAL_495>",
+    "<SPECIAL_496>",
+    "<SPECIAL_497>",
+    "<SPECIAL_498>",
+    "<SPECIAL_499>",
+    "<SPECIAL_500>",
+    "<SPECIAL_501>",
+    "<SPECIAL_502>",
+    "<SPECIAL_503>",
+    "<SPECIAL_504>",
+    "<SPECIAL_505>",
+    "<SPECIAL_506>",
+    "<SPECIAL_507>",
+    "<SPECIAL_508>",
+    "<SPECIAL_509>",
+    "<SPECIAL_510>",
+    "<SPECIAL_511>",
+    "<SPECIAL_512>",
+    "<SPECIAL_513>",
+    "<SPECIAL_514>",
+    "<SPECIAL_515>",
+    "<SPECIAL_516>",
+    "<SPECIAL_517>",
+    "<SPECIAL_518>",
+    "<SPECIAL_519>",
+    "<SPECIAL_520>",
+    "<SPECIAL_521>",
+    "<SPECIAL_522>",
+    "<SPECIAL_523>",
+    "<SPECIAL_524>",
+    "<SPECIAL_525>",
+    "<SPECIAL_526>",
+    "<SPECIAL_527>",
+    "<SPECIAL_528>",
+    "<SPECIAL_529>",
+    "<SPECIAL_530>",
+    "<SPECIAL_531>",
+    "<SPECIAL_532>",
+    "<SPECIAL_533>",
+    "<SPECIAL_534>",
+    "<SPECIAL_535>",
+    "<SPECIAL_536>",
+    "<SPECIAL_537>",
+    "<SPECIAL_538>",
+    "<SPECIAL_539>",
+    "<SPECIAL_540>",
+    "<SPECIAL_541>",
+    "<SPECIAL_542>",
+    "<SPECIAL_543>",
+    "<SPECIAL_544>",
+    "<SPECIAL_545>",
+    "<SPECIAL_546>",
+    "<SPECIAL_547>",
+    "<SPECIAL_548>",
+    "<SPECIAL_549>",
+    "<SPECIAL_550>",
+    "<SPECIAL_551>",
+    "<SPECIAL_552>",
+    "<SPECIAL_553>",
+    "<SPECIAL_554>",
+    "<SPECIAL_555>",
+    "<SPECIAL_556>",
+    "<SPECIAL_557>",
+    "<SPECIAL_558>",
+    "<SPECIAL_559>",
+    "<SPECIAL_560>",
+    "<SPECIAL_561>",
+    "<SPECIAL_562>",
+    "<SPECIAL_563>",
+    "<SPECIAL_564>",
+    "<SPECIAL_565>",
+    "<SPECIAL_566>",
+    "<SPECIAL_567>",
+    "<SPECIAL_568>",
+    "<SPECIAL_569>",
+    "<SPECIAL_570>",
+    "<SPECIAL_571>",
+    "<SPECIAL_572>",
+    "<SPECIAL_573>",
+    "<SPECIAL_574>",
+    "<SPECIAL_575>",
+    "<SPECIAL_576>",
+    "<SPECIAL_577>",
+    "<SPECIAL_578>",
+    "<SPECIAL_579>",
+    "<SPECIAL_580>",
+    "<SPECIAL_581>",
+    "<SPECIAL_582>",
+    "<SPECIAL_583>",
+    "<SPECIAL_584>",
+    "<SPECIAL_585>",
+    "<SPECIAL_586>",
+    "<SPECIAL_587>",
+    "<SPECIAL_588>",
+    "<SPECIAL_589>",
+    "<SPECIAL_590>",
+    "<SPECIAL_591>",
+    "<SPECIAL_592>",
+    "<SPECIAL_593>",
+    "<SPECIAL_594>",
+    "<SPECIAL_595>",
+    "<SPECIAL_596>",
+    "<SPECIAL_597>",
+    "<SPECIAL_598>",
+    "<SPECIAL_599>",
+    "<SPECIAL_600>",
+    "<SPECIAL_601>",
+    "<SPECIAL_602>",
+    "<SPECIAL_603>",
+    "<SPECIAL_604>",
+    "<SPECIAL_605>",
+    "<SPECIAL_606>",
+    "<SPECIAL_607>",
+    "<SPECIAL_608>",
+    "<SPECIAL_609>",
+    "<SPECIAL_610>",
+    "<SPECIAL_611>",
+    "<SPECIAL_612>",
+    "<SPECIAL_613>",
+    "<SPECIAL_614>",
+    "<SPECIAL_615>",
+    "<SPECIAL_616>",
+    "<SPECIAL_617>",
+    "<SPECIAL_618>",
+    "<SPECIAL_619>",
+    "<SPECIAL_620>",
+    "<SPECIAL_621>",
+    "<SPECIAL_622>",
+    "<SPECIAL_623>",
+    "<SPECIAL_624>",
+    "<SPECIAL_625>",
+    "<SPECIAL_626>",
+    "<SPECIAL_627>",
+    "<SPECIAL_628>",
+    "<SPECIAL_629>",
+    "<SPECIAL_630>",
+    "<SPECIAL_631>",
+    "<SPECIAL_632>",
+    "<SPECIAL_633>",
+    "<SPECIAL_634>",
+    "<SPECIAL_635>",
+    "<SPECIAL_636>",
+    "<SPECIAL_637>",
+    "<SPECIAL_638>",
+    "<SPECIAL_639>",
+    "<SPECIAL_640>",
+    "<SPECIAL_641>",
+    "<SPECIAL_642>",
+    "<SPECIAL_643>",
+    "<SPECIAL_644>",
+    "<SPECIAL_645>",
+    "<SPECIAL_646>",
+    "<SPECIAL_647>",
+    "<SPECIAL_648>",
+    "<SPECIAL_649>",
+    "<SPECIAL_650>",
+    "<SPECIAL_651>",
+    "<SPECIAL_652>",
+    "<SPECIAL_653>",
+    "<SPECIAL_654>",
+    "<SPECIAL_655>",
+    "<SPECIAL_656>",
+    "<SPECIAL_657>",
+    "<SPECIAL_658>",
+    "<SPECIAL_659>",
+    "<SPECIAL_660>",
+    "<SPECIAL_661>",
+    "<SPECIAL_662>",
+    "<SPECIAL_663>",
+    "<SPECIAL_664>",
+    "<SPECIAL_665>",
+    "<SPECIAL_666>",
+    "<SPECIAL_667>",
+    "<SPECIAL_668>",
+    "<SPECIAL_669>",
+    "<SPECIAL_670>",
+    "<SPECIAL_671>",
+    "<SPECIAL_672>",
+    "<SPECIAL_673>",
+    "<SPECIAL_674>",
+    "<SPECIAL_675>",
+    "<SPECIAL_676>",
+    "<SPECIAL_677>",
+    "<SPECIAL_678>",
+    "<SPECIAL_679>",
+    "<SPECIAL_680>",
+    "<SPECIAL_681>",
+    "<SPECIAL_682>",
+    "<SPECIAL_683>",
+    "<SPECIAL_684>",
+    "<SPECIAL_685>",
+    "<SPECIAL_686>",
+    "<SPECIAL_687>",
+    "<SPECIAL_688>",
+    "<SPECIAL_689>",
+    "<SPECIAL_690>",
+    "<SPECIAL_691>",
+    "<SPECIAL_692>",
+    "<SPECIAL_693>",
+    "<SPECIAL_694>",
+    "<SPECIAL_695>",
+    "<SPECIAL_696>",
+    "<SPECIAL_697>",
+    "<SPECIAL_698>",
+    "<SPECIAL_699>",
+    "<SPECIAL_700>",
+    "<SPECIAL_701>",
+    "<SPECIAL_702>",
+    "<SPECIAL_703>",
+    "<SPECIAL_704>",
+    "<SPECIAL_705>",
+    "<SPECIAL_706>",
+    "<SPECIAL_707>",
+    "<SPECIAL_708>",
+    "<SPECIAL_709>",
+    "<SPECIAL_710>",
+    "<SPECIAL_711>",
+    "<SPECIAL_712>",
+    "<SPECIAL_713>",
+    "<SPECIAL_714>",
+    "<SPECIAL_715>",
+    "<SPECIAL_716>",
+    "<SPECIAL_717>",
+    "<SPECIAL_718>",
+    "<SPECIAL_719>",
+    "<SPECIAL_720>",
+    "<SPECIAL_721>",
+    "<SPECIAL_722>",
+    "<SPECIAL_723>",
+    "<SPECIAL_724>",
+    "<SPECIAL_725>",
+    "<SPECIAL_726>",
+    "<SPECIAL_727>",
+    "<SPECIAL_728>",
+    "<SPECIAL_729>",
+    "<SPECIAL_730>",
+    "<SPECIAL_731>",
+    "<SPECIAL_732>",
+    "<SPECIAL_733>",
+    "<SPECIAL_734>",
+    "<SPECIAL_735>",
+    "<SPECIAL_736>",
+    "<SPECIAL_737>",
+    "<SPECIAL_738>",
+    "<SPECIAL_739>",
+    "<SPECIAL_740>",
+    "<SPECIAL_741>",
+    "<SPECIAL_742>",
+    "<SPECIAL_743>",
+    "<SPECIAL_744>",
+    "<SPECIAL_745>",
+    "<SPECIAL_746>",
+    "<SPECIAL_747>",
+    "<SPECIAL_748>",
+    "<SPECIAL_749>",
+    "<SPECIAL_750>",
+    "<SPECIAL_751>",
+    "<SPECIAL_752>",
+    "<SPECIAL_753>",
+    "<SPECIAL_754>",
+    "<SPECIAL_755>",
+    "<SPECIAL_756>",
+    "<SPECIAL_757>",
+    "<SPECIAL_758>",
+    "<SPECIAL_759>",
+    "<SPECIAL_760>",
+    "<SPECIAL_761>",
+    "<SPECIAL_762>",
+    "<SPECIAL_763>",
+    "<SPECIAL_764>",
+    "<SPECIAL_765>",
+    "<SPECIAL_766>",
+    "<SPECIAL_767>",
+    "<SPECIAL_768>",
+    "<SPECIAL_769>",
+    "<SPECIAL_770>",
+    "<SPECIAL_771>",
+    "<SPECIAL_772>",
+    "<SPECIAL_773>",
+    "<SPECIAL_774>",
+    "<SPECIAL_775>",
+    "<SPECIAL_776>",
+    "<SPECIAL_777>",
+    "<SPECIAL_778>",
+    "<SPECIAL_779>",
+    "<SPECIAL_780>",
+    "<SPECIAL_781>",
+    "<SPECIAL_782>",
+    "<SPECIAL_783>",
+    "<SPECIAL_784>",
+    "<SPECIAL_785>",
+    "<SPECIAL_786>",
+    "<SPECIAL_787>",
+    "<SPECIAL_788>",
+    "<SPECIAL_789>",
+    "<SPECIAL_790>",
+    "<SPECIAL_791>",
+    "<SPECIAL_792>",
+    "<SPECIAL_793>",
+    "<SPECIAL_794>",
+    "<SPECIAL_795>",
+    "<SPECIAL_796>",
+    "<SPECIAL_797>",
+    "<SPECIAL_798>",
+    "<SPECIAL_799>",
+    "<SPECIAL_800>",
+    "<SPECIAL_801>",
+    "<SPECIAL_802>",
+    "<SPECIAL_803>",
+    "<SPECIAL_804>",
+    "<SPECIAL_805>",
+    "<SPECIAL_806>",
+    "<SPECIAL_807>",
+    "<SPECIAL_808>",
+    "<SPECIAL_809>",
+    "<SPECIAL_810>",
+    "<SPECIAL_811>",
+    "<SPECIAL_812>",
+    "<SPECIAL_813>",
+    "<SPECIAL_814>",
+    "<SPECIAL_815>",
+    "<SPECIAL_816>",
+    "<SPECIAL_817>",
+    "<SPECIAL_818>",
+    "<SPECIAL_819>",
+    "<SPECIAL_820>",
+    "<SPECIAL_821>",
+    "<SPECIAL_822>",
+    "<SPECIAL_823>",
+    "<SPECIAL_824>",
+    "<SPECIAL_825>",
+    "<SPECIAL_826>",
+    "<SPECIAL_827>",
+    "<SPECIAL_828>",
+    "<SPECIAL_829>",
+    "<SPECIAL_830>",
+    "<SPECIAL_831>",
+    "<SPECIAL_832>",
+    "<SPECIAL_833>",
+    "<SPECIAL_834>",
+    "<SPECIAL_835>",
+    "<SPECIAL_836>",
+    "<SPECIAL_837>",
+    "<SPECIAL_838>",
+    "<SPECIAL_839>",
+    "<SPECIAL_840>",
+    "<SPECIAL_841>",
+    "<SPECIAL_842>",
+    "<SPECIAL_843>",
+    "<SPECIAL_844>",
+    "<SPECIAL_845>",
+    "<SPECIAL_846>",
+    "<SPECIAL_847>",
+    "<SPECIAL_848>",
+    "<SPECIAL_849>",
+    "<SPECIAL_850>",
+    "<SPECIAL_851>",
+    "<SPECIAL_852>",
+    "<SPECIAL_853>",
+    "<SPECIAL_854>",
+    "<SPECIAL_855>",
+    "<SPECIAL_856>",
+    "<SPECIAL_857>",
+    "<SPECIAL_858>",
+    "<SPECIAL_859>",
+    "<SPECIAL_860>",
+    "<SPECIAL_861>",
+    "<SPECIAL_862>",
+    "<SPECIAL_863>",
+    "<SPECIAL_864>",
+    "<SPECIAL_865>",
+    "<SPECIAL_866>",
+    "<SPECIAL_867>",
+    "<SPECIAL_868>",
+    "<SPECIAL_869>",
+    "<SPECIAL_870>",
+    "<SPECIAL_871>",
+    "<SPECIAL_872>",
+    "<SPECIAL_873>",
+    "<SPECIAL_874>",
+    "<SPECIAL_875>",
+    "<SPECIAL_876>",
+    "<SPECIAL_877>",
+    "<SPECIAL_878>",
+    "<SPECIAL_879>",
+    "<SPECIAL_880>",
+    "<SPECIAL_881>",
+    "<SPECIAL_882>",
+    "<SPECIAL_883>",
+    "<SPECIAL_884>",
+    "<SPECIAL_885>",
+    "<SPECIAL_886>",
+    "<SPECIAL_887>",
+    "<SPECIAL_888>",
+    "<SPECIAL_889>",
+    "<SPECIAL_890>",
+    "<SPECIAL_891>",
+    "<SPECIAL_892>",
+    "<SPECIAL_893>",
+    "<SPECIAL_894>",
+    "<SPECIAL_895>",
+    "<SPECIAL_896>",
+    "<SPECIAL_897>",
+    "<SPECIAL_898>",
+    "<SPECIAL_899>",
+    "<SPECIAL_900>",
+    "<SPECIAL_901>",
+    "<SPECIAL_902>",
+    "<SPECIAL_903>",
+    "<SPECIAL_904>",
+    "<SPECIAL_905>",
+    "<SPECIAL_906>",
+    "<SPECIAL_907>",
+    "<SPECIAL_908>",
+    "<SPECIAL_909>",
+    "<SPECIAL_910>",
+    "<SPECIAL_911>",
+    "<SPECIAL_912>",
+    "<SPECIAL_913>",
+    "<SPECIAL_914>",
+    "<SPECIAL_915>",
+    "<SPECIAL_916>",
+    "<SPECIAL_917>",
+    "<SPECIAL_918>",
+    "<SPECIAL_919>",
+    "<SPECIAL_920>",
+    "<SPECIAL_921>",
+    "<SPECIAL_922>",
+    "<SPECIAL_923>",
+    "<SPECIAL_924>",
+    "<SPECIAL_925>",
+    "<SPECIAL_926>",
+    "<SPECIAL_927>",
+    "<SPECIAL_928>",
+    "<SPECIAL_929>",
+    "<SPECIAL_930>",
+    "<SPECIAL_931>",
+    "<SPECIAL_932>",
+    "<SPECIAL_933>",
+    "<SPECIAL_934>",
+    "<SPECIAL_935>",
+    "<SPECIAL_936>",
+    "<SPECIAL_937>",
+    "<SPECIAL_938>",
+    "<SPECIAL_939>",
+    "<SPECIAL_940>",
+    "<SPECIAL_941>",
+    "<SPECIAL_942>",
+    "<SPECIAL_943>",
+    "<SPECIAL_944>",
+    "<SPECIAL_945>",
+    "<SPECIAL_946>",
+    "<SPECIAL_947>",
+    "<SPECIAL_948>",
+    "<SPECIAL_949>",
+    "<SPECIAL_950>",
+    "<SPECIAL_951>",
+    "<SPECIAL_952>",
+    "<SPECIAL_953>",
+    "<SPECIAL_954>",
+    "<SPECIAL_955>",
+    "<SPECIAL_956>",
+    "<SPECIAL_957>",
+    "<SPECIAL_958>",
+    "<SPECIAL_959>",
+    "<SPECIAL_960>",
+    "<SPECIAL_961>",
+    "<SPECIAL_962>",
+    "<SPECIAL_963>",
+    "<SPECIAL_964>",
+    "<SPECIAL_965>",
+    "<SPECIAL_966>",
+    "<SPECIAL_967>",
+    "<SPECIAL_968>",
+    "<SPECIAL_969>",
+    "<SPECIAL_970>",
+    "<SPECIAL_971>",
+    "<SPECIAL_972>",
+    "<SPECIAL_973>",
+    "<SPECIAL_974>",
+    "<SPECIAL_975>",
+    "<SPECIAL_976>",
+    "<SPECIAL_977>",
+    "<SPECIAL_978>",
+    "<SPECIAL_979>",
+    "<SPECIAL_980>",
+    "<SPECIAL_981>",
+    "<SPECIAL_982>",
+    "<SPECIAL_983>",
+    "<SPECIAL_984>",
+    "<SPECIAL_985>",
+    "<SPECIAL_986>",
+    "<SPECIAL_987>",
+    "<SPECIAL_988>",
+    "<SPECIAL_989>",
+    "<SPECIAL_990>",
+    "<SPECIAL_991>",
+    "<SPECIAL_992>",
+    "<SPECIAL_993>",
+    "<SPECIAL_994>",
+    "<SPECIAL_995>",
+    "<SPECIAL_996>",
+    "<SPECIAL_997>",
+    "<SPECIAL_998>",
+    "<SPECIAL_999>"
+  ],
+  "is_local": true,
+  "local_files_only": false,
+  "model_max_length": 2048,
+  "pad_token": "<pad>",
+  "processor_class": "PixtralProcessor",
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": "<unk>"
+}

quantization_config.json

@@ -0,0 +1,30 @@
+{
+  "add_skip_keys": false,
+  "dequantize_fp32": false,
+  "dynamic_loss_threshold": null,
+  "group_size": 0,
+  "is_integer": true,
+  "is_training": false,
+  "modules_dtype_dict": {},
+  "modules_quant_config": {},
+  "modules_to_not_convert": [],
+  "modules_to_not_use_matmul": [],
+  "non_blocking": false,
+  "quant_conv": false,
+  "quant_embedding": false,
+  "quant_method": "sdnq",
+  "quantization_device": null,
+  "quantized_matmul_dtype": null,
+  "return_device": null,
+  "sdnq_version": "0.1.9",
+  "svd_rank": 32,
+  "svd_steps": 8,
+  "use_dynamic_quantization": false,
+  "use_grad_ckpt": true,
+  "use_quantized_matmul": false,
+  "use_quantized_matmul_conv": false,
+  "use_static_quantization": true,
+  "use_stochastic_rounding": false,
+  "use_svd": false,
+  "weights_dtype": "uint4"
+}

quantization_metrics.json

@@ -0,0 +1,22 @@
+{
+  "name": "quantize_and_save",
+  "seconds": 71.19859568402171,
+  "gpu_start_mib": 270,
+  "gpu_end_mib": 312,
+  "gpu_peak_mib": 1276,
+  "torch_peak_allocated_mib": 870,
+  "torch_peak_reserved_mib": 964,
+  "base_model": "baidu/ERNIE-Image-Turbo",
+  "output": "artifacts/ERNIE-Image-Turbo-SDNQ-uint4-static",
+  "output_size_bytes": 10048077414,
+  "recipe": {
+    "weights_dtype": "uint4",
+    "group_size": 0,
+    "use_svd": false,
+    "svd_rank": 32,
+    "svd_steps": 8,
+    "dynamic_loss_threshold": null,
+    "use_quantized_matmul": false,
+    "dequantize_fp32": false
+  }
+}

scheduler/scheduler_config.json

@@ -0,0 +1,18 @@
+{
+  "_class_name": "FlowMatchEulerDiscreteScheduler",
+  "_diffusers_version": "0.39.0.dev0",
+  "base_image_seq_len": 256,
+  "base_shift": 0.5,
+  "invert_sigmas": false,
+  "max_image_seq_len": 4096,
+  "max_shift": 1.15,
+  "num_train_timesteps": 1000,
+  "shift": 4.0,
+  "shift_terminal": null,
+  "stochastic_sampling": false,
+  "time_shift_type": "exponential",
+  "use_beta_sigmas": false,
+  "use_dynamic_shifting": false,
+  "use_exponential_sigmas": false,
+  "use_karras_sigmas": false
+}

text_encoder/config.json

@@ -0,0 +1,231 @@
+{
+  "architectures": [
+    "Mistral3Model"
+  ],
+  "dtype": "bfloat16",
+  "image_token_index": 10,
+  "model_type": "mistral3",
+  "multimodal_projector_bias": false,
+  "projector_hidden_act": "gelu",
+  "quantization_config": {
+    "add_skip_keys": false,
+    "dequantize_fp32": false,
+    "dynamic_loss_threshold": null,
+    "group_size": 0,
+    "is_integer": true,
+    "is_training": false,
+    "modules_dtype_dict": {},
+    "modules_quant_config": {},
+    "modules_to_not_convert": [
+      ".img_out",
+      ".proj_out",
+      ".emb_in",
+      ".final_layer",
+      "patch_embed",
+      "language_model.norm",
+      ".time_embed",
+      "multi_modal_projector",
+      ".condition_embedder",
+      ".t_embedder",
+      "wte",
+      "lm_head",
+      ".txt_out",
+      "time_text_embed",
+      ".context_embedder",
+      ".txt_in",
+      ".emb_out",
+      ".norm_out",
+      ".img_in",
+      ".vid_in",
+      ".x_embedder",
+      "patch_embedding",
+      "patch_emb",
+      ".vid_out",
+      ".y_embedder",
+      "language_model.embed_tokens.weight",
+      "language_model.layers.0.input_layernorm.weight",
+      "language_model.layers.0.post_attention_layernorm.weight",
+      "language_model.layers.1.input_layernorm.weight",
+      "language_model.layers.1.post_attention_layernorm.weight",
+      "language_model.layers.2.input_layernorm.weight",
+      "language_model.layers.2.post_attention_layernorm.weight",
+      "language_model.layers.3.input_layernorm.weight",
+      "language_model.layers.3.post_attention_layernorm.weight",
+      "language_model.layers.4.input_layernorm.weight",
+      "language_model.layers.4.post_attention_layernorm.weight",
+      "language_model.layers.5.input_layernorm.weight",
+      "language_model.layers.5.post_attention_layernorm.weight",
+      "language_model.layers.6.input_layernorm.weight",
+      "language_model.layers.6.post_attention_layernorm.weight",
+      "language_model.layers.7.input_layernorm.weight",
+      "language_model.layers.7.post_attention_layernorm.weight",
+      "language_model.layers.8.input_layernorm.weight",
+      "language_model.layers.8.post_attention_layernorm.weight",
+      "language_model.layers.9.input_layernorm.weight",
+      "language_model.layers.9.post_attention_layernorm.weight",
+      "language_model.layers.10.input_layernorm.weight",
+      "language_model.layers.10.post_attention_layernorm.weight",
+      "language_model.layers.11.input_layernorm.weight",
+      "language_model.layers.11.post_attention_layernorm.weight",
+      "language_model.layers.12.input_layernorm.weight",
+      "language_model.layers.12.post_attention_layernorm.weight",
+      "language_model.layers.13.input_layernorm.weight",
+      "language_model.layers.13.post_attention_layernorm.weight",
+      "language_model.layers.14.input_layernorm.weight",
+      "language_model.layers.14.post_attention_layernorm.weight",
+      "language_model.layers.15.input_layernorm.weight",
+      "language_model.layers.15.post_attention_layernorm.weight",
+      "language_model.layers.16.input_layernorm.weight",
+      "language_model.layers.16.post_attention_layernorm.weight",
+      "language_model.layers.17.input_layernorm.weight",
+      "language_model.layers.17.post_attention_layernorm.weight",
+      "language_model.layers.18.input_layernorm.weight",
+      "language_model.layers.18.post_attention_layernorm.weight",
+      "language_model.layers.19.input_layernorm.weight",
+      "language_model.layers.19.post_attention_layernorm.weight",
+      "language_model.layers.20.input_layernorm.weight",
+      "language_model.layers.20.post_attention_layernorm.weight",
+      "language_model.layers.21.input_layernorm.weight",
+      "language_model.layers.21.post_attention_layernorm.weight",
+      "language_model.layers.22.input_layernorm.weight",
+      "language_model.layers.22.post_attention_layernorm.weight",
+      "language_model.layers.23.input_layernorm.weight",
+      "language_model.layers.23.post_attention_layernorm.weight",
+      "language_model.layers.24.input_layernorm.weight",
+      "language_model.layers.24.post_attention_layernorm.weight",
+      "language_model.layers.25.input_layernorm.weight",
+      "language_model.layers.25.post_attention_layernorm.weight",
+      "language_model.norm.weight",
+      "multi_modal_projector.linear_1.weight",
+      "multi_modal_projector.linear_2.weight",
+      "multi_modal_projector.norm.weight",
+      "multi_modal_projector.patch_merger.merging_layer.weight",
+      "vision_tower.ln_pre.weight",
+      "vision_tower.patch_conv.weight",
+      "vision_tower.transformer.layers.0.attention_norm.weight",
+      "vision_tower.transformer.layers.0.ffn_norm.weight",
+      "vision_tower.transformer.layers.1.attention_norm.weight",
+      "vision_tower.transformer.layers.1.ffn_norm.weight",
+      "vision_tower.transformer.layers.2.attention_norm.weight",
+      "vision_tower.transformer.layers.2.ffn_norm.weight",
+      "vision_tower.transformer.layers.3.attention_norm.weight",
+      "vision_tower.transformer.layers.3.ffn_norm.weight",
+      "vision_tower.transformer.layers.4.attention_norm.weight",
+      "vision_tower.transformer.layers.4.ffn_norm.weight",
+      "vision_tower.transformer.layers.5.attention_norm.weight",
+      "vision_tower.transformer.layers.5.ffn_norm.weight",
+      "vision_tower.transformer.layers.6.attention_norm.weight",
+      "vision_tower.transformer.layers.6.ffn_norm.weight",
+      "vision_tower.transformer.layers.7.attention_norm.weight",
+      "vision_tower.transformer.layers.7.ffn_norm.weight",
+      "vision_tower.transformer.layers.8.attention_norm.weight",
+      "vision_tower.transformer.layers.8.ffn_norm.weight",
+      "vision_tower.transformer.layers.9.attention_norm.weight",
+      "vision_tower.transformer.layers.9.ffn_norm.weight",
+      "vision_tower.transformer.layers.10.attention_norm.weight",
+      "vision_tower.transformer.layers.10.ffn_norm.weight",
+      "vision_tower.transformer.layers.11.attention_norm.weight",
+      "vision_tower.transformer.layers.11.ffn_norm.weight",
+      "vision_tower.transformer.layers.12.attention_norm.weight",
+      "vision_tower.transformer.layers.12.ffn_norm.weight",
+      "vision_tower.transformer.layers.13.attention_norm.weight",
+      "vision_tower.transformer.layers.13.ffn_norm.weight",
+      "vision_tower.transformer.layers.14.attention_norm.weight",
+      "vision_tower.transformer.layers.14.ffn_norm.weight",
+      "vision_tower.transformer.layers.15.attention_norm.weight",
+      "vision_tower.transformer.layers.15.ffn_norm.weight",
+      "vision_tower.transformer.layers.16.attention_norm.weight",
+      "vision_tower.transformer.layers.16.ffn_norm.weight",
+      "vision_tower.transformer.layers.17.attention_norm.weight",
+      "vision_tower.transformer.layers.17.ffn_norm.weight",
+      "vision_tower.transformer.layers.18.attention_norm.weight",
+      "vision_tower.transformer.layers.18.ffn_norm.weight",
+      "vision_tower.transformer.layers.19.attention_norm.weight",
+      "vision_tower.transformer.layers.19.ffn_norm.weight",
+      "vision_tower.transformer.layers.20.attention_norm.weight",
+      "vision_tower.transformer.layers.20.ffn_norm.weight",
+      "vision_tower.transformer.layers.21.attention_norm.weight",
+      "vision_tower.transformer.layers.21.ffn_norm.weight",
+      "vision_tower.transformer.layers.22.attention_norm.weight",
+      "vision_tower.transformer.layers.22.ffn_norm.weight",
+      "vision_tower.transformer.layers.23.attention_norm.weight",
+      "vision_tower.transformer.layers.23.ffn_norm.weight"
+    ],
+    "modules_to_not_use_matmul": [],
+    "non_blocking": false,
+    "quant_conv": false,
+    "quant_embedding": false,
+    "quant_method": "sdnq",
+    "quantization_device": null,
+    "quantized_matmul_dtype": null,
+    "return_device": null,
+    "sdnq_version": "0.1.9",
+    "svd_rank": 32,
+    "svd_steps": 8,
+    "use_dynamic_quantization": false,
+    "use_grad_ckpt": true,
+    "use_quantized_matmul": false,
+    "use_quantized_matmul_conv": false,
+    "use_static_quantization": true,
+    "use_stochastic_rounding": false,
+    "use_svd": false,
+    "weights_dtype": "uint4"
+  },
+  "spatial_merge_size": 2,
+  "text_config": {
+    "attention_dropout": 0.0,
+    "bos_token_id": 1,
+    "dtype": "bfloat16",
+    "eos_token_id": 2,
+    "head_dim": 128,
+    "hidden_act": "silu",
+    "hidden_size": 3072,
+    "initializer_range": 0.02,
+    "intermediate_size": 9216,
+    "max_position_embeddings": 262144,
+    "model_type": "mistral",
+    "num_attention_heads": 32,
+    "num_hidden_layers": 26,
+    "num_key_value_heads": 8,
+    "pad_token_id": null,
+    "rms_norm_eps": 1e-05,
+    "rope_parameters": {
+      "beta_fast": 32.0,
+      "beta_slow": 1.0,
+      "factor": 16.0,
+      "llama_4_scaling_beta": 0.1,
+      "mscale": 1.0,
+      "mscale_all_dim": 1.0,
+      "original_max_position_embeddings": 16384,
+      "rope_theta": 1000000.0,
+      "rope_type": "yarn",
+      "type": "yarn"
+    },
+    "sliding_window": null,
+    "tie_word_embeddings": true,
+    "use_cache": true,
+    "vocab_size": 131072
+  },
+  "tie_word_embeddings": true,
+  "transformers_version": "5.8.0.dev0",
+  "vision_config": {
+    "attention_dropout": 0.0,
+    "dtype": "bfloat16",
+    "head_dim": 64,
+    "hidden_act": "silu",
+    "hidden_size": 1024,
+    "image_size": 1540,
+    "initializer_range": 0.02,
+    "intermediate_size": 4096,
+    "model_type": "pixtral",
+    "num_attention_heads": 16,
+    "num_channels": 3,
+    "num_hidden_layers": 24,
+    "patch_size": 14,
+    "rope_parameters": {
+      "rope_theta": 10000.0,
+      "rope_type": "default"
+    }
+  },
+  "vision_feature_layer": -1
+}

text_encoder/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4a102fd5903b8562219c41c5bf4d17181bae894d78f4bf9036a651c2067ba3c
+size 2769371768

text_encoder/quantization_config.json

@@ -0,0 +1,164 @@
+{
+  "add_skip_keys": false,
+  "dequantize_fp32": false,
+  "dynamic_loss_threshold": null,
+  "group_size": 0,
+  "is_integer": true,
+  "is_training": false,
+  "modules_dtype_dict": {},
+  "modules_quant_config": {},
+  "modules_to_not_convert": [
+    ".img_out",
+    ".proj_out",
+    ".emb_in",
+    ".final_layer",
+    "patch_embed",
+    "language_model.norm",
+    ".time_embed",
+    "multi_modal_projector",
+    ".condition_embedder",
+    ".t_embedder",
+    "wte",
+    "lm_head",
+    ".txt_out",
+    "time_text_embed",
+    ".context_embedder",
+    ".txt_in",
+    ".emb_out",
+    ".norm_out",
+    ".img_in",
+    ".vid_in",
+    ".x_embedder",
+    "patch_embedding",
+    "patch_emb",
+    ".vid_out",
+    ".y_embedder",
+    "language_model.embed_tokens.weight",
+    "language_model.layers.0.input_layernorm.weight",
+    "language_model.layers.0.post_attention_layernorm.weight",
+    "language_model.layers.1.input_layernorm.weight",
+    "language_model.layers.1.post_attention_layernorm.weight",
+    "language_model.layers.2.input_layernorm.weight",
+    "language_model.layers.2.post_attention_layernorm.weight",
+    "language_model.layers.3.input_layernorm.weight",
+    "language_model.layers.3.post_attention_layernorm.weight",
+    "language_model.layers.4.input_layernorm.weight",
+    "language_model.layers.4.post_attention_layernorm.weight",
+    "language_model.layers.5.input_layernorm.weight",
+    "language_model.layers.5.post_attention_layernorm.weight",
+    "language_model.layers.6.input_layernorm.weight",
+    "language_model.layers.6.post_attention_layernorm.weight",
+    "language_model.layers.7.input_layernorm.weight",
+    "language_model.layers.7.post_attention_layernorm.weight",
+    "language_model.layers.8.input_layernorm.weight",
+    "language_model.layers.8.post_attention_layernorm.weight",
+    "language_model.layers.9.input_layernorm.weight",
+    "language_model.layers.9.post_attention_layernorm.weight",
+    "language_model.layers.10.input_layernorm.weight",
+    "language_model.layers.10.post_attention_layernorm.weight",
+    "language_model.layers.11.input_layernorm.weight",
+    "language_model.layers.11.post_attention_layernorm.weight",
+    "language_model.layers.12.input_layernorm.weight",
+    "language_model.layers.12.post_attention_layernorm.weight",
+    "language_model.layers.13.input_layernorm.weight",
+    "language_model.layers.13.post_attention_layernorm.weight",
+    "language_model.layers.14.input_layernorm.weight",
+    "language_model.layers.14.post_attention_layernorm.weight",
+    "language_model.layers.15.input_layernorm.weight",
+    "language_model.layers.15.post_attention_layernorm.weight",
+    "language_model.layers.16.input_layernorm.weight",
+    "language_model.layers.16.post_attention_layernorm.weight",
+    "language_model.layers.17.input_layernorm.weight",
+    "language_model.layers.17.post_attention_layernorm.weight",
+    "language_model.layers.18.input_layernorm.weight",
+    "language_model.layers.18.post_attention_layernorm.weight",
+    "language_model.layers.19.input_layernorm.weight",
+    "language_model.layers.19.post_attention_layernorm.weight",
+    "language_model.layers.20.input_layernorm.weight",
+    "language_model.layers.20.post_attention_layernorm.weight",
+    "language_model.layers.21.input_layernorm.weight",
+    "language_model.layers.21.post_attention_layernorm.weight",
+    "language_model.layers.22.input_layernorm.weight",
+    "language_model.layers.22.post_attention_layernorm.weight",
+    "language_model.layers.23.input_layernorm.weight",
+    "language_model.layers.23.post_attention_layernorm.weight",
+    "language_model.layers.24.input_layernorm.weight",
+    "language_model.layers.24.post_attention_layernorm.weight",
+    "language_model.layers.25.input_layernorm.weight",
+    "language_model.layers.25.post_attention_layernorm.weight",
+    "language_model.norm.weight",
+    "multi_modal_projector.linear_1.weight",
+    "multi_modal_projector.linear_2.weight",
+    "multi_modal_projector.norm.weight",
+    "multi_modal_projector.patch_merger.merging_layer.weight",
+    "vision_tower.ln_pre.weight",
+    "vision_tower.patch_conv.weight",
+    "vision_tower.transformer.layers.0.attention_norm.weight",
+    "vision_tower.transformer.layers.0.ffn_norm.weight",
+    "vision_tower.transformer.layers.1.attention_norm.weight",
+    "vision_tower.transformer.layers.1.ffn_norm.weight",
+    "vision_tower.transformer.layers.2.attention_norm.weight",
+    "vision_tower.transformer.layers.2.ffn_norm.weight",
+    "vision_tower.transformer.layers.3.attention_norm.weight",
+    "vision_tower.transformer.layers.3.ffn_norm.weight",
+    "vision_tower.transformer.layers.4.attention_norm.weight",
+    "vision_tower.transformer.layers.4.ffn_norm.weight",
+    "vision_tower.transformer.layers.5.attention_norm.weight",
+    "vision_tower.transformer.layers.5.ffn_norm.weight",
+    "vision_tower.transformer.layers.6.attention_norm.weight",
+    "vision_tower.transformer.layers.6.ffn_norm.weight",
+    "vision_tower.transformer.layers.7.attention_norm.weight",
+    "vision_tower.transformer.layers.7.ffn_norm.weight",
+    "vision_tower.transformer.layers.8.attention_norm.weight",
+    "vision_tower.transformer.layers.8.ffn_norm.weight",
+    "vision_tower.transformer.layers.9.attention_norm.weight",
+    "vision_tower.transformer.layers.9.ffn_norm.weight",
+    "vision_tower.transformer.layers.10.attention_norm.weight",
+    "vision_tower.transformer.layers.10.ffn_norm.weight",
+    "vision_tower.transformer.layers.11.attention_norm.weight",
+    "vision_tower.transformer.layers.11.ffn_norm.weight",
+    "vision_tower.transformer.layers.12.attention_norm.weight",
+    "vision_tower.transformer.layers.12.ffn_norm.weight",
+    "vision_tower.transformer.layers.13.attention_norm.weight",
+    "vision_tower.transformer.layers.13.ffn_norm.weight",
+    "vision_tower.transformer.layers.14.attention_norm.weight",
+    "vision_tower.transformer.layers.14.ffn_norm.weight",
+    "vision_tower.transformer.layers.15.attention_norm.weight",
+    "vision_tower.transformer.layers.15.ffn_norm.weight",
+    "vision_tower.transformer.layers.16.attention_norm.weight",
+    "vision_tower.transformer.layers.16.ffn_norm.weight",
+    "vision_tower.transformer.layers.17.attention_norm.weight",
+    "vision_tower.transformer.layers.17.ffn_norm.weight",
+    "vision_tower.transformer.layers.18.attention_norm.weight",
+    "vision_tower.transformer.layers.18.ffn_norm.weight",
+    "vision_tower.transformer.layers.19.attention_norm.weight",
+    "vision_tower.transformer.layers.19.ffn_norm.weight",
+    "vision_tower.transformer.layers.20.attention_norm.weight",
+    "vision_tower.transformer.layers.20.ffn_norm.weight",
+    "vision_tower.transformer.layers.21.attention_norm.weight",
+    "vision_tower.transformer.layers.21.ffn_norm.weight",
+    "vision_tower.transformer.layers.22.attention_norm.weight",
+    "vision_tower.transformer.layers.22.ffn_norm.weight",
+    "vision_tower.transformer.layers.23.attention_norm.weight",
+    "vision_tower.transformer.layers.23.ffn_norm.weight"
+  ],
+  "modules_to_not_use_matmul": [],
+  "non_blocking": false,
+  "quant_conv": false,
+  "quant_embedding": false,
+  "quant_method": "sdnq",
+  "quantization_device": null,
+  "quantized_matmul_dtype": null,
+  "return_device": null,
+  "sdnq_version": "0.1.9",
+  "svd_rank": 32,
+  "svd_steps": 8,
+  "use_dynamic_quantization": false,
+  "use_grad_ckpt": true,
+  "use_quantized_matmul": false,
+  "use_quantized_matmul_conv": false,
+  "use_static_quantization": true,
+  "use_stochastic_rounding": false,
+  "use_svd": false,
+  "weights_dtype": "uint4"
+}

tokenizer/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:577575622324b2e099e2648be26bdeb5e5815ffe66d7004e9e3ddbf421db6bf1
+size 17078110

tokenizer/tokenizer_config.json

@@ -0,0 +1,16 @@
+{
+  "add_prefix_space": null,
+  "backend": "tokenizers",
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "is_local": true,
+  "legacy": true,
+  "local_files_only": false,
+  "model_max_length": 2048,
+  "pad_token": "<pad>",
+  "processor_class": "PixtralProcessor",
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

transformer/config.json

@@ -0,0 +1,225 @@
+{
+  "_class_name": "ErnieImageTransformer2DModel",
+  "_diffusers_version": "0.39.0.dev0",
+  "_name_or_path": "/root/ernie-sdnq-lab/hf-cache/hub/models--baidu--ERNIE-Image-Turbo/snapshots/bc68c81e2a1730a394d5fc9fae70713dee940140/transformer",
+  "eps": 1e-06,
+  "ffn_hidden_size": 12288,
+  "hidden_size": 4096,
+  "in_channels": 128,
+  "lora_rank": 4,
+  "num_attention_heads": 32,
+  "num_layers": 36,
+  "out_channels": 128,
+  "patch_size": 1,
+  "qk_layernorm": true,
+  "quantization_config": {
+    "add_skip_keys": true,
+    "dequantize_fp32": false,
+    "dynamic_loss_threshold": null,
+    "group_size": 0,
+    "is_integer": true,
+    "is_training": false,
+    "modules_dtype_dict": {},
+    "modules_quant_config": {},
+    "modules_to_not_convert": [
+      ".img_out",
+      ".proj_out",
+      ".emb_in",
+      ".final_layer",
+      "patch_embed",
+      ".time_embed",
+      "multi_modal_projector",
+      ".condition_embedder",
+      ".t_embedder",
+      "wte",
+      "lm_head",
+      ".txt_out",
+      "time_text_embed",
+      ".context_embedder",
+      ".txt_in",
+      ".emb_out",
+      ".norm_out",
+      ".img_in",
+      ".vid_in",
+      ".x_embedder",
+      "patch_embedding",
+      "patch_emb",
+      ".vid_out",
+      ".y_embedder",
+      "layers.0.adaLN_mlp_ln.weight",
+      "layers.0.adaLN_sa_ln.weight",
+      "layers.0.self_attention.norm_k.weight",
+      "layers.0.self_attention.norm_q.weight",
+      "layers.1.adaLN_mlp_ln.weight",
+      "layers.1.adaLN_sa_ln.weight",
+      "layers.1.self_attention.norm_k.weight",
+      "layers.1.self_attention.norm_q.weight",
+      "layers.10.adaLN_mlp_ln.weight",
+      "layers.10.adaLN_sa_ln.weight",
+      "layers.10.self_attention.norm_k.weight",
+      "layers.10.self_attention.norm_q.weight",
+      "layers.11.adaLN_mlp_ln.weight",
+      "layers.11.adaLN_sa_ln.weight",
+      "layers.11.self_attention.norm_k.weight",
+      "layers.11.self_attention.norm_q.weight",
+      "layers.12.adaLN_mlp_ln.weight",
+      "layers.12.adaLN_sa_ln.weight",
+      "layers.12.self_attention.norm_k.weight",
+      "layers.12.self_attention.norm_q.weight",
+      "layers.13.adaLN_mlp_ln.weight",
+      "layers.13.adaLN_sa_ln.weight",
+      "layers.13.self_attention.norm_k.weight",
+      "layers.13.self_attention.norm_q.weight",
+      "layers.14.adaLN_mlp_ln.weight",
+      "layers.14.adaLN_sa_ln.weight",
+      "layers.14.self_attention.norm_k.weight",
+      "layers.14.self_attention.norm_q.weight",
+      "layers.15.adaLN_mlp_ln.weight",
+      "layers.15.adaLN_sa_ln.weight",
+      "layers.15.self_attention.norm_k.weight",
+      "layers.15.self_attention.norm_q.weight",
+      "layers.16.adaLN_mlp_ln.weight",
+      "layers.16.adaLN_sa_ln.weight",
+      "layers.16.self_attention.norm_k.weight",
+      "layers.16.self_attention.norm_q.weight",
+      "layers.17.adaLN_mlp_ln.weight",
+      "layers.17.adaLN_sa_ln.weight",
+      "layers.17.self_attention.norm_k.weight",
+      "layers.17.self_attention.norm_q.weight",
+      "layers.18.adaLN_mlp_ln.weight",
+      "layers.18.adaLN_sa_ln.weight",
+      "layers.18.self_attention.norm_k.weight",
+      "layers.18.self_attention.norm_q.weight",
+      "layers.19.adaLN_mlp_ln.weight",
+      "layers.19.adaLN_sa_ln.weight",
+      "layers.19.self_attention.norm_k.weight",
+      "layers.19.self_attention.norm_q.weight",
+      "layers.2.adaLN_mlp_ln.weight",
+      "layers.2.adaLN_sa_ln.weight",
+      "layers.2.self_attention.norm_k.weight",
+      "layers.2.self_attention.norm_q.weight",
+      "layers.20.adaLN_mlp_ln.weight",
+      "layers.20.adaLN_sa_ln.weight",
+      "layers.20.self_attention.norm_k.weight",
+      "layers.20.self_attention.norm_q.weight",
+      "layers.21.adaLN_mlp_ln.weight",
+      "layers.21.adaLN_sa_ln.weight",
+      "layers.21.self_attention.norm_k.weight",
+      "layers.21.self_attention.norm_q.weight",
+      "layers.22.adaLN_sa_ln.weight",
+      "layers.3.adaLN_mlp_ln.weight",
+      "layers.3.adaLN_sa_ln.weight",
+      "layers.3.self_attention.norm_k.weight",
+      "layers.3.self_attention.norm_q.weight",
+      "layers.4.adaLN_mlp_ln.weight",
+      "layers.4.adaLN_sa_ln.weight",
+      "layers.4.self_attention.norm_k.weight",
+      "layers.4.self_attention.norm_q.weight",
+      "layers.5.adaLN_mlp_ln.weight",
+      "layers.5.adaLN_sa_ln.weight",
+      "layers.5.self_attention.norm_k.weight",
+      "layers.5.self_attention.norm_q.weight",
+      "layers.6.adaLN_mlp_ln.weight",
+      "layers.6.adaLN_sa_ln.weight",
+      "layers.6.self_attention.norm_k.weight",
+      "layers.6.self_attention.norm_q.weight",
+      "layers.7.adaLN_mlp_ln.weight",
+      "layers.7.adaLN_sa_ln.weight",
+      "layers.7.self_attention.norm_k.weight",
+      "layers.7.self_attention.norm_q.weight",
+      "layers.8.adaLN_mlp_ln.weight",
+      "layers.8.adaLN_sa_ln.weight",
+      "layers.8.self_attention.norm_k.weight",
+      "layers.8.self_attention.norm_q.weight",
+      "layers.9.adaLN_mlp_ln.weight",
+      "layers.9.adaLN_sa_ln.weight",
+      "layers.9.self_attention.norm_k.weight",
+      "layers.9.self_attention.norm_q.weight",
+      "time_embedding.linear_1.weight",
+      "time_embedding.linear_2.weight",
+      "x_embedder.proj.weight",
+      "layers.22.adaLN_mlp_ln.weight",
+      "layers.22.self_attention.norm_k.weight",
+      "layers.22.self_attention.norm_q.weight",
+      "layers.23.adaLN_mlp_ln.weight",
+      "layers.23.adaLN_sa_ln.weight",
+      "layers.23.self_attention.norm_k.weight",
+      "layers.23.self_attention.norm_q.weight",
+      "layers.24.adaLN_mlp_ln.weight",
+      "layers.24.adaLN_sa_ln.weight",
+      "layers.24.self_attention.norm_k.weight",
+      "layers.24.self_attention.norm_q.weight",
+      "layers.25.adaLN_mlp_ln.weight",
+      "layers.25.adaLN_sa_ln.weight",
+      "layers.25.self_attention.norm_k.weight",
+      "layers.25.self_attention.norm_q.weight",
+      "layers.26.adaLN_mlp_ln.weight",
+      "layers.26.adaLN_sa_ln.weight",
+      "layers.26.self_attention.norm_k.weight",
+      "layers.26.self_attention.norm_q.weight",
+      "layers.27.adaLN_mlp_ln.weight",
+      "layers.27.adaLN_sa_ln.weight",
+      "layers.27.self_attention.norm_k.weight",
+      "layers.27.self_attention.norm_q.weight",
+      "layers.28.adaLN_mlp_ln.weight",
+      "layers.28.adaLN_sa_ln.weight",
+      "layers.28.self_attention.norm_k.weight",
+      "layers.28.self_attention.norm_q.weight",
+      "layers.29.adaLN_mlp_ln.weight",
+      "layers.29.adaLN_sa_ln.weight",
+      "layers.29.self_attention.norm_k.weight",
+      "layers.29.self_attention.norm_q.weight",
+      "layers.30.adaLN_mlp_ln.weight",
+      "layers.30.adaLN_sa_ln.weight",
+      "layers.30.self_attention.norm_k.weight",
+      "layers.30.self_attention.norm_q.weight",
+      "layers.31.adaLN_mlp_ln.weight",
+      "layers.31.adaLN_sa_ln.weight",
+      "layers.31.self_attention.norm_k.weight",
+      "layers.31.self_attention.norm_q.weight",
+      "layers.32.adaLN_mlp_ln.weight",
+      "layers.32.adaLN_sa_ln.weight",
+      "layers.32.self_attention.norm_k.weight",
+      "layers.32.self_attention.norm_q.weight",
+      "layers.33.adaLN_mlp_ln.weight",
+      "layers.33.adaLN_sa_ln.weight",
+      "layers.33.self_attention.norm_k.weight",
+      "layers.33.self_attention.norm_q.weight",
+      "layers.34.adaLN_mlp_ln.weight",
+      "layers.34.adaLN_sa_ln.weight",
+      "layers.34.self_attention.norm_k.weight",
+      "layers.34.self_attention.norm_q.weight",
+      "layers.35.adaLN_mlp_ln.weight",
+      "layers.35.adaLN_sa_ln.weight",
+      "layers.35.self_attention.norm_k.weight",
+      "layers.35.self_attention.norm_q.weight"
+    ],
+    "modules_to_not_use_matmul": [],
+    "non_blocking": false,
+    "quant_conv": false,
+    "quant_embedding": false,
+    "quant_method": "sdnq",
+    "quantization_device": "cuda",
+    "quantized_matmul_dtype": null,
+    "return_device": "cpu",
+    "sdnq_version": "0.1.9",
+    "svd_rank": 32,
+    "svd_steps": 8,
+    "use_dynamic_quantization": false,
+    "use_grad_ckpt": true,
+    "use_quantized_matmul": false,
+    "use_quantized_matmul_conv": false,
+    "use_static_quantization": true,
+    "use_stochastic_rounding": false,
+    "use_svd": false,
+    "weights_dtype": "uint4"
+  },
+  "rope_axes_dim": [
+    32,
+    48,
+    48
+  ],
+  "rope_theta": 256,
+  "text_in_dim": 3072,
+  "use_lora": false
+}

transformer/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d7eb7f834ab5ad523886ad021c117b4e71b69649308f280477f48dbb9b0bf48
+size 4568429440

transformer/quantization_config.json

@@ -0,0 +1,202 @@
+{
+  "add_skip_keys": false,
+  "dequantize_fp32": false,
+  "dynamic_loss_threshold": null,
+  "group_size": 0,
+  "is_integer": true,
+  "is_training": false,
+  "modules_dtype_dict": {},
+  "modules_quant_config": {},
+  "modules_to_not_convert": [
+    ".img_out",
+    ".proj_out",
+    ".emb_in",
+    ".final_layer",
+    "patch_embed",
+    ".time_embed",
+    "multi_modal_projector",
+    ".condition_embedder",
+    ".t_embedder",
+    "wte",
+    "lm_head",
+    ".txt_out",
+    "time_text_embed",
+    ".context_embedder",
+    ".txt_in",
+    ".emb_out",
+    ".norm_out",
+    ".img_in",
+    ".vid_in",
+    ".x_embedder",
+    "patch_embedding",
+    "patch_emb",
+    ".vid_out",
+    ".y_embedder",
+    "layers.0.adaLN_mlp_ln.weight",
+    "layers.0.adaLN_sa_ln.weight",
+    "layers.0.self_attention.norm_k.weight",
+    "layers.0.self_attention.norm_q.weight",
+    "layers.1.adaLN_mlp_ln.weight",
+    "layers.1.adaLN_sa_ln.weight",
+    "layers.1.self_attention.norm_k.weight",
+    "layers.1.self_attention.norm_q.weight",
+    "layers.10.adaLN_mlp_ln.weight",
+    "layers.10.adaLN_sa_ln.weight",
+    "layers.10.self_attention.norm_k.weight",
+    "layers.10.self_attention.norm_q.weight",
+    "layers.11.adaLN_mlp_ln.weight",
+    "layers.11.adaLN_sa_ln.weight",
+    "layers.11.self_attention.norm_k.weight",
+    "layers.11.self_attention.norm_q.weight",
+    "layers.12.adaLN_mlp_ln.weight",
+    "layers.12.adaLN_sa_ln.weight",
+    "layers.12.self_attention.norm_k.weight",
+    "layers.12.self_attention.norm_q.weight",
+    "layers.13.adaLN_mlp_ln.weight",
+    "layers.13.adaLN_sa_ln.weight",
+    "layers.13.self_attention.norm_k.weight",
+    "layers.13.self_attention.norm_q.weight",
+    "layers.14.adaLN_mlp_ln.weight",
+    "layers.14.adaLN_sa_ln.weight",
+    "layers.14.self_attention.norm_k.weight",
+    "layers.14.self_attention.norm_q.weight",
+    "layers.15.adaLN_mlp_ln.weight",
+    "layers.15.adaLN_sa_ln.weight",
+    "layers.15.self_attention.norm_k.weight",
+    "layers.15.self_attention.norm_q.weight",
+    "layers.16.adaLN_mlp_ln.weight",
+    "layers.16.adaLN_sa_ln.weight",
+    "layers.16.self_attention.norm_k.weight",
+    "layers.16.self_attention.norm_q.weight",
+    "layers.17.adaLN_mlp_ln.weight",
+    "layers.17.adaLN_sa_ln.weight",
+    "layers.17.self_attention.norm_k.weight",
+    "layers.17.self_attention.norm_q.weight",
+    "layers.18.adaLN_mlp_ln.weight",
+    "layers.18.adaLN_sa_ln.weight",
+    "layers.18.self_attention.norm_k.weight",
+    "layers.18.self_attention.norm_q.weight",
+    "layers.19.adaLN_mlp_ln.weight",
+    "layers.19.adaLN_sa_ln.weight",
+    "layers.19.self_attention.norm_k.weight",
+    "layers.19.self_attention.norm_q.weight",
+    "layers.2.adaLN_mlp_ln.weight",
+    "layers.2.adaLN_sa_ln.weight",
+    "layers.2.self_attention.norm_k.weight",
+    "layers.2.self_attention.norm_q.weight",
+    "layers.20.adaLN_mlp_ln.weight",
+    "layers.20.adaLN_sa_ln.weight",
+    "layers.20.self_attention.norm_k.weight",
+    "layers.20.self_attention.norm_q.weight",
+    "layers.21.adaLN_mlp_ln.weight",
+    "layers.21.adaLN_sa_ln.weight",
+    "layers.21.self_attention.norm_k.weight",
+    "layers.21.self_attention.norm_q.weight",
+    "layers.22.adaLN_sa_ln.weight",
+    "layers.3.adaLN_mlp_ln.weight",
+    "layers.3.adaLN_sa_ln.weight",
+    "layers.3.self_attention.norm_k.weight",
+    "layers.3.self_attention.norm_q.weight",
+    "layers.4.adaLN_mlp_ln.weight",
+    "layers.4.adaLN_sa_ln.weight",
+    "layers.4.self_attention.norm_k.weight",
+    "layers.4.self_attention.norm_q.weight",
+    "layers.5.adaLN_mlp_ln.weight",
+    "layers.5.adaLN_sa_ln.weight",
+    "layers.5.self_attention.norm_k.weight",
+    "layers.5.self_attention.norm_q.weight",
+    "layers.6.adaLN_mlp_ln.weight",
+    "layers.6.adaLN_sa_ln.weight",
+    "layers.6.self_attention.norm_k.weight",
+    "layers.6.self_attention.norm_q.weight",
+    "layers.7.adaLN_mlp_ln.weight",
+    "layers.7.adaLN_sa_ln.weight",
+    "layers.7.self_attention.norm_k.weight",
+    "layers.7.self_attention.norm_q.weight",
+    "layers.8.adaLN_mlp_ln.weight",
+    "layers.8.adaLN_sa_ln.weight",
+    "layers.8.self_attention.norm_k.weight",
+    "layers.8.self_attention.norm_q.weight",
+    "layers.9.adaLN_mlp_ln.weight",
+    "layers.9.adaLN_sa_ln.weight",
+    "layers.9.self_attention.norm_k.weight",
+    "layers.9.self_attention.norm_q.weight",
+    "time_embedding.linear_1.weight",
+    "time_embedding.linear_2.weight",
+    "x_embedder.proj.weight",
+    "layers.22.adaLN_mlp_ln.weight",
+    "layers.22.self_attention.norm_k.weight",
+    "layers.22.self_attention.norm_q.weight",
+    "layers.23.adaLN_mlp_ln.weight",
+    "layers.23.adaLN_sa_ln.weight",
+    "layers.23.self_attention.norm_k.weight",
+    "layers.23.self_attention.norm_q.weight",
+    "layers.24.adaLN_mlp_ln.weight",
+    "layers.24.adaLN_sa_ln.weight",
+    "layers.24.self_attention.norm_k.weight",
+    "layers.24.self_attention.norm_q.weight",
+    "layers.25.adaLN_mlp_ln.weight",
+    "layers.25.adaLN_sa_ln.weight",
+    "layers.25.self_attention.norm_k.weight",
+    "layers.25.self_attention.norm_q.weight",
+    "layers.26.adaLN_mlp_ln.weight",
+    "layers.26.adaLN_sa_ln.weight",
+    "layers.26.self_attention.norm_k.weight",
+    "layers.26.self_attention.norm_q.weight",
+    "layers.27.adaLN_mlp_ln.weight",
+    "layers.27.adaLN_sa_ln.weight",
+    "layers.27.self_attention.norm_k.weight",
+    "layers.27.self_attention.norm_q.weight",
+    "layers.28.adaLN_mlp_ln.weight",
+    "layers.28.adaLN_sa_ln.weight",
+    "layers.28.self_attention.norm_k.weight",
+    "layers.28.self_attention.norm_q.weight",
+    "layers.29.adaLN_mlp_ln.weight",
+    "layers.29.adaLN_sa_ln.weight",
+    "layers.29.self_attention.norm_k.weight",
+    "layers.29.self_attention.norm_q.weight",
+    "layers.30.adaLN_mlp_ln.weight",
+    "layers.30.adaLN_sa_ln.weight",
+    "layers.30.self_attention.norm_k.weight",
+    "layers.30.self_attention.norm_q.weight",
+    "layers.31.adaLN_mlp_ln.weight",
+    "layers.31.adaLN_sa_ln.weight",
+    "layers.31.self_attention.norm_k.weight",
+    "layers.31.self_attention.norm_q.weight",
+    "layers.32.adaLN_mlp_ln.weight",
+    "layers.32.adaLN_sa_ln.weight",
+    "layers.32.self_attention.norm_k.weight",
+    "layers.32.self_attention.norm_q.weight",
+    "layers.33.adaLN_mlp_ln.weight",
+    "layers.33.adaLN_sa_ln.weight",
+    "layers.33.self_attention.norm_k.weight",
+    "layers.33.self_attention.norm_q.weight",
+    "layers.34.adaLN_mlp_ln.weight",
+    "layers.34.adaLN_sa_ln.weight",
+    "layers.34.self_attention.norm_k.weight",
+    "layers.34.self_attention.norm_q.weight",
+    "layers.35.adaLN_mlp_ln.weight",
+    "layers.35.adaLN_sa_ln.weight",
+    "layers.35.self_attention.norm_k.weight",
+    "layers.35.self_attention.norm_q.weight"
+  ],
+  "modules_to_not_use_matmul": [],
+  "non_blocking": false,
+  "quant_conv": false,
+  "quant_embedding": false,
+  "quant_method": "sdnq",
+  "quantization_device": null,
+  "quantized_matmul_dtype": null,
+  "return_device": null,
+  "sdnq_version": "0.1.9",
+  "svd_rank": 32,
+  "svd_steps": 8,
+  "use_dynamic_quantization": false,
+  "use_grad_ckpt": true,
+  "use_quantized_matmul": false,
+  "use_quantized_matmul_conv": false,
+  "use_static_quantization": true,
+  "use_stochastic_rounding": false,
+  "use_svd": false,
+  "weights_dtype": "uint4"
+}

vae/config.json

@@ -0,0 +1,41 @@
+{
+  "_class_name": "AutoencoderKLFlux2",
+  "_diffusers_version": "0.39.0.dev0",
+  "_name_or_path": "/root/ernie-sdnq-lab/hf-cache/hub/models--baidu--ERNIE-Image-Turbo/snapshots/bc68c81e2a1730a394d5fc9fae70713dee940140/vae",
+  "act_fn": "silu",
+  "batch_norm_eps": 0.0001,
+  "batch_norm_momentum": 0.1,
+  "block_out_channels": [
+    128,
+    256,
+    512,
+    512
+  ],
+  "decoder_block_out_channels": null,
+  "down_block_types": [
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D"
+  ],
+  "force_upcast": true,
+  "in_channels": 3,
+  "latent_channels": 32,
+  "layers_per_block": 2,
+  "mid_block_add_attention": true,
+  "norm_num_groups": 32,
+  "out_channels": 3,
+  "patch_size": [
+    2,
+    2
+  ],
+  "sample_size": 1024,
+  "up_block_types": [
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D"
+  ],
+  "use_post_quant_conv": true,
+  "use_quant_conv": true
+}

vae/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca70d2202afe6415bdbcb8793ba8cd99fd159cfe6192381504d6c4d3036e0f04
+size 168120878