fxtdstudios commited on
Commit
018c46d
·
verified ·
1 Parent(s): 53b0fbb

RUDRA ACES 2 and AI fill v3 (Apache 2.0)

Browse files
README.md CHANGED
@@ -1,7 +1,6 @@
1
  ---
2
  license: apache-2.0
3
  library_name: rudra
4
- inference: false
5
  pipeline_tag: image-to-image
6
  base_model:
7
  - Wan-AI/Wan2.1-VACE-1.3B-diffusers
@@ -23,12 +22,12 @@ commercial use permitted.**
23
 
24
  | Folder | What it is | Size |
25
  |---|---|---|
26
- | [`models/RUDRA_ACES_2/`](models/RUDRA_ACES_2) | **RUDRA ACES 2**: the learned ACES 2.0 reconstruction. 8-bit SDR in, an ACEScg master out, whose ACES 2.0 SDR view keeps the source's 8-bit codes. Ships as TorchScript and ONNX (DirectML, Core ML or CPU). | 2.2 MB |
27
  | [`models/rudra_genfill_v3/`](models/rudra_genfill_v3) | **RUDRA AI fill v3**: a LoRA for Wan2.1-VACE-1.3B that draws the *shape* inside fully clipped highlights, and a small brightness estimator (`brightness.json`) that sets *how bright* they are. | 271 MB |
28
 
29
  ## Use
30
 
31
- **In the RUDRA app** (Windows and macOS):
32
 
33
  - **RUDRA ACES 2:** File › Model packages, or Look › Download the model. Every
34
  file is checked against its SHA-256 before it is used.
@@ -36,8 +35,8 @@ commercial use permitted.**
36
  this LoRA and `brightness.json`. It also needs:
37
  - the base model [Wan2.1-VACE-1.3B-diffusers](https://huggingface.co/Wan-AI/Wan2.1-VACE-1.3B-diffusers)
38
  (set its folder in Deliver › AI fill › Setup…, or put it in `<RUDRA folder>/models/genfill/`);
39
- - Python (`scripts\GENFILL.ps1 -Setup` on Windows, `scripts/genfill_setup.sh` on a Mac);
40
- - a 16 GB GPU (on a Mac, 16 GB of memory or more: it runs on Metal).
41
 
42
  Then use Deliver › AI fill.
43
 
@@ -52,6 +51,21 @@ python -m rudra.genfill.run --master <EXR folder> --out <new folder> \
52
  `run.py` loads `brightness.json` from the LoRA's folder by default. Add
53
  `--brightness-offset <stops>` to shift the brightness (±3).
54
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
55
  ## What the AI fill guarantees
56
 
57
  - **Codes kept:** no 8-bit code of the master's ACES 2.0 SDR 100-nit view
 
1
  ---
2
  license: apache-2.0
3
  library_name: rudra
 
4
  pipeline_tag: image-to-image
5
  base_model:
6
  - Wan-AI/Wan2.1-VACE-1.3B-diffusers
 
22
 
23
  | Folder | What it is | Size |
24
  |---|---|---|
25
+ | [`models/RUDRA_ACES_2/`](models/RUDRA_ACES_2) | **RUDRA ACES 2**: the learned ACES 2.0 reconstruction. 8-bit SDR in, an ACEScg master out, whose ACES 2.0 SDR view keeps the source's 8-bit codes. Ships as TorchScript and ONNX (DirectML/CPU). | 2.2 MB |
26
  | [`models/rudra_genfill_v3/`](models/rudra_genfill_v3) | **RUDRA AI fill v3**: a LoRA for Wan2.1-VACE-1.3B that draws the *shape* inside fully clipped highlights, and a small brightness estimator (`brightness.json`) that sets *how bright* they are. | 271 MB |
27
 
28
  ## Use
29
 
30
+ **In the RUDRA app** (Windows; macOS builds are unsigned for now):
31
 
32
  - **RUDRA ACES 2:** File › Model packages, or Look › Download the model. Every
33
  file is checked against its SHA-256 before it is used.
 
35
  this LoRA and `brightness.json`. It also needs:
36
  - the base model [Wan2.1-VACE-1.3B-diffusers](https://huggingface.co/Wan-AI/Wan2.1-VACE-1.3B-diffusers)
37
  (set its folder in Deliver › AI fill › Setup…, or put it in `<RUDRA folder>/models/genfill/`);
38
+ - Python (`scripts\GENFILL.ps1 -Setup`);
39
+ - a 16 GB GPU.
40
 
41
  Then use Deliver › AI fill.
42
 
 
51
  `run.py` loads `brightness.json` from the LoRA's folder by default. Add
52
  `--brightness-offset <stops>` to shift the brightness (±3).
53
 
54
+ ## RUDRA ACES 2: what it guarantees, and its evidence
55
+
56
+ - **Codes kept:** the master's ACES 2.0 SDR 100-nit view gives back the
57
+ source's 8-bit codes within 1 code, fully clipped pixels included.
58
+ - **2.0 (w4 + lock):** in partially clipped highlights the channels the camera
59
+ recorded are held to the source (the lock), and only the clipped channel is
60
+ rebuilt. Needs RUDRA 1.2.0 or later; older apps refuse the package.
61
+ - **Held-out test** (The Crew, ARRI ALEXA LogC3, 56 cases from 18 scenes,
62
+ registered before it was run, run once): better than the plain ACES 2.0
63
+ inverse at −3 EV and +3 EV and in highlight colour; at the delivered
64
+ exposure **within 0.1 code RMS of it, not better** (0.311 against 0.242).
65
+ One production and one camera: a result on that footage, not on all footage.
66
+ - **Rebuilt highlights are estimates**, marked in the `rudra.reconstructed`
67
+ EXR channel. Not camera RAW.
68
+
69
  ## What the AI fill guarantees
70
 
71
  - **Codes kept:** no 8-bit code of the master's ACES 2.0 SDR 100-nit view
models/RUDRA_ACES_2/manifest.json CHANGED
@@ -1,108 +1,162 @@
1
- {
2
- "name": "RUDRA_ACES_2",
3
- "title": "RUDRA ACES 2",
4
- "source_package": "aces_v2_partial_pcgrad_duration6400_retry1_20261025",
5
- "contract": "rudra_native_aces_v2_staged",
6
- "contract_major": 2,
7
- "model_contract": {
8
- "id": "aces2_ap1_reconstruction_v9_full_context_experimental",
9
- "config": "cg-config-v4.0.0_aces-v2.0_ocio-v2.5",
10
- "display": "Rec.1886 Rec.709 - Display",
11
- "view": "ACES 2.0 - SDR 100 nits (Rec.709)",
12
- "working_space": "ACEScg",
13
- "ocio_version": "2.5.2",
14
- "config_sha256": "24ec81841048fc5db160a7bad882263246183385c5d49d0e86e11464917ead15",
15
- "input": "Rec709 Rec1886 display RGB codes",
16
- "output": "ACEScg linear",
17
- "source_code_bound": 0.0019215686274509803,
18
- "reconstruction_policy": "observed_clip_v1",
19
- "residual": "signed_linear_shared_rgb_scale",
20
- "residual_scale_floor": 0.001,
21
- "radiance_clamp": false,
22
- "scene_exposure_recovered": false,
23
- "release_authorized": false,
24
- "normalization": "pixel_rms",
25
- "restoration": "local_codes",
26
- "restoration_support": "source reconstruction_weight == 0",
27
- "restoration_context_kernels": [
28
- 5,
29
- 17,
30
- 65
31
- ],
32
- "shadow": "analytic",
33
- "shadow_support": "source reconstruction_weight > 0 and max source <= .5",
34
- "shadow_output": "w*exact source inverse + (1-w)*exact inverse of bounded restored source",
35
- "shadow_reconstructed": false,
36
- "shadow_reason": "5-seed study 24 Oct 2026: no trained shadow branch reliably beat the analytic inverse",
37
- "highlight_context": "independent_pixel_rms",
38
- "highlight_support": "source reconstruction_weight>0 and max source>.5",
39
- "frozen_parent_fields": true,
40
- "temporal_inference": false,
41
- "temporal_training": "known subpixel planar-motion residual consistency; no recurrent inference",
42
- "full_context": "independent_pixel_rms",
43
- "full_support": "reconstruction_weight>0 and min source>=254.5/255",
44
- "full_transition": "hard source support; boundary and temporal qualification required"
45
- },
46
- "checkpoint_sha256": "f73f14c7e74d5bb54d0cbf949b94365f2eeb12a3cc50ed3ea45d1ad1b163256e",
47
- "g1_sha256": "fb30e3df748f4fa9ae17f7db189c8fe49a58d9863af6c1c45a408a72f3b7a59b",
48
- "g1_file": "rudra_g1.ts",
49
- "g1_inputs": [
50
- "source BCHW float32 display codes [0,1] Rec709 Rec1886",
51
- "baseline BCHW float32 ACEScg = OCIO inverse(source)"
52
- ],
53
- "g1_outputs": [
54
- "restored_source B3HW",
55
- "highlight_fields B3HW",
56
- "full_fields B3HW",
57
- "weight B1HW"
58
- ],
59
- "ocio": {
60
- "ocio_version": "2.5.2",
61
- "config": "cg-config-v4.0.0_aces-v2.0_ocio-v2.5",
62
- "view": "ACES 2.0 - SDR 100 nits (Rec.709)",
63
- "display": "Rec.1886 Rec.709 - Display"
64
- },
65
- "stages": [
66
- "inverse_view(source)->baseline",
67
- "G1",
68
- "inverse_view(restored_source)->restored",
69
- "compose",
70
- "clipped_smoothing",
71
- "appearance_constraint(include_fully_clipped)",
72
- "mask"
73
- ],
74
- "compose": {
75
- "scale": "max_c|restored|+0.001",
76
- "high": "w>0 & max(src)>0.5",
77
- "full": "w>0 & min(src)>=254.5/255",
78
- "shadow": "w>0 & max(src)<=0.5 -> w>=1 ? baseline : w*baseline+(1-w)*restored",
79
- "else": "restored"
80
- },
81
- "clipped_smoothing": {
82
- "name": "hybrid64",
83
- "sigma": 64.0,
84
- "ramp": 1.0,
85
- "start": 0.0,
86
- "mode": "hybrid",
87
- "erode": "cross 3x3 x2, border 0",
88
- "components": 8,
89
- "gaussian": "k=int(4*sigma)|1, BORDER_REFLECT_101"
90
- },
91
- "appearance_constraint": {
92
- "max_display_error_codes": 1,
93
- "include_fully_clipped": true,
94
- "method": "halving backtrack toward baseline",
95
- "forbid_new_negative_y": true
96
- },
97
- "onnx_file": "rudra_g1.onnx",
98
- "onnx_sha256": "8ba97cb0ed36f5033958e752afad95d585d50bcefefc3bc66a61b27a6618f237",
99
- "onnx_tolerance": {
100
- "final_max_display_code_diff": 0.10628391802310944,
101
- "final_max_rel": 0.0007460270426236093,
102
- "mask_exact": true,
103
- "limit_display_codes": 0.25,
104
- "limit_rel": 0.005
105
- },
106
- "parity_pass": true,
107
- "release_authorized": true
108
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "RUDRA_ACES_2_w4",
3
+ "title": "RUDRA ACES 2",
4
+ "contract": "rudra_native_aces_v2_staged",
5
+ "contract_major": 3,
6
+ "model_contract": {
7
+ "id": "aces2_ap1_reconstruction_v9_full_context_experimental",
8
+ "config": "cg-config-v4.0.0_aces-v2.0_ocio-v2.5",
9
+ "display": "Rec.1886 Rec.709 - Display",
10
+ "view": "ACES 2.0 - SDR 100 nits (Rec.709)",
11
+ "working_space": "ACEScg",
12
+ "ocio_version": "2.5.2",
13
+ "config_sha256": "24ec81841048fc5db160a7bad882263246183385c5d49d0e86e11464917ead15",
14
+ "input": "Rec709 Rec1886 display RGB codes",
15
+ "output": "ACEScg linear",
16
+ "source_code_bound": 0.0019215686274509803,
17
+ "reconstruction_policy": "observed_clip_v1",
18
+ "residual": "signed_linear_shared_rgb_scale",
19
+ "residual_scale_floor": 0.001,
20
+ "radiance_clamp": false,
21
+ "scene_exposure_recovered": false,
22
+ "release_authorized": false,
23
+ "normalization": "pixel_rms",
24
+ "restoration": "local_codes",
25
+ "restoration_support": "source reconstruction_weight == 0",
26
+ "restoration_context_kernels": [
27
+ 5,
28
+ 17,
29
+ 65
30
+ ],
31
+ "shadow": "analytic",
32
+ "shadow_support": "source reconstruction_weight > 0 and max source <= .5",
33
+ "shadow_output": "w*exact source inverse + (1-w)*exact inverse of bounded restored source",
34
+ "shadow_reconstructed": false,
35
+ "shadow_reason": "5-seed study 24 Oct 2026: no trained shadow branch reliably beat the analytic inverse",
36
+ "highlight_context": "independent_pixel_rms",
37
+ "highlight_support": "source reconstruction_weight>0 and max source>.5",
38
+ "frozen_parent_fields": true,
39
+ "temporal_inference": false,
40
+ "temporal_training": "known subpixel planar-motion residual consistency; no recurrent inference",
41
+ "full_context": "independent_pixel_rms",
42
+ "full_support": "reconstruction_weight>0 and min source>=254.5/255",
43
+ "full_transition": "hard source support; boundary and temporal qualification required"
44
+ },
45
+ "checkpoint_sha256": "b9a903a58dccfcd2e1d2aa15b7d01d33cf0858dacb09cbe5910df05998f2825f",
46
+ "g1_sha256": "7082eb869d265d6e8c4fc209f7cac7e5dce2f40c10df5e73a889fb85efb9dbcc",
47
+ "g1_file": "rudra_g1.ts",
48
+ "g1_inputs": [
49
+ "source BCHW float32 display codes [0,1] Rec709 Rec1886",
50
+ "baseline BCHW float32 ACEScg = OCIO inverse(source)"
51
+ ],
52
+ "g1_outputs": [
53
+ "restored_source B3HW",
54
+ "highlight_fields B3HW",
55
+ "full_fields B3HW",
56
+ "weight B1HW"
57
+ ],
58
+ "ocio": {
59
+ "ocio_version": "2.5.2",
60
+ "config": "cg-config-v4.0.0_aces-v2.0_ocio-v2.5",
61
+ "view": "ACES 2.0 - SDR 100 nits (Rec.709)",
62
+ "display": "Rec.1886 Rec.709 - Display"
63
+ },
64
+ "stages": [
65
+ "inverse_view(source)->baseline",
66
+ "G1",
67
+ "inverse_view(restored_source)->restored",
68
+ "compose",
69
+ "clipped_smoothing",
70
+ "appearance_constraint(include_fully_clipped)",
71
+ "lock",
72
+ "mask"
73
+ ],
74
+ "compose": {
75
+ "scale": "max_c|restored|+0.001",
76
+ "high": "w>0 & max(src)>0.5",
77
+ "full": "w>0 & min(src)>=254.5/255",
78
+ "shadow": "w>0 & max(src)<=0.5 -> w>=1 ? baseline : w*baseline+(1-w)*restored",
79
+ "else": "restored"
80
+ },
81
+ "clipped_smoothing": {
82
+ "name": "hybrid64",
83
+ "sigma": 64.0,
84
+ "ramp": 1.0,
85
+ "start": 0.0,
86
+ "mode": "hybrid",
87
+ "erode": "cross 3x3 x2, border 0",
88
+ "components": 8,
89
+ "gaussian": "k=int(4*sigma)|1, BORDER_REFLECT_101"
90
+ },
91
+ "appearance_constraint": {
92
+ "max_display_error_codes": 1,
93
+ "include_fully_clipped": true,
94
+ "method": "halving backtrack toward baseline",
95
+ "forbid_new_negative_y": true
96
+ },
97
+ "onnx_file": "rudra_g1.onnx",
98
+ "onnx_sha256": "880e7b698a099b8dd60549444392d51af3022efbae8f68f05c9f0f3a29f8ee49",
99
+ "onnx_tolerance": {
100
+ "final_max_display_code_diff": 0.10628391872160137,
101
+ "final_max_rel": 0.018843110650777817,
102
+ "mask_exact": true,
103
+ "limit_display_codes": 0.25,
104
+ "limit_rel": 0.005,
105
+ "rel_fraction_limit": 0.001,
106
+ "pixels_rel_over_limit": 1,
107
+ "per_probe_rel_over_limit": [
108
+ 0,
109
+ 0,
110
+ 1
111
+ ],
112
+ "locked": {
113
+ "p999_display_code_diff": 0.014085752964021859,
114
+ "max_display_code_diff": 0.7523596286773682,
115
+ "max_rel": 0.3374467194080353,
116
+ "pixels_over_limit": 3,
117
+ "pixels": 1855488,
118
+ "limit_p999_codes": 0.25,
119
+ "limit_max_codes": 1.0,
120
+ "pixels_rel_over_limit": 26,
121
+ "pass": true
122
+ },
123
+ "locked_per_probe_rel_over_limit": [
124
+ 21,
125
+ 0,
126
+ 5
127
+ ]
128
+ },
129
+ "lock": {
130
+ "name": "lock_dqs050g",
131
+ "mode": "lock",
132
+ "strength": 0.5,
133
+ "target": "dequant",
134
+ "iters": 4,
135
+ "band_codes": 0.0,
136
+ "dq_iters": 200,
137
+ "dq_max_lap": 1.5,
138
+ "registration": "configs/experiments/holdout_crew_eval_20261107.json",
139
+ "registration_sha256": "a7271ad62db39ec0af97085bfd39cabfdf5d379cd93f65d0fed36551c69ac106"
140
+ },
141
+ "provenance": {
142
+ "model": "RUDRA ACES 2",
143
+ "version": "2.0",
144
+ "checkpoint": {
145
+ "run": "research_bin_hinge_w4_20261102",
146
+ "step": 1600,
147
+ "sha256": "b9a903a58dccfcd2e1d2aa15b7d01d33cf0858dacb09cbe5910df05998f2825f"
148
+ },
149
+ "lock": "lock_dqs050g",
150
+ "holdout": {
151
+ "registration": "configs\\experiments\\holdout_crew_eval_20261107.json",
152
+ "registration_sha256": "a7271ad62db39ec0af97085bfd39cabfdf5d379cd93f65d0fed36551c69ac106",
153
+ "verdict_sha256": "c6710884e187a2367050b3339e41cf6393fea7cdea85e5f7e00a334bc8cf7034",
154
+ "result": "docs/competitive/RESULT_CREW_HOLDOUT_2026_11_07.md",
155
+ "cases": 56,
156
+ "scenes": 18
157
+ },
158
+ "limits": "One production (The Crew), one camera (ARRI ALEXA LogC3), single frames."
159
+ },
160
+ "parity_pass": true,
161
+ "release_authorized": false
162
+ }
models/RUDRA_ACES_2/parity.json CHANGED
@@ -35,8 +35,8 @@
35
  }
36
  ],
37
  "onnx": {
38
- "final_max_display_code_diff": 0.10628391802310944,
39
- "final_max_rel": 0.0007460270426236093,
40
  "mask_exact": true,
41
  "limit_display_codes": 0.25,
42
  "limit_rel": 0.005,
@@ -48,8 +48,12 @@
48
  128,
49
  3
50
  ],
51
- "max_display_code_diff": 0.10628391802310944,
52
- "max_rel": 0.0003397462423890829,
 
 
 
 
53
  "mask_exact": true
54
  },
55
  {
@@ -58,8 +62,12 @@
58
  1280,
59
  3
60
  ],
61
- "max_display_code_diff": 0.0005167722702026367,
 
 
 
62
  "max_rel": 0.00022069606347940862,
 
63
  "mask_exact": true
64
  },
65
  {
@@ -68,11 +76,72 @@
68
  1280,
69
  3
70
  ],
71
- "max_display_code_diff": 0.0005927681922912598,
72
- "max_rel": 0.0007460270426236093,
 
 
 
 
73
  "mask_exact": true
74
  }
75
  ],
76
- "pass": true
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
77
  }
78
  }
 
35
  }
36
  ],
37
  "onnx": {
38
+ "final_max_display_code_diff": 0.10628391872160137,
39
+ "final_max_rel": 0.018843110650777817,
40
  "mask_exact": true,
41
  "limit_display_codes": 0.25,
42
  "limit_rel": 0.005,
 
48
  128,
49
  3
50
  ],
51
+ "max_display_code_diff": 0.10628391872160137,
52
+ "p999_display_code_diff": 0.0014061886444687933,
53
+ "pixels_over_limit": 0,
54
+ "pixels": 12288,
55
+ "max_rel": 6.0725302319042385e-05,
56
+ "pixels_rel_over_limit": 0,
57
  "mask_exact": true
58
  },
59
  {
 
62
  1280,
63
  3
64
  ],
65
+ "max_display_code_diff": 0.0015503168106079102,
66
+ "p999_display_code_diff": 0.00021278858184814453,
67
+ "pixels_over_limit": 0,
68
+ "pixels": 921600,
69
  "max_rel": 0.00022069606347940862,
70
+ "pixels_rel_over_limit": 0,
71
  "mask_exact": true
72
  },
73
  {
 
76
  1280,
77
  3
78
  ],
79
+ "max_display_code_diff": 0.0016415119171142578,
80
+ "p999_display_code_diff": 0.00025838613510131836,
81
+ "pixels_over_limit": 0,
82
+ "pixels": 921600,
83
+ "max_rel": 0.018843110650777817,
84
+ "pixels_rel_over_limit": 1,
85
  "mask_exact": true
86
  }
87
  ],
88
+ "rel_fraction_limit": 0.001,
89
+ "pixels_rel_over_limit": 1,
90
+ "pass": true,
91
+ "locked": {
92
+ "p999_display_code_diff": 0.014085752964021859,
93
+ "max_display_code_diff": 0.7523596286773682,
94
+ "max_rel": 0.3374467194080353,
95
+ "pixels_over_limit": 3,
96
+ "pixels": 1855488,
97
+ "limit_p999_codes": 0.25,
98
+ "limit_max_codes": 1.0,
99
+ "probes": [
100
+ {
101
+ "shape": [
102
+ 96,
103
+ 128,
104
+ 3
105
+ ],
106
+ "max_display_code_diff": 0.7523596286773682,
107
+ "p999_display_code_diff": 0.014085752964021859,
108
+ "pixels_over_limit": 3,
109
+ "pixels": 12288,
110
+ "max_rel": 0.3374467194080353,
111
+ "pixels_rel_over_limit": 21,
112
+ "mask_exact": true
113
+ },
114
+ {
115
+ "shape": [
116
+ 720,
117
+ 1280,
118
+ 3
119
+ ],
120
+ "max_display_code_diff": 0.0015503168106079102,
121
+ "p999_display_code_diff": 0.00021278858184814453,
122
+ "pixels_over_limit": 0,
123
+ "pixels": 921600,
124
+ "max_rel": 0.0007740037399344146,
125
+ "pixels_rel_over_limit": 0,
126
+ "mask_exact": true
127
+ },
128
+ {
129
+ "shape": [
130
+ 720,
131
+ 1280,
132
+ 3
133
+ ],
134
+ "max_display_code_diff": 0.013907253742218018,
135
+ "p999_display_code_diff": 0.00025838613510131836,
136
+ "pixels_over_limit": 0,
137
+ "pixels": 921600,
138
+ "max_rel": 0.018843110650777817,
139
+ "pixels_rel_over_limit": 5,
140
+ "mask_exact": true
141
+ }
142
+ ],
143
+ "pixels_rel_over_limit": 26,
144
+ "pass": true
145
+ }
146
  }
147
  }
models/RUDRA_ACES_2/rudra_g1.onnx CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8ba97cb0ed36f5033958e752afad95d585d50bcefefc3bc66a61b27a6618f237
3
  size 1157294
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:880e7b698a099b8dd60549444392d51af3022efbae8f68f05c9f0f3a29f8ee49
3
  size 1157294
models/RUDRA_ACES_2/rudra_g1.ts CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:fb30e3df748f4fa9ae17f7db189c8fe49a58d9863af6c1c45a408a72f3b7a59b
3
  size 1037772
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7082eb869d265d6e8c4fc209f7cac7e5dce2f40c10df5e73a889fb85efb9dbcc
3
  size 1037772
models/rudra_genfill_v3/prompt_cache/19805c5814225ef6a3c96bad.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"text": "clear sky with soft clouds over the sea, distant coastline", "tokens": 14}
models/rudra_genfill_v3/prompt_cache/19805c5814225ef6a3c96bad.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a25186b90f16ad0ae397e72bd748a8150ebbfbcedc107057605e64245d36f243
3
+ size 8390432