Upload folder using huggingface_hub
Browse files
.gitattributes
CHANGED
|
@@ -1 +1,2 @@
|
|
| 1 |
ckpt_new/denoiser_step0350000.pt filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 1 |
ckpt_new/denoiser_step0350000.pt filter=lfs diff=lfs merge=lfs -text
|
| 2 |
+
decoder/ckpt_new/denoiser_step0350000.pt filter=lfs diff=lfs merge=lfs -text
|
decoder/ckpt_new/denoiser_step0350000.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"name": "ControlledSparseStructureFlowModel",
|
| 3 |
+
"args": {
|
| 4 |
+
"resolution": 16,
|
| 5 |
+
"in_channels": 8,
|
| 6 |
+
"out_channels": 8,
|
| 7 |
+
"model_channels": 1024,
|
| 8 |
+
"cond_channels": 1024,
|
| 9 |
+
"num_blocks": 24,
|
| 10 |
+
"num_heads": 16,
|
| 11 |
+
"mlp_ratio": 4,
|
| 12 |
+
"patch_size": 1,
|
| 13 |
+
"pe_mode": "ape",
|
| 14 |
+
"qk_rms_norm": true,
|
| 15 |
+
"use_fp16": true
|
| 16 |
+
}
|
| 17 |
+
}
|
decoder/ckpt_new/denoiser_step0350000.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5031661edcf12d754556184644813a1cd327ae6200f99a7e21375070dbb1d78c
|
| 3 |
+
size 3470728767
|
decoder/pipeline.json
ADDED
|
@@ -0,0 +1,63 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"name": "TrellisImageTo3DPipeline",
|
| 3 |
+
"args": {
|
| 4 |
+
"models": {
|
| 5 |
+
"sparse_structure_decoder": "../trellis/ckpts/ss_dec_conv3d_16l8_fp16",
|
| 6 |
+
"sparse_structure_encoder": "../trellis/ckpts/ss_enc_conv3d_16l8_fp16",
|
| 7 |
+
|
| 8 |
+
"sparse_structure_flow_model": "./ckpt_new/denoiser_step0350000.pt",
|
| 9 |
+
|
| 10 |
+
"slat_decoder_gs": "../trellis/ckpts/slat_dec_gs_swin8_B_64l8gs32_fp16",
|
| 11 |
+
"slat_decoder_rf": "../trellis/ckpts/slat_dec_rf_swin8_B_64l8r16_fp16",
|
| 12 |
+
"slat_decoder_mesh": "../trellis/ckpts/slat_dec_mesh_swin8_B_64l8m256c_fp16",
|
| 13 |
+
"slat_flow_model": "../trellis/ckpts/slat_flow_img_dit_L_64l8p2_fp16"
|
| 14 |
+
},
|
| 15 |
+
"sparse_structure_sampler": {
|
| 16 |
+
"name": "FlowEulerGuidanceIntervalSampler_Control",
|
| 17 |
+
"args": {
|
| 18 |
+
"sigma_min": 1e-5
|
| 19 |
+
},
|
| 20 |
+
"params": {
|
| 21 |
+
"steps": 25,
|
| 22 |
+
"cfg_strength": 5.0,
|
| 23 |
+
"cfg_interval": [0.5, 1.0],
|
| 24 |
+
"rescale_t": 3.0
|
| 25 |
+
}
|
| 26 |
+
},
|
| 27 |
+
"slat_sampler": {
|
| 28 |
+
"name": "FlowEulerGuidanceIntervalSampler",
|
| 29 |
+
"args": {
|
| 30 |
+
"sigma_min": 1e-5
|
| 31 |
+
},
|
| 32 |
+
"params": {
|
| 33 |
+
"steps": 25,
|
| 34 |
+
"cfg_strength": 5.0,
|
| 35 |
+
"cfg_interval": [0.5, 1.0],
|
| 36 |
+
"rescale_t": 3.0
|
| 37 |
+
}
|
| 38 |
+
},
|
| 39 |
+
"slat_normalization": {
|
| 40 |
+
"mean": [
|
| 41 |
+
-2.1687545776367188,
|
| 42 |
+
-0.004347046371549368,
|
| 43 |
+
-0.13352349400520325,
|
| 44 |
+
-0.08418072760105133,
|
| 45 |
+
-0.5271206498146057,
|
| 46 |
+
0.7238689064979553,
|
| 47 |
+
-1.1414450407028198,
|
| 48 |
+
1.2039363384246826
|
| 49 |
+
],
|
| 50 |
+
"std": [
|
| 51 |
+
2.377650737762451,
|
| 52 |
+
2.386378288269043,
|
| 53 |
+
2.124418020248413,
|
| 54 |
+
2.1748552322387695,
|
| 55 |
+
2.663944721221924,
|
| 56 |
+
2.371192216873169,
|
| 57 |
+
2.6217446327209473,
|
| 58 |
+
2.684523105621338
|
| 59 |
+
]
|
| 60 |
+
},
|
| 61 |
+
"image_cond_model": "dinov2_vitl14_reg"
|
| 62 |
+
}
|
| 63 |
+
}
|