Add model from convert_d_fine_original_pytorch_checkpoint_to_hf.py

config.json

@@ -21,29 +21,6 @@
       2,
       3,
       4
-    ],
-    "stage_in_channels": [
-      64,
-      128,
-      512,
-      1024
-    ],
-    "stage_mid_channels": [
-      64,
-      128,
-      256,
-      512
-    ],
-    "stage_num_blocks": [
-      1,
-      2,
-      5,
-      2
-    ],
-    "stem_channels": [
-      3,
-      32,
-      64
     ]
   },
   "backbone_kwargs": null,
@@ -54,17 +31,13 @@
   "decoder_attention_heads": 8,
   "decoder_ffn_dim": 1024,
   "decoder_in_channels": [
-    384,
-    384,
-    384
+    256,
+    256,
+    256
   ],
   "decoder_layers": 6,
   "decoder_method": "default",
-  "decoder_n_points": [
-    3,
-    6,
-    3
-  ],
+  "decoder_n_points": 4,
   "decoder_offset_scale": 0.5,
   "depth_mult": 1.0,
   "disable_custom_kernels": true,
@@ -74,8 +47,8 @@
   ],
   "encoder_activation_function": "gelu",
   "encoder_attention_heads": 8,
-  "encoder_ffn_dim": 2048,
-  "encoder_hidden_dim": 384,
+  "encoder_ffn_dim": 1024,
+  "encoder_hidden_dim": 256,
   "encoder_in_channels": [
     512,
     1024,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3ac6ea37c6c995ed65a10beeda703249e97aac1bb40bda18cfaa1c2f6cdca4c5
-size 251619400
+oid sha256:4e9d77ba937ab6eb8537b183c1c05372262daaa552e7682d721417042e24a8dc
+size 125618504