V3 - More samples - Burnice_White

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "/home/ralvarez22/Adrian/Text2Latex/checkpoints/detr/Tayzzyronth/V_9",
+  "_name_or_path": "./finetuned/rtdetr/Burnice_White__RTDETR/V_6",
   "activation_dropout": 0.0,
   "activation_function": "silu",
   "anchor_image_size": null,
@@ -11,18 +11,11 @@
   "backbone": null,
   "backbone_config": {
     "depths": [
-      2,
-      2,
-      2,
-      2
-    ],
-    "hidden_sizes": [
-      64,
-      128,
-      256,
-      512
+      3,
+      4,
+      23,
+      3
     ],
-    "layer_type": "basic",
     "model_type": "rt_detr_resnet",
     "out_features": [
       "stage2",
@@ -43,11 +36,11 @@
   "decoder_attention_heads": 8,
   "decoder_ffn_dim": 1024,
   "decoder_in_channels": [
-    256,
-    256,
-    256
+    384,
+    384,
+    384
   ],
-  "decoder_layers": 3,
+  "decoder_layers": 6,
   "decoder_n_points": 4,
   "disable_custom_kernels": true,
   "dropout": 0.0,
@@ -56,12 +49,12 @@
   ],
   "encoder_activation_function": "gelu",
   "encoder_attention_heads": 8,
-  "encoder_ffn_dim": 1024,
-  "encoder_hidden_dim": 256,
+  "encoder_ffn_dim": 2048,
+  "encoder_hidden_dim": 384,
   "encoder_in_channels": [
-    128,
-    256,
-    512
+    512,
+    1024,
+    2048
   ],
   "encoder_layers": 1,
   "eos_coefficient": 0.0001,
@@ -74,21 +67,21 @@
   "focal_loss_alpha": 0.75,
   "focal_loss_gamma": 2.0,
   "freeze_backbone_batch_norms": true,
-  "hidden_expansion": 0.5,
+  "hidden_expansion": 1.0,
   "id2label": {
-    "0": "word",
-    "1": "text",
-    "2": "latex",
-    "3": "image"
+    "0": "LABEL_0",
+    "1": "LABEL_1",
+    "2": "LABEL_2",
+    "3": "LABEL_3"
   },
   "initializer_bias_prior_prob": null,
   "initializer_range": 0.01,
   "is_encoder_decoder": true,
   "label2id": {
-    "image": 3,
-    "latex": 2,
-    "text": 1,
-    "word": 0
+    "LABEL_0": 0,
+    "LABEL_1": 1,
+    "LABEL_2": 2,
+    "LABEL_3": 3
   },
   "label_noise_ratio": 0.5,
   "layer_norm_eps": 1e-05,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:035a80308d86babf53235a4fd7c57b2d16687a25b07f598ec48eb2db390493cb
-size 80513800
+oid sha256:34fd453268ba06400bb137d95f698f643a857e1c39521eb10090c5cb9fbec5cb
+size 306706264