Upload model

config.json

@@ -1,6 +1,7 @@
 {
+  "_name_or_path": "/Users/vik/datalab/experiments/texify_train",
   "architectures": [
-    "OCREncoderDecoderModel"
+    "TexifyModel"
   ],
   "decoder": {
     "_attn_implementation_autoset": true,
@@ -36,6 +37,7 @@
     "decoder_start_token_id": null,
     "diversity_penalty": 0.0,
     "do_sample": false,
+    "double_residual_flow": false,
     "dropout": 0.1,
     "early_stopping": false,
     "encoder_cross_attn_layers": [
@@ -91,7 +93,7 @@
     "lru_width": 512,
     "max_length": 20,
     "min_length": 0,
-    "model_type": "recurrent_gemma",
+    "model_type": "texify",
     "no_repeat_ngram_size": 0,
     "num_attention_heads": 8,
     "num_beam_groups": 1,
@@ -145,7 +147,7 @@
   "decoder_end_token_id": 1,
   "decoder_start_token_id": 1,
   "encoder": {
-    "_attn_implementation_autoset": false,
+    "_attn_implementation_autoset": true,
     "_name_or_path": "",
     "add_cross_attention": false,
     "architectures": null,
@@ -244,6 +246,7 @@
     "typical_p": 1.0,
     "use_absolute_embeddings": false,
     "use_bfloat16": false,
+    "use_positional_embeddings": true,
     "window_size": 8
   },
   "eos_token_id": 1,
@@ -251,6 +254,6 @@
   "model_type": "vision-encoder-decoder",
   "pad_token_id": 0,
   "tie_word_embeddings": false,
-  "torch_dtype": "float32",
-  "transformers_version": "4.46.3"
+  "torch_dtype": "float16",
+  "transformers_version": "4.48.1"
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e9a9bc6f22db88352b2efd8bffd580555a9f37b520eec5901d3bb6eb9435632e
-size 683901280
+oid sha256:3ee5c3efcd79246f20fccc5292ce95bd5e4c0e26e6ef7abc8e6543f1a159e1cf
+size 338587064