new download format for elated-lake-212 model with data stored separately for faster retreival

config.json

@@ -1,61 +1,64 @@
 {
-  "_name_or_path": "elated-lake-212",
-  "architectures": [
-    "T5ForConditionalGeneration"
-  ],
-  "classifier_dropout": 0.0,
-  "d_ff": 1024,
-  "d_kv": 64,
-  "d_model": 512,
-  "decoder_start_token_id": 0,
-  "dense_act_fn": "gelu_new",
-  "dropout_rate": 0.1,
-  "eos_token_id": 1,
-  "feed_forward_proj": "gated-gelu",
-  "initializer_factor": 1.0,
-  "is_encoder_decoder": true,
-  "is_gated_act": true,
-  "layer_norm_epsilon": 1e-06,
-  "model_type": "t5",
-  "n_positions": 512,
-  "num_decoder_layers": 2,
-  "num_heads": 6,
-  "num_layers": 8,
-  "output_past": true,
-  "pad_token_id": 0,
-  "relative_attention_max_distance": 128,
-  "relative_attention_num_buckets": 32,
-  "task_specific_params": {
-    "summarization": {
-      "early_stopping": true,
-      "length_penalty": 2.0,
-      "max_length": 200,
-      "min_length": 30,
-      "no_repeat_ngram_size": 3,
-      "num_beams": 4,
-      "prefix": "summarize: "
+    "_name_or_path": "elated-lake-212",
+    "architectures": [
+        "T5ForConditionalGeneration"
+    ],
+    "classifier_dropout": 0.0,
+    "d_ff": 1024,
+    "d_kv": 64,
+    "d_model": 512,
+    "decoder_start_token_id": 0,
+    "dense_act_fn": "gelu_new",
+    "dropout_rate": 0.1,
+    "eos_token_id": 1,
+    "feed_forward_proj": "gated-gelu",
+    "initializer_factor": 1.0,
+    "is_encoder_decoder": true,
+    "is_gated_act": true,
+    "layer_norm_epsilon": 1e-06,
+    "model_type": "t5",
+    "n_positions": 512,
+    "num_decoder_layers": 2,
+    "num_heads": 6,
+    "num_layers": 8,
+    "output_past": true,
+    "pad_token_id": 0,
+    "relative_attention_max_distance": 128,
+    "relative_attention_num_buckets": 32,
+    "task_specific_params": {
+        "summarization": {
+            "early_stopping": true,
+            "length_penalty": 2.0,
+            "max_length": 200,
+            "min_length": 30,
+            "no_repeat_ngram_size": 3,
+            "num_beams": 4,
+            "prefix": "summarize: "
+        },
+        "translation_en_to_de": {
+            "early_stopping": true,
+            "max_length": 300,
+            "num_beams": 4,
+            "prefix": "translate English to German: "
+        },
+        "translation_en_to_fr": {
+            "early_stopping": true,
+            "max_length": 300,
+            "num_beams": 4,
+            "prefix": "translate English to French: "
+        },
+        "translation_en_to_ro": {
+            "early_stopping": true,
+            "max_length": 300,
+            "num_beams": 4,
+            "prefix": "translate English to Romanian: "
+        }
     },
-    "translation_en_to_de": {
-      "early_stopping": true,
-      "max_length": 300,
-      "num_beams": 4,
-      "prefix": "translate English to German: "
-    },
-    "translation_en_to_fr": {
-      "early_stopping": true,
-      "max_length": 300,
-      "num_beams": 4,
-      "prefix": "translate English to French: "
-    },
-    "translation_en_to_ro": {
-      "early_stopping": true,
-      "max_length": 300,
-      "num_beams": 4,
-      "prefix": "translate English to Romanian: "
+    "tie_word_embeddings": false,
+    "transformers_version": "4.43.4",
+    "use_cache": true,
+    "vocab_size": 32128,
+    "transformers.js_config": {
+        "use_external_data_format": true
     }
-  },
-  "tie_word_embeddings": false,
-  "transformers_version": "4.43.4",
-  "use_cache": true,
-  "vocab_size": 32128
-}
+}

onnx/decoder_model.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:967b8ca37a9000bb8744cfd9dc75fca9001c065b8e7352ecb4beb060f2904672
-size 156855183
+oid sha256:0f0917a6356c30c16edf61db3d65ec651bdbc69fa0894f18938780c5154b7890
+size 80917

onnx/decoder_model.onnx_data

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd57c2b50bba0aa2fd1f32d52e08de864914c46c0c43773bbdc83e0aa806cdaf
+size 156776448

onnx/decoder_model_merged.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e24d87e71c31ef05d50966c673f30ec9da9a6917801a8a30f888c4b97d99f80d
-size 156935090
+oid sha256:f6b332727c9e4650f644a0bc898a68e35a06931ac30c12d1554d89ddcc800bc3
+size 161054

onnx/decoder_model_merged.onnx_data

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca17984db2d10fb738d692c3f00bac476706fe0dee0020c6115778dcbad26a34
+size 156776448

onnx/decoder_model_merged_quantized.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ae58988133a4bf53e74072058b72e75a607270ffec805a7578e4ad7e575f6f59
-size 39634616
+oid sha256:047291bc4dc783282486336a9a954991be953cbfc45bbcb817f130224d2ead45
+size 432464

onnx/decoder_model_merged_quantized.onnx_data

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:17edaa480f281be856a68bdf15b0298dc539d04361b90660e82cc3fc1409b71d
+size 39204864

onnx/decoder_model_quantized.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5626af1398c1a45f4f72297a714309fe0830c257697cb394339a57f93bf73bb4
-size 39530972
+oid sha256:a3d6730dabdd14036c4d0e4b4933f6a9451137ec8f511973ce3e58b6ab66fef2
+size 328569

onnx/decoder_model_quantized.onnx_data

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:950848da25276500c41d62ee8b517adcb31b45bc25a5cf525acce94dcb005842
+size 39204864

onnx/decoder_with_past_model.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ad67949c7009da990658a301802fad935a60adae4acf43ea2c196dc7fb6f0643
-size 153699835
+oid sha256:0018eea69567573479b6f2101bb442b4101393cc5bdb7130252560d1bb567b24
+size 71294

onnx/decoder_with_past_model.onnx_data

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:908cde32ddb49278f9236be1a9cb6421b5bddec86f84904a8ccde0bed4d634e3
+size 153630720

onnx/decoder_with_past_model_quantized.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d965705e855008ff2d4778bb3abfae2a8e79f7e9674ba556a0bd88d60f498364
-size 38723193
+oid sha256:4fb3b930e98e235dd81ef5547aa5c882ed4aa163acf0e2323c7a2988a705cf38
+size 307198

onnx/decoder_with_past_model_quantized.onnx_data

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3361716a287b4377334cc90861bde6e9b7250a124744c7021dad54e4d1e07530
+size 38418432

onnx/encoder_model.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4c8493adcf02e3e9afcf8e06a2e28ecf822b12de2c7133f9068e85589e2dedaf
-size 141467235
+oid sha256:35de487ada66c4963c4f8a09d774c1913c210d6481a906746abcd45ee80e7437
+size 142063

onnx/encoder_model.onnx_data

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d30f4a7eb22efea57c1fdbf69600a8d6b302ef7ca711496df59cc1778da2dd8
+size 141330432

onnx/encoder_model_quantized.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:90a7041527a249dc7494b7714f09c2e51b4844b8838816ccf8581de277f07f98
-size 35726832
+oid sha256:3a83b15590f4ed47ba4fec268388c06a400cbdf514a59193617f84f967d515f0
+size 374009

onnx/encoder_model_quantized.onnx_data

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09a9c1b1427227ef5c02e12319c92385100cef863644a06efea32130cdb8e99b
+size 35358720