Upload ONNX weights (#1)

- Upload ONNX weights (276c40be3f547f0ea8efb8352f018d9b03827dda)
- Update config.json (8389978ed15ec2581acbb1e102d39266045ff7b9)


Co-authored-by: Joshua <Xenova@users.noreply.huggingface.co>

config.json

@@ -27,6 +27,12 @@
   "tie_word_embeddings": true,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.42.3",
+  "transformers.js_config": {
+    "kv_cache_dtype": {
+      "q4f16": "float16",
+      "fp16": "float16"
+    }
+  },
   "use_cache": true,
   "vocab_size": 49152
 }

onnx/model.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d6f8ed23d9b947a0f415630dc9848156677834f5353666fce8510dba62cea2d
+size 1638675174

onnx/model_fp16.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e54d13b44aa0457cc2a50815f5b47e39f20bce3a6aab3d954280042b38003ec
+size 819612409

onnx/model_q4.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bcfd45cdfdc459425a306afb360228e9bbdf9b7ce8954f8f8097964662cf89f8
+size 417376345

onnx/model_q4f16.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:027d0926afdd99a4d214dcf13026032ba56e9926b77564d21b94d30de752253f
+size 299634248