Add 4,096-token text buckets (prompts up to 4,096 tokens)

lite/README.md

@@ -19,7 +19,7 @@ tags:
 
 | File | What |
 |---|---|
-| `d3_E256` / `d3_E640` / `d3_E1536.mlpackage` | text graph (fp16 compute, int8 weights) per prompt-length bucket; input = token embeddings with image/video tokens spliced in |
+| `d3_E256` / `d3_E640` / `d3_E1536` / `d3_E4096.mlpackage` | text graph (fp16 compute, int8 weights) per prompt-length bucket (prompts up to 4,096 tokens); input = token embeddings with image/video tokens spliced in |
 | `vision_P320` / `vision_P1536` / `vision_P3072.mlpackage` | vision tower (fp32) per patch bucket; video frames go through it one frame pair at a time (block mask) |
 | `embeddings.f16`, `pos_embed.f32`, `readout.f32`, `tokenizer.json`, `config.json`, `coreml_config.json` | host-side tables and config |
 | `python/` | Python host (coremltools): `d3_coreml_text.py`, `d3_coreml_image.py`, `d3_coreml_video.py` |

lite/coreml_config.json

@@ -307,6 +307,7 @@
  "text_buckets": [
   256,
   640,
-  1536
+  1536,
+  4096
  ]
-}
+}

lite/d3_E4096.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49ee6c2b03212d41380da35ba67b658fe538647c3b316c7e0548095cb3bef849
+size 9753702

lite/d3_E4096.mlpackage/Data/com.apple.CoreML/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b7d9cbc03dbc5d3bff68f38c0d2df0de8256a96afbc5f1b90451c33d5601efe0
+size 502251904

lite/d3_E4096.mlpackage/Manifest.json

@@ -0,0 +1,18 @@
+{
+    "fileFormatVersion": "1.0.0",
+    "itemInfoEntries": {
+        "58BE847F-D9F0-40A3-83D1-FDAE24414FEF": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Weights",
+            "name": "weights",
+            "path": "com.apple.CoreML/weights"
+        },
+        "5A36C8D8-ABB3-490E-84D0-F101B3D31270": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Specification",
+            "name": "model.mlmodel",
+            "path": "com.apple.CoreML/model.mlmodel"
+        }
+    },
+    "rootModelIdentifier": "5A36C8D8-ABB3-490E-84D0-F101B3D31270"
+}

mini/README.md

@@ -19,7 +19,7 @@ tags:
 
 | File | What |
 |---|---|
-| `d3_E256` / `d3_E640` / `d3_E1536.mlpackage` | text graph (fp16 compute, int8 weights) per prompt-length bucket; input = token embeddings with image/video tokens spliced in |
+| `d3_E256` / `d3_E640` / `d3_E1536` / `d3_E4096.mlpackage` | text graph (fp16 compute, int8 weights) per prompt-length bucket (prompts up to 4,096 tokens); input = token embeddings with image/video tokens spliced in |
 | `vision_P320` / `vision_P1536` / `vision_P3072.mlpackage` | vision tower (fp32) per patch bucket; video frames go through it one frame pair at a time (block mask) |
 | `embeddings.f16`, `pos_embed.f32`, `readout.f32`, `tokenizer.json`, `config.json`, `coreml_config.json` | host-side tables and config |
 | `python/` | Python host (coremltools): `d3_coreml_text.py`, `d3_coreml_image.py`, `d3_coreml_video.py` |

mini/coreml_config.json

@@ -307,6 +307,7 @@
  "text_buckets": [
   256,
   640,
-  1536
+  1536,
+  4096
  ]
-}
+}

mini/d3_E4096.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f1988453c0c8a99852bbacad9ff43e7ec0a76cc67b45185e3dc5fefed2e7b60
+size 13037683

mini/d3_E4096.mlpackage/Data/com.apple.CoreML/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:caaf8175b2a879c9fde8a801d9330ce489fb13752328ba2180f35b7951263485
+size 3578663744

mini/d3_E4096.mlpackage/Manifest.json

@@ -0,0 +1,18 @@
+{
+    "fileFormatVersion": "1.0.0",
+    "itemInfoEntries": {
+        "5943285C-0AF3-4E34-B041-0CFB8ADBCCD9": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Weights",
+            "name": "weights",
+            "path": "com.apple.CoreML/weights"
+        },
+        "B9FEEB1E-D394-4E8B-B5F4-9AD35D805F3C": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Specification",
+            "name": "model.mlmodel",
+            "path": "com.apple.CoreML/model.mlmodel"
+        }
+    },
+    "rootModelIdentifier": "B9FEEB1E-D394-4E8B-B5F4-9AD35D805F3C"
+}

nano/README.md

@@ -19,7 +19,7 @@ tags:
 
 | File | What |
 |---|---|
-| `d3_E256` / `d3_E640` / `d3_E1536.mlpackage` | text graph (fp16 compute, int8 weights) per prompt-length bucket; input = token embeddings with image/video tokens spliced in |
+| `d3_E256` / `d3_E640` / `d3_E1536` / `d3_E4096.mlpackage` | text graph (fp16 compute, int8 weights) per prompt-length bucket (prompts up to 4,096 tokens); input = token embeddings with image/video tokens spliced in |
 | `vision_P320` / `vision_P1536` / `vision_P3072.mlpackage` | vision tower (fp32) per patch bucket; video frames go through it one frame pair at a time (block mask) |
 | `embeddings.f16`, `pos_embed.f32`, `readout.f32`, `tokenizer.json`, `config.json`, `coreml_config.json` | host-side tables and config |
 | `python/` | Python host (coremltools): `d3_coreml_text.py`, `d3_coreml_image.py`, `d3_coreml_video.py` |

nano/coreml_config.json

@@ -307,6 +307,7 @@
  "text_buckets": [
   256,
   640,
-  1536
+  1536,
+  4096
  ]
-}
+}

nano/d3_E4096.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c9418d0bcbe0e994567429a2181ccf14a7bc3f51762a1557f77f3a3b449bed4
+size 9753570

nano/d3_E4096.mlpackage/Data/com.apple.CoreML/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:217131052f06fb7a4d0a8c0eaf468a9de3048aa0048f5add1074f4042fb558b4
+size 1377789184

nano/d3_E4096.mlpackage/Manifest.json

@@ -0,0 +1,18 @@
+{
+    "fileFormatVersion": "1.0.0",
+    "itemInfoEntries": {
+        "CD6938DF-7339-4306-B20F-FB6FDF979298": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Weights",
+            "name": "weights",
+            "path": "com.apple.CoreML/weights"
+        },
+        "F49C3D11-1ECB-499C-9D28-F9C2F6B9EED0": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Specification",
+            "name": "model.mlmodel",
+            "path": "com.apple.CoreML/model.mlmodel"
+        }
+    },
+    "rootModelIdentifier": "F49C3D11-1ECB-499C-9D28-F9C2F6B9EED0"
+}