Add verification reports
Browse files
reports/ane-fallback-L128.json
ADDED
|
@@ -0,0 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"hardware": {
|
| 3 |
+
"device": "arm64",
|
| 4 |
+
"chip": "Apple M5 Pro",
|
| 5 |
+
"ram": "24GB",
|
| 6 |
+
"os_version": "macOS 27.0",
|
| 7 |
+
"timestamp": "2026-09-21T03:58:43.638868+00:00"
|
| 8 |
+
},
|
| 9 |
+
"models": [
|
| 10 |
+
{
|
| 11 |
+
"model_path": "/Users/hanweng/Documents/mobius-laya/models/computer-use/laya/coreml/build/laya_multilingual_fp16_L128_options32.mlmodelc",
|
| 12 |
+
"model_name": "laya_multilingual_fp16_L128_options32",
|
| 13 |
+
"fallback": {
|
| 14 |
+
"compute_units": "cpu_and_neural_engine",
|
| 15 |
+
"total_ops": 978,
|
| 16 |
+
"ane_ops": 973,
|
| 17 |
+
"gpu_ops": 0,
|
| 18 |
+
"cpu_ops": 5,
|
| 19 |
+
"ane_percent": 99.5,
|
| 20 |
+
"reasons": [
|
| 21 |
+
{
|
| 22 |
+
"reason": "Unsupported tensor data type: int32",
|
| 23 |
+
"count": 4,
|
| 24 |
+
"estimated_cpu_runtime_ms": 5.7622,
|
| 25 |
+
"op_types": {
|
| 26 |
+
"ios17.cast": 1,
|
| 27 |
+
"ios17.add": 1,
|
| 28 |
+
"select": 1,
|
| 29 |
+
"ios17.gather": 1
|
| 30 |
+
},
|
| 31 |
+
"ops": [
|
| 32 |
+
"attention_mask_to_fp16",
|
| 33 |
+
"add_0",
|
| 34 |
+
"select_0",
|
| 35 |
+
"input_3_cast_fp16"
|
| 36 |
+
]
|
| 37 |
+
},
|
| 38 |
+
{
|
| 39 |
+
"reason": "Unable to resolve operation input \"y\".",
|
| 40 |
+
"count": 1,
|
| 41 |
+
"estimated_cpu_runtime_ms": 0.001,
|
| 42 |
+
"op_types": {
|
| 43 |
+
"ios17.greater_equal": 1
|
| 44 |
+
},
|
| 45 |
+
"ops": [
|
| 46 |
+
"greater_equal_0"
|
| 47 |
+
]
|
| 48 |
+
}
|
| 49 |
+
]
|
| 50 |
+
}
|
| 51 |
+
}
|
| 52 |
+
]
|
| 53 |
+
}
|
reports/ane-fallback-L256.json
ADDED
|
@@ -0,0 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"hardware": {
|
| 3 |
+
"device": "arm64",
|
| 4 |
+
"chip": "Apple M5 Pro",
|
| 5 |
+
"ram": "24GB",
|
| 6 |
+
"os_version": "macOS 27.0",
|
| 7 |
+
"timestamp": "2026-09-21T03:58:48.759760+00:00"
|
| 8 |
+
},
|
| 9 |
+
"models": [
|
| 10 |
+
{
|
| 11 |
+
"model_path": "/Users/hanweng/Documents/mobius-laya/models/computer-use/laya/coreml/build/laya_multilingual_fp16_L256_options32.mlmodelc",
|
| 12 |
+
"model_name": "laya_multilingual_fp16_L256_options32",
|
| 13 |
+
"fallback": {
|
| 14 |
+
"compute_units": "cpu_and_neural_engine",
|
| 15 |
+
"total_ops": 978,
|
| 16 |
+
"ane_ops": 973,
|
| 17 |
+
"gpu_ops": 0,
|
| 18 |
+
"cpu_ops": 5,
|
| 19 |
+
"ane_percent": 99.5,
|
| 20 |
+
"reasons": [
|
| 21 |
+
{
|
| 22 |
+
"reason": "Unsupported tensor data type: int32",
|
| 23 |
+
"count": 4,
|
| 24 |
+
"estimated_cpu_runtime_ms": 5.7651,
|
| 25 |
+
"op_types": {
|
| 26 |
+
"ios17.cast": 1,
|
| 27 |
+
"ios17.add": 1,
|
| 28 |
+
"select": 1,
|
| 29 |
+
"ios17.gather": 1
|
| 30 |
+
},
|
| 31 |
+
"ops": [
|
| 32 |
+
"attention_mask_to_fp16",
|
| 33 |
+
"add_0",
|
| 34 |
+
"select_0",
|
| 35 |
+
"input_3_cast_fp16"
|
| 36 |
+
]
|
| 37 |
+
},
|
| 38 |
+
{
|
| 39 |
+
"reason": "Unable to resolve operation input \"y\".",
|
| 40 |
+
"count": 1,
|
| 41 |
+
"estimated_cpu_runtime_ms": 0.001,
|
| 42 |
+
"op_types": {
|
| 43 |
+
"ios17.greater_equal": 1
|
| 44 |
+
},
|
| 45 |
+
"ops": [
|
| 46 |
+
"greater_equal_0"
|
| 47 |
+
]
|
| 48 |
+
}
|
| 49 |
+
]
|
| 50 |
+
}
|
| 51 |
+
}
|
| 52 |
+
]
|
| 53 |
+
}
|
reports/ane-fallback-L512.json
ADDED
|
@@ -0,0 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"hardware": {
|
| 3 |
+
"device": "arm64",
|
| 4 |
+
"chip": "Apple M5 Pro",
|
| 5 |
+
"ram": "24GB",
|
| 6 |
+
"os_version": "macOS 27.0",
|
| 7 |
+
"timestamp": "2026-09-21T03:58:54.139733+00:00"
|
| 8 |
+
},
|
| 9 |
+
"models": [
|
| 10 |
+
{
|
| 11 |
+
"model_path": "/Users/hanweng/Documents/mobius-laya/models/computer-use/laya/coreml/build/laya_multilingual_fp16_L512_options32.mlmodelc",
|
| 12 |
+
"model_name": "laya_multilingual_fp16_L512_options32",
|
| 13 |
+
"fallback": {
|
| 14 |
+
"compute_units": "cpu_and_neural_engine",
|
| 15 |
+
"total_ops": 978,
|
| 16 |
+
"ane_ops": 973,
|
| 17 |
+
"gpu_ops": 0,
|
| 18 |
+
"cpu_ops": 5,
|
| 19 |
+
"ane_percent": 99.5,
|
| 20 |
+
"reasons": [
|
| 21 |
+
{
|
| 22 |
+
"reason": "Unsupported tensor data type: int32",
|
| 23 |
+
"count": 4,
|
| 24 |
+
"estimated_cpu_runtime_ms": 5.7709,
|
| 25 |
+
"op_types": {
|
| 26 |
+
"ios17.cast": 1,
|
| 27 |
+
"ios17.add": 1,
|
| 28 |
+
"select": 1,
|
| 29 |
+
"ios17.gather": 1
|
| 30 |
+
},
|
| 31 |
+
"ops": [
|
| 32 |
+
"attention_mask_to_fp16",
|
| 33 |
+
"add_0",
|
| 34 |
+
"select_0",
|
| 35 |
+
"input_3_cast_fp16"
|
| 36 |
+
]
|
| 37 |
+
},
|
| 38 |
+
{
|
| 39 |
+
"reason": "Unable to resolve operation input \"y\".",
|
| 40 |
+
"count": 1,
|
| 41 |
+
"estimated_cpu_runtime_ms": 0.001,
|
| 42 |
+
"op_types": {
|
| 43 |
+
"ios17.greater_equal": 1
|
| 44 |
+
},
|
| 45 |
+
"ops": [
|
| 46 |
+
"greater_equal_0"
|
| 47 |
+
]
|
| 48 |
+
}
|
| 49 |
+
]
|
| 50 |
+
}
|
| 51 |
+
}
|
| 52 |
+
]
|
| 53 |
+
}
|
reports/verification-multilingual-L128.json
ADDED
|
@@ -0,0 +1,579 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"package": "laya_multilingual_fp16_L128_options32.mlpackage",
|
| 3 |
+
"package_files_sha256": {
|
| 4 |
+
"Data/com.apple.CoreML/model.mlmodel": "a536e177f174cc14a0dbe539357db29016a10ad579cdbf040b98525f32e7a4a7",
|
| 5 |
+
"Data/com.apple.CoreML/weights/weight.bin": "506eb53e74aaebcdfb3f55304fbdd0b9c1678026efaef60454a1c66557f35d90",
|
| 6 |
+
"Manifest.json": "9957a84277c0a722bcdbbc69ae3ce065122d9dd14de85ee9c76a2e44b2cb7146"
|
| 7 |
+
},
|
| 8 |
+
"source_repo": "convaiinnovations/laya",
|
| 9 |
+
"source_revision": "1c5edc17a7acd8701df6fc341c0d179f1c62c982",
|
| 10 |
+
"variant": "multilingual",
|
| 11 |
+
"length": 128,
|
| 12 |
+
"max_options": 32,
|
| 13 |
+
"environment": {
|
| 14 |
+
"chip": "Apple M5 Pro",
|
| 15 |
+
"os": "27.0",
|
| 16 |
+
"python": "3.12.13",
|
| 17 |
+
"torch": "2.7.0",
|
| 18 |
+
"coremltools": "9.0"
|
| 19 |
+
},
|
| 20 |
+
"gates": {
|
| 21 |
+
"argmax_agreement": "100%",
|
| 22 |
+
"max_probability_error": 0.02,
|
| 23 |
+
"max_action_probability_error": 0.02,
|
| 24 |
+
"finite": true,
|
| 25 |
+
"padding_masked": true
|
| 26 |
+
},
|
| 27 |
+
"cases_sha256": "4de4e34309c421913524209b62bb7c8a0a232928bd228ec12e884347ce65eeed",
|
| 28 |
+
"runs": {
|
| 29 |
+
"ALL": {
|
| 30 |
+
"compute_units": "ALL",
|
| 31 |
+
"load_seconds": 1.9205929159652442,
|
| 32 |
+
"questions": 16,
|
| 33 |
+
"argmax_agreements": 16,
|
| 34 |
+
"max_probability_error": 0.0021350979804992676,
|
| 35 |
+
"max_action_probability_error": 0.0,
|
| 36 |
+
"latency_ms": {
|
| 37 |
+
"p50": 3.874978981912136,
|
| 38 |
+
"p95": 4.150477421353571,
|
| 39 |
+
"min": 3.734458005055785,
|
| 40 |
+
"max": 6.0334589798003435
|
| 41 |
+
},
|
| 42 |
+
"skipped": [],
|
| 43 |
+
"rows": [
|
| 44 |
+
{
|
| 45 |
+
"case": "tetris_clean",
|
| 46 |
+
"question": "clean",
|
| 47 |
+
"type": "noul",
|
| 48 |
+
"tokens": 80,
|
| 49 |
+
"options": 2,
|
| 50 |
+
"argmax_agrees": true,
|
| 51 |
+
"coreml_argmax": 0,
|
| 52 |
+
"reference_argmax": 0,
|
| 53 |
+
"max_probability_error": 0.0001417025923728943,
|
| 54 |
+
"max_action_probability_error": 0.0,
|
| 55 |
+
"finite": true,
|
| 56 |
+
"padding_masked": true,
|
| 57 |
+
"probability_sum": 1.00006103515625,
|
| 58 |
+
"median_ms": 3.948333003791049
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"case": "tetris_flat",
|
| 62 |
+
"question": "clean",
|
| 63 |
+
"type": "noul",
|
| 64 |
+
"tokens": 84,
|
| 65 |
+
"options": 2,
|
| 66 |
+
"argmax_agrees": true,
|
| 67 |
+
"coreml_argmax": 1,
|
| 68 |
+
"reference_argmax": 1,
|
| 69 |
+
"max_probability_error": 0.00014770030975341797,
|
| 70 |
+
"max_action_probability_error": 0.0,
|
| 71 |
+
"finite": true,
|
| 72 |
+
"padding_masked": true,
|
| 73 |
+
"probability_sum": 0.9998626708984375,
|
| 74 |
+
"median_ms": 3.872645494993776
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"case": "tetris_lane",
|
| 78 |
+
"question": "lane",
|
| 79 |
+
"type": "choice",
|
| 80 |
+
"tokens": 62,
|
| 81 |
+
"options": 4,
|
| 82 |
+
"argmax_agrees": true,
|
| 83 |
+
"coreml_argmax": 1,
|
| 84 |
+
"reference_argmax": 1,
|
| 85 |
+
"max_probability_error": 0.0005466341972351074,
|
| 86 |
+
"max_action_probability_error": 0.0,
|
| 87 |
+
"finite": true,
|
| 88 |
+
"padding_masked": true,
|
| 89 |
+
"probability_sum": 1.0001220703125,
|
| 90 |
+
"median_ms": 4.006020986707881
|
| 91 |
+
},
|
| 92 |
+
{
|
| 93 |
+
"case": "tetris_lane",
|
| 94 |
+
"question": "risk",
|
| 95 |
+
"type": "score",
|
| 96 |
+
"tokens": 64,
|
| 97 |
+
"options": 3,
|
| 98 |
+
"argmax_agrees": true,
|
| 99 |
+
"coreml_argmax": 1,
|
| 100 |
+
"reference_argmax": 1,
|
| 101 |
+
"max_probability_error": 0.00014688074588775635,
|
| 102 |
+
"max_action_probability_error": 0.0,
|
| 103 |
+
"finite": true,
|
| 104 |
+
"padding_masked": true,
|
| 105 |
+
"probability_sum": 1.0,
|
| 106 |
+
"median_ms": 3.9132919919211417
|
| 107 |
+
},
|
| 108 |
+
{
|
| 109 |
+
"case": "form_field",
|
| 110 |
+
"question": "action",
|
| 111 |
+
"type": "choice",
|
| 112 |
+
"tokens": 104,
|
| 113 |
+
"options": 3,
|
| 114 |
+
"argmax_agrees": true,
|
| 115 |
+
"coreml_argmax": 0,
|
| 116 |
+
"reference_argmax": 0,
|
| 117 |
+
"max_probability_error": 0.0021350979804992676,
|
| 118 |
+
"max_action_probability_error": 0.0,
|
| 119 |
+
"finite": true,
|
| 120 |
+
"padding_masked": true,
|
| 121 |
+
"probability_sum": 0.9998779296875,
|
| 122 |
+
"median_ms": 3.8335000281222165
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"case": "support_triage",
|
| 126 |
+
"question": "intent",
|
| 127 |
+
"type": "choice",
|
| 128 |
+
"tokens": 62,
|
| 129 |
+
"options": 4,
|
| 130 |
+
"argmax_agrees": true,
|
| 131 |
+
"coreml_argmax": 1,
|
| 132 |
+
"reference_argmax": 1,
|
| 133 |
+
"max_probability_error": 5.608797073364258e-05,
|
| 134 |
+
"max_action_probability_error": 0.0,
|
| 135 |
+
"finite": true,
|
| 136 |
+
"padding_masked": true,
|
| 137 |
+
"probability_sum": 1.0000602006912231,
|
| 138 |
+
"median_ms": 3.799437516136095
|
| 139 |
+
},
|
| 140 |
+
{
|
| 141 |
+
"case": "support_triage",
|
| 142 |
+
"question": "frustration",
|
| 143 |
+
"type": "score",
|
| 144 |
+
"tokens": 69,
|
| 145 |
+
"options": 3,
|
| 146 |
+
"argmax_agrees": true,
|
| 147 |
+
"coreml_argmax": 1,
|
| 148 |
+
"reference_argmax": 1,
|
| 149 |
+
"max_probability_error": 0.0008405148983001709,
|
| 150 |
+
"max_action_probability_error": 0.0,
|
| 151 |
+
"finite": true,
|
| 152 |
+
"padding_masked": true,
|
| 153 |
+
"probability_sum": 1.00006103515625,
|
| 154 |
+
"median_ms": 3.865458507789299
|
| 155 |
+
},
|
| 156 |
+
{
|
| 157 |
+
"case": "support_triage",
|
| 158 |
+
"question": "churn",
|
| 159 |
+
"type": "noul",
|
| 160 |
+
"tokens": 71,
|
| 161 |
+
"options": 2,
|
| 162 |
+
"argmax_agrees": true,
|
| 163 |
+
"coreml_argmax": 0,
|
| 164 |
+
"reference_argmax": 0,
|
| 165 |
+
"max_probability_error": 3.820657730102539e-05,
|
| 166 |
+
"max_action_probability_error": 0.0,
|
| 167 |
+
"finite": true,
|
| 168 |
+
"padding_masked": true,
|
| 169 |
+
"probability_sum": 0.9999732971191406,
|
| 170 |
+
"median_ms": 3.829437482636422
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"case": "zh",
|
| 174 |
+
"question": "intent",
|
| 175 |
+
"type": "choice",
|
| 176 |
+
"tokens": 37,
|
| 177 |
+
"options": 3,
|
| 178 |
+
"argmax_agrees": true,
|
| 179 |
+
"coreml_argmax": 0,
|
| 180 |
+
"reference_argmax": 0,
|
| 181 |
+
"max_probability_error": 0.00023287534713745117,
|
| 182 |
+
"max_action_probability_error": 0.0,
|
| 183 |
+
"finite": true,
|
| 184 |
+
"padding_masked": true,
|
| 185 |
+
"probability_sum": 0.9997743368148804,
|
| 186 |
+
"median_ms": 3.8122705300338566
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"case": "zh",
|
| 190 |
+
"question": "urgent",
|
| 191 |
+
"type": "noul",
|
| 192 |
+
"tokens": 48,
|
| 193 |
+
"options": 2,
|
| 194 |
+
"argmax_agrees": true,
|
| 195 |
+
"coreml_argmax": 1,
|
| 196 |
+
"reference_argmax": 1,
|
| 197 |
+
"max_probability_error": 0.00011944770812988281,
|
| 198 |
+
"max_action_probability_error": 0.0,
|
| 199 |
+
"finite": true,
|
| 200 |
+
"padding_masked": true,
|
| 201 |
+
"probability_sum": 0.999786376953125,
|
| 202 |
+
"median_ms": 3.9022920245770365
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"case": "ja",
|
| 206 |
+
"question": "intent",
|
| 207 |
+
"type": "choice",
|
| 208 |
+
"tokens": 45,
|
| 209 |
+
"options": 3,
|
| 210 |
+
"argmax_agrees": true,
|
| 211 |
+
"coreml_argmax": 0,
|
| 212 |
+
"reference_argmax": 0,
|
| 213 |
+
"max_probability_error": 0.00017899274826049805,
|
| 214 |
+
"max_action_probability_error": 0.0,
|
| 215 |
+
"finite": true,
|
| 216 |
+
"padding_masked": true,
|
| 217 |
+
"probability_sum": 1.000178337097168,
|
| 218 |
+
"median_ms": 3.817520511802286
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"case": "de",
|
| 222 |
+
"question": "intent",
|
| 223 |
+
"type": "choice",
|
| 224 |
+
"tokens": 44,
|
| 225 |
+
"options": 3,
|
| 226 |
+
"argmax_agrees": true,
|
| 227 |
+
"coreml_argmax": 0,
|
| 228 |
+
"reference_argmax": 0,
|
| 229 |
+
"max_probability_error": 7.051229476928711e-05,
|
| 230 |
+
"max_action_probability_error": 0.0,
|
| 231 |
+
"finite": true,
|
| 232 |
+
"padding_masked": true,
|
| 233 |
+
"probability_sum": 0.9998912811279297,
|
| 234 |
+
"median_ms": 3.8392704736907035
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"case": "many_options",
|
| 238 |
+
"question": "lane",
|
| 239 |
+
"type": "choice",
|
| 240 |
+
"tokens": 128,
|
| 241 |
+
"options": 20,
|
| 242 |
+
"argmax_agrees": true,
|
| 243 |
+
"coreml_argmax": 3,
|
| 244 |
+
"reference_argmax": 3,
|
| 245 |
+
"max_probability_error": 0.0005449950695037842,
|
| 246 |
+
"max_action_probability_error": 0.0,
|
| 247 |
+
"finite": true,
|
| 248 |
+
"padding_masked": true,
|
| 249 |
+
"probability_sum": 0.9999656677246094,
|
| 250 |
+
"median_ms": 4.023146029794589
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"case": "guard",
|
| 254 |
+
"question": "injection",
|
| 255 |
+
"type": "noul",
|
| 256 |
+
"tokens": 48,
|
| 257 |
+
"options": 2,
|
| 258 |
+
"argmax_agrees": true,
|
| 259 |
+
"coreml_argmax": 1,
|
| 260 |
+
"reference_argmax": 1,
|
| 261 |
+
"max_probability_error": 0.0011126399040222168,
|
| 262 |
+
"max_action_probability_error": 0.0,
|
| 263 |
+
"finite": true,
|
| 264 |
+
"padding_masked": true,
|
| 265 |
+
"probability_sum": 1.0,
|
| 266 |
+
"median_ms": 3.868666506605223
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"case": "long_state",
|
| 270 |
+
"question": "ready",
|
| 271 |
+
"type": "noul",
|
| 272 |
+
"tokens": 128,
|
| 273 |
+
"options": 2,
|
| 274 |
+
"argmax_agrees": true,
|
| 275 |
+
"coreml_argmax": 1,
|
| 276 |
+
"reference_argmax": 1,
|
| 277 |
+
"max_probability_error": 0.0001234356313943863,
|
| 278 |
+
"max_action_probability_error": 0.0,
|
| 279 |
+
"finite": true,
|
| 280 |
+
"padding_masked": true,
|
| 281 |
+
"probability_sum": 0.999847412109375,
|
| 282 |
+
"median_ms": 3.8872919976711273
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"case": "long_state",
|
| 286 |
+
"question": "owner",
|
| 287 |
+
"type": "choice",
|
| 288 |
+
"tokens": 128,
|
| 289 |
+
"options": 3,
|
| 290 |
+
"argmax_agrees": true,
|
| 291 |
+
"coreml_argmax": 2,
|
| 292 |
+
"reference_argmax": 2,
|
| 293 |
+
"max_probability_error": 0.0010861754417419434,
|
| 294 |
+
"max_action_probability_error": 0.0,
|
| 295 |
+
"finite": true,
|
| 296 |
+
"padding_masked": true,
|
| 297 |
+
"probability_sum": 0.9999961853027344,
|
| 298 |
+
"median_ms": 3.830937494058162
|
| 299 |
+
}
|
| 300 |
+
],
|
| 301 |
+
"passed": true
|
| 302 |
+
},
|
| 303 |
+
"CPU_AND_NE": {
|
| 304 |
+
"compute_units": "CPU_AND_NE",
|
| 305 |
+
"load_seconds": 3.0328945000073873,
|
| 306 |
+
"questions": 16,
|
| 307 |
+
"argmax_agreements": 16,
|
| 308 |
+
"max_probability_error": 0.012559235095977783,
|
| 309 |
+
"max_action_probability_error": 0.0,
|
| 310 |
+
"latency_ms": {
|
| 311 |
+
"p50": 3.6408329906407744,
|
| 312 |
+
"p95": 3.7919832742772996,
|
| 313 |
+
"min": 3.5940000088885427,
|
| 314 |
+
"max": 6.553583953063935
|
| 315 |
+
},
|
| 316 |
+
"skipped": [],
|
| 317 |
+
"rows": [
|
| 318 |
+
{
|
| 319 |
+
"case": "tetris_clean",
|
| 320 |
+
"question": "clean",
|
| 321 |
+
"type": "noul",
|
| 322 |
+
"tokens": 80,
|
| 323 |
+
"options": 2,
|
| 324 |
+
"argmax_agrees": true,
|
| 325 |
+
"coreml_argmax": 0,
|
| 326 |
+
"reference_argmax": 0,
|
| 327 |
+
"max_probability_error": 0.0004076361656188965,
|
| 328 |
+
"max_action_probability_error": 0.0,
|
| 329 |
+
"finite": true,
|
| 330 |
+
"padding_masked": true,
|
| 331 |
+
"probability_sum": 1.00018310546875,
|
| 332 |
+
"median_ms": 3.6329374997876585
|
| 333 |
+
},
|
| 334 |
+
{
|
| 335 |
+
"case": "tetris_flat",
|
| 336 |
+
"question": "clean",
|
| 337 |
+
"type": "noul",
|
| 338 |
+
"tokens": 84,
|
| 339 |
+
"options": 2,
|
| 340 |
+
"argmax_agrees": true,
|
| 341 |
+
"coreml_argmax": 1,
|
| 342 |
+
"reference_argmax": 1,
|
| 343 |
+
"max_probability_error": 0.00034058094024658203,
|
| 344 |
+
"max_action_probability_error": 0.0,
|
| 345 |
+
"finite": true,
|
| 346 |
+
"padding_masked": true,
|
| 347 |
+
"probability_sum": 1.0003662109375,
|
| 348 |
+
"median_ms": 3.6503754963632673
|
| 349 |
+
},
|
| 350 |
+
{
|
| 351 |
+
"case": "tetris_lane",
|
| 352 |
+
"question": "lane",
|
| 353 |
+
"type": "choice",
|
| 354 |
+
"tokens": 62,
|
| 355 |
+
"options": 4,
|
| 356 |
+
"argmax_agrees": true,
|
| 357 |
+
"coreml_argmax": 1,
|
| 358 |
+
"reference_argmax": 1,
|
| 359 |
+
"max_probability_error": 0.0033596158027648926,
|
| 360 |
+
"max_action_probability_error": 0.0,
|
| 361 |
+
"finite": true,
|
| 362 |
+
"padding_masked": true,
|
| 363 |
+
"probability_sum": 0.9998779296875,
|
| 364 |
+
"median_ms": 3.6310830037109554
|
| 365 |
+
},
|
| 366 |
+
{
|
| 367 |
+
"case": "tetris_lane",
|
| 368 |
+
"question": "risk",
|
| 369 |
+
"type": "score",
|
| 370 |
+
"tokens": 64,
|
| 371 |
+
"options": 3,
|
| 372 |
+
"argmax_agrees": true,
|
| 373 |
+
"coreml_argmax": 1,
|
| 374 |
+
"reference_argmax": 1,
|
| 375 |
+
"max_probability_error": 0.0014896541833877563,
|
| 376 |
+
"max_action_probability_error": 0.0,
|
| 377 |
+
"finite": true,
|
| 378 |
+
"padding_masked": true,
|
| 379 |
+
"probability_sum": 0.9996337890625,
|
| 380 |
+
"median_ms": 3.644520998932421
|
| 381 |
+
},
|
| 382 |
+
{
|
| 383 |
+
"case": "form_field",
|
| 384 |
+
"question": "action",
|
| 385 |
+
"type": "choice",
|
| 386 |
+
"tokens": 104,
|
| 387 |
+
"options": 3,
|
| 388 |
+
"argmax_agrees": true,
|
| 389 |
+
"coreml_argmax": 0,
|
| 390 |
+
"reference_argmax": 0,
|
| 391 |
+
"max_probability_error": 0.0034949183464050293,
|
| 392 |
+
"max_action_probability_error": 0.0,
|
| 393 |
+
"finite": true,
|
| 394 |
+
"padding_masked": true,
|
| 395 |
+
"probability_sum": 0.9996337890625,
|
| 396 |
+
"median_ms": 3.6403754784259945
|
| 397 |
+
},
|
| 398 |
+
{
|
| 399 |
+
"case": "support_triage",
|
| 400 |
+
"question": "intent",
|
| 401 |
+
"type": "choice",
|
| 402 |
+
"tokens": 62,
|
| 403 |
+
"options": 4,
|
| 404 |
+
"argmax_agrees": true,
|
| 405 |
+
"coreml_argmax": 1,
|
| 406 |
+
"reference_argmax": 1,
|
| 407 |
+
"max_probability_error": 0.0004321932792663574,
|
| 408 |
+
"max_action_probability_error": 0.0,
|
| 409 |
+
"finite": true,
|
| 410 |
+
"padding_masked": true,
|
| 411 |
+
"probability_sum": 0.9995682239532471,
|
| 412 |
+
"median_ms": 3.631958010373637
|
| 413 |
+
},
|
| 414 |
+
{
|
| 415 |
+
"case": "support_triage",
|
| 416 |
+
"question": "frustration",
|
| 417 |
+
"type": "score",
|
| 418 |
+
"tokens": 69,
|
| 419 |
+
"options": 3,
|
| 420 |
+
"argmax_agrees": true,
|
| 421 |
+
"coreml_argmax": 1,
|
| 422 |
+
"reference_argmax": 1,
|
| 423 |
+
"max_probability_error": 0.001600891351699829,
|
| 424 |
+
"max_action_probability_error": 0.0,
|
| 425 |
+
"finite": true,
|
| 426 |
+
"padding_masked": true,
|
| 427 |
+
"probability_sum": 0.99981689453125,
|
| 428 |
+
"median_ms": 3.6398124939296395
|
| 429 |
+
},
|
| 430 |
+
{
|
| 431 |
+
"case": "support_triage",
|
| 432 |
+
"question": "churn",
|
| 433 |
+
"type": "noul",
|
| 434 |
+
"tokens": 71,
|
| 435 |
+
"options": 2,
|
| 436 |
+
"argmax_agrees": true,
|
| 437 |
+
"coreml_argmax": 0,
|
| 438 |
+
"reference_argmax": 0,
|
| 439 |
+
"max_probability_error": 3.820657730102539e-05,
|
| 440 |
+
"max_action_probability_error": 0.0,
|
| 441 |
+
"finite": true,
|
| 442 |
+
"padding_masked": true,
|
| 443 |
+
"probability_sum": 0.9999771118164062,
|
| 444 |
+
"median_ms": 3.6401874967850745
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"case": "zh",
|
| 448 |
+
"question": "intent",
|
| 449 |
+
"type": "choice",
|
| 450 |
+
"tokens": 37,
|
| 451 |
+
"options": 3,
|
| 452 |
+
"argmax_agrees": true,
|
| 453 |
+
"coreml_argmax": 0,
|
| 454 |
+
"reference_argmax": 0,
|
| 455 |
+
"max_probability_error": 0.00023287534713745117,
|
| 456 |
+
"max_action_probability_error": 0.0,
|
| 457 |
+
"finite": true,
|
| 458 |
+
"padding_masked": true,
|
| 459 |
+
"probability_sum": 0.999773383140564,
|
| 460 |
+
"median_ms": 3.7934584834147245
|
| 461 |
+
},
|
| 462 |
+
{
|
| 463 |
+
"case": "zh",
|
| 464 |
+
"question": "urgent",
|
| 465 |
+
"type": "noul",
|
| 466 |
+
"tokens": 48,
|
| 467 |
+
"options": 2,
|
| 468 |
+
"argmax_agrees": true,
|
| 469 |
+
"coreml_argmax": 1,
|
| 470 |
+
"reference_argmax": 1,
|
| 471 |
+
"max_probability_error": 0.0004093870520591736,
|
| 472 |
+
"max_action_probability_error": 0.0,
|
| 473 |
+
"finite": true,
|
| 474 |
+
"padding_masked": true,
|
| 475 |
+
"probability_sum": 1.0002899169921875,
|
| 476 |
+
"median_ms": 3.753063007025048
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"case": "ja",
|
| 480 |
+
"question": "intent",
|
| 481 |
+
"type": "choice",
|
| 482 |
+
"tokens": 45,
|
| 483 |
+
"options": 3,
|
| 484 |
+
"argmax_agrees": true,
|
| 485 |
+
"coreml_argmax": 0,
|
| 486 |
+
"reference_argmax": 0,
|
| 487 |
+
"max_probability_error": 0.00017899274826049805,
|
| 488 |
+
"max_action_probability_error": 0.0,
|
| 489 |
+
"finite": true,
|
| 490 |
+
"padding_masked": true,
|
| 491 |
+
"probability_sum": 1.0001769065856934,
|
| 492 |
+
"median_ms": 3.62402101745829
|
| 493 |
+
},
|
| 494 |
+
{
|
| 495 |
+
"case": "de",
|
| 496 |
+
"question": "intent",
|
| 497 |
+
"type": "choice",
|
| 498 |
+
"tokens": 44,
|
| 499 |
+
"options": 3,
|
| 500 |
+
"argmax_agrees": true,
|
| 501 |
+
"coreml_argmax": 0,
|
| 502 |
+
"reference_argmax": 0,
|
| 503 |
+
"max_probability_error": 7.051229476928711e-05,
|
| 504 |
+
"max_action_probability_error": 0.0,
|
| 505 |
+
"finite": true,
|
| 506 |
+
"padding_masked": true,
|
| 507 |
+
"probability_sum": 0.9998779296875,
|
| 508 |
+
"median_ms": 3.6391039902810007
|
| 509 |
+
},
|
| 510 |
+
{
|
| 511 |
+
"case": "many_options",
|
| 512 |
+
"question": "lane",
|
| 513 |
+
"type": "choice",
|
| 514 |
+
"tokens": 128,
|
| 515 |
+
"options": 20,
|
| 516 |
+
"argmax_agrees": true,
|
| 517 |
+
"coreml_argmax": 3,
|
| 518 |
+
"reference_argmax": 3,
|
| 519 |
+
"max_probability_error": 0.004200160503387451,
|
| 520 |
+
"max_action_probability_error": 0.0,
|
| 521 |
+
"finite": true,
|
| 522 |
+
"padding_masked": true,
|
| 523 |
+
"probability_sum": 1.0005359649658203,
|
| 524 |
+
"median_ms": 3.6420005199033767
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"case": "guard",
|
| 528 |
+
"question": "injection",
|
| 529 |
+
"type": "noul",
|
| 530 |
+
"tokens": 48,
|
| 531 |
+
"options": 2,
|
| 532 |
+
"argmax_agrees": true,
|
| 533 |
+
"coreml_argmax": 1,
|
| 534 |
+
"reference_argmax": 1,
|
| 535 |
+
"max_probability_error": 0.012559235095977783,
|
| 536 |
+
"max_action_probability_error": 0.0,
|
| 537 |
+
"finite": true,
|
| 538 |
+
"padding_masked": true,
|
| 539 |
+
"probability_sum": 1.000244140625,
|
| 540 |
+
"median_ms": 3.6424169957172126
|
| 541 |
+
},
|
| 542 |
+
{
|
| 543 |
+
"case": "long_state",
|
| 544 |
+
"question": "ready",
|
| 545 |
+
"type": "noul",
|
| 546 |
+
"tokens": 128,
|
| 547 |
+
"options": 2,
|
| 548 |
+
"argmax_agrees": true,
|
| 549 |
+
"coreml_argmax": 1,
|
| 550 |
+
"reference_argmax": 1,
|
| 551 |
+
"max_probability_error": 0.0002122577279806137,
|
| 552 |
+
"max_action_probability_error": 0.0,
|
| 553 |
+
"finite": true,
|
| 554 |
+
"padding_masked": true,
|
| 555 |
+
"probability_sum": 1.00018310546875,
|
| 556 |
+
"median_ms": 3.634604043327272
|
| 557 |
+
},
|
| 558 |
+
{
|
| 559 |
+
"case": "long_state",
|
| 560 |
+
"question": "owner",
|
| 561 |
+
"type": "choice",
|
| 562 |
+
"tokens": 128,
|
| 563 |
+
"options": 3,
|
| 564 |
+
"argmax_agrees": true,
|
| 565 |
+
"coreml_argmax": 2,
|
| 566 |
+
"reference_argmax": 2,
|
| 567 |
+
"max_probability_error": 0.0022311657667160034,
|
| 568 |
+
"max_action_probability_error": 0.0,
|
| 569 |
+
"finite": true,
|
| 570 |
+
"padding_masked": true,
|
| 571 |
+
"probability_sum": 1.0003395080566406,
|
| 572 |
+
"median_ms": 3.628145990660414
|
| 573 |
+
}
|
| 574 |
+
],
|
| 575 |
+
"passed": true
|
| 576 |
+
}
|
| 577 |
+
},
|
| 578 |
+
"passed": true
|
| 579 |
+
}
|
reports/verification-multilingual-L256.json
ADDED
|
@@ -0,0 +1,579 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"package": "laya_multilingual_fp16_L256_options32.mlpackage",
|
| 3 |
+
"package_files_sha256": {
|
| 4 |
+
"Data/com.apple.CoreML/model.mlmodel": "e6afbab8c933fc317626764a872bbbede233faeb4591222b0742330d77b829f8",
|
| 5 |
+
"Data/com.apple.CoreML/weights/weight.bin": "41e287a96c4260faec5c1d2aa37209666279a69541652ea450879e5853881d73",
|
| 6 |
+
"Manifest.json": "4a04e53c9966399b9b1b38f93ac2688db2ef79a3b24d4866bdb93015c162c13a"
|
| 7 |
+
},
|
| 8 |
+
"source_repo": "convaiinnovations/laya",
|
| 9 |
+
"source_revision": "1c5edc17a7acd8701df6fc341c0d179f1c62c982",
|
| 10 |
+
"variant": "multilingual",
|
| 11 |
+
"length": 256,
|
| 12 |
+
"max_options": 32,
|
| 13 |
+
"environment": {
|
| 14 |
+
"chip": "Apple M5 Pro",
|
| 15 |
+
"os": "27.0",
|
| 16 |
+
"python": "3.12.13",
|
| 17 |
+
"torch": "2.7.0",
|
| 18 |
+
"coremltools": "9.0"
|
| 19 |
+
},
|
| 20 |
+
"gates": {
|
| 21 |
+
"argmax_agreement": "100%",
|
| 22 |
+
"max_probability_error": 0.02,
|
| 23 |
+
"max_action_probability_error": 0.02,
|
| 24 |
+
"finite": true,
|
| 25 |
+
"padding_masked": true
|
| 26 |
+
},
|
| 27 |
+
"cases_sha256": "4de4e34309c421913524209b62bb7c8a0a232928bd228ec12e884347ce65eeed",
|
| 28 |
+
"runs": {
|
| 29 |
+
"ALL": {
|
| 30 |
+
"compute_units": "ALL",
|
| 31 |
+
"load_seconds": 1.829418750014156,
|
| 32 |
+
"questions": 16,
|
| 33 |
+
"argmax_agreements": 16,
|
| 34 |
+
"max_probability_error": 0.0021350979804992676,
|
| 35 |
+
"max_action_probability_error": 0.0,
|
| 36 |
+
"latency_ms": {
|
| 37 |
+
"p50": 5.186791997402906,
|
| 38 |
+
"p95": 5.620151464245282,
|
| 39 |
+
"min": 5.075542023405433,
|
| 40 |
+
"max": 6.205584038980305
|
| 41 |
+
},
|
| 42 |
+
"skipped": [],
|
| 43 |
+
"rows": [
|
| 44 |
+
{
|
| 45 |
+
"case": "tetris_clean",
|
| 46 |
+
"question": "clean",
|
| 47 |
+
"type": "noul",
|
| 48 |
+
"tokens": 80,
|
| 49 |
+
"options": 2,
|
| 50 |
+
"argmax_agrees": true,
|
| 51 |
+
"coreml_argmax": 0,
|
| 52 |
+
"reference_argmax": 0,
|
| 53 |
+
"max_probability_error": 0.0001417025923728943,
|
| 54 |
+
"max_action_probability_error": 0.0,
|
| 55 |
+
"finite": true,
|
| 56 |
+
"padding_masked": true,
|
| 57 |
+
"probability_sum": 1.00006103515625,
|
| 58 |
+
"median_ms": 5.240916507318616
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"case": "tetris_flat",
|
| 62 |
+
"question": "clean",
|
| 63 |
+
"type": "noul",
|
| 64 |
+
"tokens": 84,
|
| 65 |
+
"options": 2,
|
| 66 |
+
"argmax_agrees": true,
|
| 67 |
+
"coreml_argmax": 1,
|
| 68 |
+
"reference_argmax": 1,
|
| 69 |
+
"max_probability_error": 0.00014770030975341797,
|
| 70 |
+
"max_action_probability_error": 0.0,
|
| 71 |
+
"finite": true,
|
| 72 |
+
"padding_masked": true,
|
| 73 |
+
"probability_sum": 0.9998626708984375,
|
| 74 |
+
"median_ms": 5.267270491458476
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"case": "tetris_lane",
|
| 78 |
+
"question": "lane",
|
| 79 |
+
"type": "choice",
|
| 80 |
+
"tokens": 62,
|
| 81 |
+
"options": 4,
|
| 82 |
+
"argmax_agrees": true,
|
| 83 |
+
"coreml_argmax": 1,
|
| 84 |
+
"reference_argmax": 1,
|
| 85 |
+
"max_probability_error": 0.0005466341972351074,
|
| 86 |
+
"max_action_probability_error": 0.0,
|
| 87 |
+
"finite": true,
|
| 88 |
+
"padding_masked": true,
|
| 89 |
+
"probability_sum": 1.0001220703125,
|
| 90 |
+
"median_ms": 5.622666503768414
|
| 91 |
+
},
|
| 92 |
+
{
|
| 93 |
+
"case": "tetris_lane",
|
| 94 |
+
"question": "risk",
|
| 95 |
+
"type": "score",
|
| 96 |
+
"tokens": 64,
|
| 97 |
+
"options": 3,
|
| 98 |
+
"argmax_agrees": true,
|
| 99 |
+
"coreml_argmax": 1,
|
| 100 |
+
"reference_argmax": 1,
|
| 101 |
+
"max_probability_error": 0.00014688074588775635,
|
| 102 |
+
"max_action_probability_error": 0.0,
|
| 103 |
+
"finite": true,
|
| 104 |
+
"padding_masked": true,
|
| 105 |
+
"probability_sum": 1.0,
|
| 106 |
+
"median_ms": 5.3875620069447905
|
| 107 |
+
},
|
| 108 |
+
{
|
| 109 |
+
"case": "form_field",
|
| 110 |
+
"question": "action",
|
| 111 |
+
"type": "choice",
|
| 112 |
+
"tokens": 104,
|
| 113 |
+
"options": 3,
|
| 114 |
+
"argmax_agrees": true,
|
| 115 |
+
"coreml_argmax": 0,
|
| 116 |
+
"reference_argmax": 0,
|
| 117 |
+
"max_probability_error": 0.0021350979804992676,
|
| 118 |
+
"max_action_probability_error": 0.0,
|
| 119 |
+
"finite": true,
|
| 120 |
+
"padding_masked": true,
|
| 121 |
+
"probability_sum": 0.9998779296875,
|
| 122 |
+
"median_ms": 5.193667020648718
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"case": "support_triage",
|
| 126 |
+
"question": "intent",
|
| 127 |
+
"type": "choice",
|
| 128 |
+
"tokens": 62,
|
| 129 |
+
"options": 4,
|
| 130 |
+
"argmax_agrees": true,
|
| 131 |
+
"coreml_argmax": 1,
|
| 132 |
+
"reference_argmax": 1,
|
| 133 |
+
"max_probability_error": 5.608797073364258e-05,
|
| 134 |
+
"max_action_probability_error": 0.0,
|
| 135 |
+
"finite": true,
|
| 136 |
+
"padding_masked": true,
|
| 137 |
+
"probability_sum": 1.0000602006912231,
|
| 138 |
+
"median_ms": 5.3670419729314744
|
| 139 |
+
},
|
| 140 |
+
{
|
| 141 |
+
"case": "support_triage",
|
| 142 |
+
"question": "frustration",
|
| 143 |
+
"type": "score",
|
| 144 |
+
"tokens": 69,
|
| 145 |
+
"options": 3,
|
| 146 |
+
"argmax_agrees": true,
|
| 147 |
+
"coreml_argmax": 1,
|
| 148 |
+
"reference_argmax": 1,
|
| 149 |
+
"max_probability_error": 0.0008405148983001709,
|
| 150 |
+
"max_action_probability_error": 0.0,
|
| 151 |
+
"finite": true,
|
| 152 |
+
"padding_masked": true,
|
| 153 |
+
"probability_sum": 1.00006103515625,
|
| 154 |
+
"median_ms": 5.145104514667764
|
| 155 |
+
},
|
| 156 |
+
{
|
| 157 |
+
"case": "support_triage",
|
| 158 |
+
"question": "churn",
|
| 159 |
+
"type": "noul",
|
| 160 |
+
"tokens": 71,
|
| 161 |
+
"options": 2,
|
| 162 |
+
"argmax_agrees": true,
|
| 163 |
+
"coreml_argmax": 0,
|
| 164 |
+
"reference_argmax": 0,
|
| 165 |
+
"max_probability_error": 3.820657730102539e-05,
|
| 166 |
+
"max_action_probability_error": 0.0,
|
| 167 |
+
"finite": true,
|
| 168 |
+
"padding_masked": true,
|
| 169 |
+
"probability_sum": 0.9999732971191406,
|
| 170 |
+
"median_ms": 5.1331460126675665
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"case": "zh",
|
| 174 |
+
"question": "intent",
|
| 175 |
+
"type": "choice",
|
| 176 |
+
"tokens": 37,
|
| 177 |
+
"options": 3,
|
| 178 |
+
"argmax_agrees": true,
|
| 179 |
+
"coreml_argmax": 0,
|
| 180 |
+
"reference_argmax": 0,
|
| 181 |
+
"max_probability_error": 0.00023287534713745117,
|
| 182 |
+
"max_action_probability_error": 0.0,
|
| 183 |
+
"finite": true,
|
| 184 |
+
"padding_masked": true,
|
| 185 |
+
"probability_sum": 0.9997743368148804,
|
| 186 |
+
"median_ms": 5.142499983776361
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"case": "zh",
|
| 190 |
+
"question": "urgent",
|
| 191 |
+
"type": "noul",
|
| 192 |
+
"tokens": 48,
|
| 193 |
+
"options": 2,
|
| 194 |
+
"argmax_agrees": true,
|
| 195 |
+
"coreml_argmax": 1,
|
| 196 |
+
"reference_argmax": 1,
|
| 197 |
+
"max_probability_error": 0.00011944770812988281,
|
| 198 |
+
"max_action_probability_error": 0.0,
|
| 199 |
+
"finite": true,
|
| 200 |
+
"padding_masked": true,
|
| 201 |
+
"probability_sum": 0.999786376953125,
|
| 202 |
+
"median_ms": 5.46637549996376
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"case": "ja",
|
| 206 |
+
"question": "intent",
|
| 207 |
+
"type": "choice",
|
| 208 |
+
"tokens": 45,
|
| 209 |
+
"options": 3,
|
| 210 |
+
"argmax_agrees": true,
|
| 211 |
+
"coreml_argmax": 0,
|
| 212 |
+
"reference_argmax": 0,
|
| 213 |
+
"max_probability_error": 0.00017899274826049805,
|
| 214 |
+
"max_action_probability_error": 0.0,
|
| 215 |
+
"finite": true,
|
| 216 |
+
"padding_masked": true,
|
| 217 |
+
"probability_sum": 1.000178337097168,
|
| 218 |
+
"median_ms": 5.178166495170444
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"case": "de",
|
| 222 |
+
"question": "intent",
|
| 223 |
+
"type": "choice",
|
| 224 |
+
"tokens": 44,
|
| 225 |
+
"options": 3,
|
| 226 |
+
"argmax_agrees": true,
|
| 227 |
+
"coreml_argmax": 0,
|
| 228 |
+
"reference_argmax": 0,
|
| 229 |
+
"max_probability_error": 7.051229476928711e-05,
|
| 230 |
+
"max_action_probability_error": 0.0,
|
| 231 |
+
"finite": true,
|
| 232 |
+
"padding_masked": true,
|
| 233 |
+
"probability_sum": 0.9998912811279297,
|
| 234 |
+
"median_ms": 5.163750000065193
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"case": "many_options",
|
| 238 |
+
"question": "lane",
|
| 239 |
+
"type": "choice",
|
| 240 |
+
"tokens": 157,
|
| 241 |
+
"options": 20,
|
| 242 |
+
"argmax_agrees": true,
|
| 243 |
+
"coreml_argmax": 3,
|
| 244 |
+
"reference_argmax": 3,
|
| 245 |
+
"max_probability_error": 0.0014334022998809814,
|
| 246 |
+
"max_action_probability_error": 0.0,
|
| 247 |
+
"finite": true,
|
| 248 |
+
"padding_masked": true,
|
| 249 |
+
"probability_sum": 0.9998855590820312,
|
| 250 |
+
"median_ms": 5.150437500560656
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"case": "guard",
|
| 254 |
+
"question": "injection",
|
| 255 |
+
"type": "noul",
|
| 256 |
+
"tokens": 48,
|
| 257 |
+
"options": 2,
|
| 258 |
+
"argmax_agrees": true,
|
| 259 |
+
"coreml_argmax": 1,
|
| 260 |
+
"reference_argmax": 1,
|
| 261 |
+
"max_probability_error": 0.0011126399040222168,
|
| 262 |
+
"max_action_probability_error": 0.0,
|
| 263 |
+
"finite": true,
|
| 264 |
+
"padding_masked": true,
|
| 265 |
+
"probability_sum": 1.0,
|
| 266 |
+
"median_ms": 5.1465209980960935
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"case": "long_state",
|
| 270 |
+
"question": "ready",
|
| 271 |
+
"type": "noul",
|
| 272 |
+
"tokens": 174,
|
| 273 |
+
"options": 2,
|
| 274 |
+
"argmax_agrees": true,
|
| 275 |
+
"coreml_argmax": 1,
|
| 276 |
+
"reference_argmax": 1,
|
| 277 |
+
"max_probability_error": 0.0001442432403564453,
|
| 278 |
+
"max_action_probability_error": 0.0,
|
| 279 |
+
"finite": true,
|
| 280 |
+
"padding_masked": true,
|
| 281 |
+
"probability_sum": 1.0000152587890625,
|
| 282 |
+
"median_ms": 5.122436996316537
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"case": "long_state",
|
| 286 |
+
"question": "owner",
|
| 287 |
+
"type": "choice",
|
| 288 |
+
"tokens": 158,
|
| 289 |
+
"options": 3,
|
| 290 |
+
"argmax_agrees": true,
|
| 291 |
+
"coreml_argmax": 2,
|
| 292 |
+
"reference_argmax": 2,
|
| 293 |
+
"max_probability_error": 0.0009380578994750977,
|
| 294 |
+
"max_action_probability_error": 0.0,
|
| 295 |
+
"finite": true,
|
| 296 |
+
"padding_masked": true,
|
| 297 |
+
"probability_sum": 0.999969482421875,
|
| 298 |
+
"median_ms": 5.1526665047276765
|
| 299 |
+
}
|
| 300 |
+
],
|
| 301 |
+
"passed": true
|
| 302 |
+
},
|
| 303 |
+
"CPU_AND_NE": {
|
| 304 |
+
"compute_units": "CPU_AND_NE",
|
| 305 |
+
"load_seconds": 4.439885583007708,
|
| 306 |
+
"questions": 16,
|
| 307 |
+
"argmax_agreements": 16,
|
| 308 |
+
"max_probability_error": 0.012559235095977783,
|
| 309 |
+
"max_action_probability_error": 0.0,
|
| 310 |
+
"latency_ms": {
|
| 311 |
+
"p50": 9.878875018330291,
|
| 312 |
+
"p95": 10.141766036394984,
|
| 313 |
+
"min": 9.49920795392245,
|
| 314 |
+
"max": 11.666958045680076
|
| 315 |
+
},
|
| 316 |
+
"skipped": [],
|
| 317 |
+
"rows": [
|
| 318 |
+
{
|
| 319 |
+
"case": "tetris_clean",
|
| 320 |
+
"question": "clean",
|
| 321 |
+
"type": "noul",
|
| 322 |
+
"tokens": 80,
|
| 323 |
+
"options": 2,
|
| 324 |
+
"argmax_agrees": true,
|
| 325 |
+
"coreml_argmax": 0,
|
| 326 |
+
"reference_argmax": 0,
|
| 327 |
+
"max_probability_error": 0.0004076361656188965,
|
| 328 |
+
"max_action_probability_error": 0.0,
|
| 329 |
+
"finite": true,
|
| 330 |
+
"padding_masked": true,
|
| 331 |
+
"probability_sum": 1.00018310546875,
|
| 332 |
+
"median_ms": 9.773458004929125
|
| 333 |
+
},
|
| 334 |
+
{
|
| 335 |
+
"case": "tetris_flat",
|
| 336 |
+
"question": "clean",
|
| 337 |
+
"type": "noul",
|
| 338 |
+
"tokens": 84,
|
| 339 |
+
"options": 2,
|
| 340 |
+
"argmax_agrees": true,
|
| 341 |
+
"coreml_argmax": 1,
|
| 342 |
+
"reference_argmax": 1,
|
| 343 |
+
"max_probability_error": 0.00034058094024658203,
|
| 344 |
+
"max_action_probability_error": 0.0,
|
| 345 |
+
"finite": true,
|
| 346 |
+
"padding_masked": true,
|
| 347 |
+
"probability_sum": 1.0003662109375,
|
| 348 |
+
"median_ms": 9.998270979849622
|
| 349 |
+
},
|
| 350 |
+
{
|
| 351 |
+
"case": "tetris_lane",
|
| 352 |
+
"question": "lane",
|
| 353 |
+
"type": "choice",
|
| 354 |
+
"tokens": 62,
|
| 355 |
+
"options": 4,
|
| 356 |
+
"argmax_agrees": true,
|
| 357 |
+
"coreml_argmax": 1,
|
| 358 |
+
"reference_argmax": 1,
|
| 359 |
+
"max_probability_error": 0.0033596158027648926,
|
| 360 |
+
"max_action_probability_error": 0.0,
|
| 361 |
+
"finite": true,
|
| 362 |
+
"padding_masked": true,
|
| 363 |
+
"probability_sum": 0.9998779296875,
|
| 364 |
+
"median_ms": 10.107375041116029
|
| 365 |
+
},
|
| 366 |
+
{
|
| 367 |
+
"case": "tetris_lane",
|
| 368 |
+
"question": "risk",
|
| 369 |
+
"type": "score",
|
| 370 |
+
"tokens": 64,
|
| 371 |
+
"options": 3,
|
| 372 |
+
"argmax_agrees": true,
|
| 373 |
+
"coreml_argmax": 1,
|
| 374 |
+
"reference_argmax": 1,
|
| 375 |
+
"max_probability_error": 0.0014896541833877563,
|
| 376 |
+
"max_action_probability_error": 0.0,
|
| 377 |
+
"finite": true,
|
| 378 |
+
"padding_masked": true,
|
| 379 |
+
"probability_sum": 0.9996337890625,
|
| 380 |
+
"median_ms": 9.907895990181714
|
| 381 |
+
},
|
| 382 |
+
{
|
| 383 |
+
"case": "form_field",
|
| 384 |
+
"question": "action",
|
| 385 |
+
"type": "choice",
|
| 386 |
+
"tokens": 104,
|
| 387 |
+
"options": 3,
|
| 388 |
+
"argmax_agrees": true,
|
| 389 |
+
"coreml_argmax": 0,
|
| 390 |
+
"reference_argmax": 0,
|
| 391 |
+
"max_probability_error": 0.0034949183464050293,
|
| 392 |
+
"max_action_probability_error": 0.0,
|
| 393 |
+
"finite": true,
|
| 394 |
+
"padding_masked": true,
|
| 395 |
+
"probability_sum": 0.9996337890625,
|
| 396 |
+
"median_ms": 9.904021018883213
|
| 397 |
+
},
|
| 398 |
+
{
|
| 399 |
+
"case": "support_triage",
|
| 400 |
+
"question": "intent",
|
| 401 |
+
"type": "choice",
|
| 402 |
+
"tokens": 62,
|
| 403 |
+
"options": 4,
|
| 404 |
+
"argmax_agrees": true,
|
| 405 |
+
"coreml_argmax": 1,
|
| 406 |
+
"reference_argmax": 1,
|
| 407 |
+
"max_probability_error": 0.0004321932792663574,
|
| 408 |
+
"max_action_probability_error": 0.0,
|
| 409 |
+
"finite": true,
|
| 410 |
+
"padding_masked": true,
|
| 411 |
+
"probability_sum": 0.9995682239532471,
|
| 412 |
+
"median_ms": 9.766937495442107
|
| 413 |
+
},
|
| 414 |
+
{
|
| 415 |
+
"case": "support_triage",
|
| 416 |
+
"question": "frustration",
|
| 417 |
+
"type": "score",
|
| 418 |
+
"tokens": 69,
|
| 419 |
+
"options": 3,
|
| 420 |
+
"argmax_agrees": true,
|
| 421 |
+
"coreml_argmax": 1,
|
| 422 |
+
"reference_argmax": 1,
|
| 423 |
+
"max_probability_error": 0.001600891351699829,
|
| 424 |
+
"max_action_probability_error": 0.0,
|
| 425 |
+
"finite": true,
|
| 426 |
+
"padding_masked": true,
|
| 427 |
+
"probability_sum": 0.99981689453125,
|
| 428 |
+
"median_ms": 9.736333013279364
|
| 429 |
+
},
|
| 430 |
+
{
|
| 431 |
+
"case": "support_triage",
|
| 432 |
+
"question": "churn",
|
| 433 |
+
"type": "noul",
|
| 434 |
+
"tokens": 71,
|
| 435 |
+
"options": 2,
|
| 436 |
+
"argmax_agrees": true,
|
| 437 |
+
"coreml_argmax": 0,
|
| 438 |
+
"reference_argmax": 0,
|
| 439 |
+
"max_probability_error": 3.820657730102539e-05,
|
| 440 |
+
"max_action_probability_error": 0.0,
|
| 441 |
+
"finite": true,
|
| 442 |
+
"padding_masked": true,
|
| 443 |
+
"probability_sum": 0.9999771118164062,
|
| 444 |
+
"median_ms": 9.809916489757597
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"case": "zh",
|
| 448 |
+
"question": "intent",
|
| 449 |
+
"type": "choice",
|
| 450 |
+
"tokens": 37,
|
| 451 |
+
"options": 3,
|
| 452 |
+
"argmax_agrees": true,
|
| 453 |
+
"coreml_argmax": 0,
|
| 454 |
+
"reference_argmax": 0,
|
| 455 |
+
"max_probability_error": 0.00023287534713745117,
|
| 456 |
+
"max_action_probability_error": 0.0,
|
| 457 |
+
"finite": true,
|
| 458 |
+
"padding_masked": true,
|
| 459 |
+
"probability_sum": 0.999773383140564,
|
| 460 |
+
"median_ms": 9.79193698731251
|
| 461 |
+
},
|
| 462 |
+
{
|
| 463 |
+
"case": "zh",
|
| 464 |
+
"question": "urgent",
|
| 465 |
+
"type": "noul",
|
| 466 |
+
"tokens": 48,
|
| 467 |
+
"options": 2,
|
| 468 |
+
"argmax_agrees": true,
|
| 469 |
+
"coreml_argmax": 1,
|
| 470 |
+
"reference_argmax": 1,
|
| 471 |
+
"max_probability_error": 0.0004093870520591736,
|
| 472 |
+
"max_action_probability_error": 0.0,
|
| 473 |
+
"finite": true,
|
| 474 |
+
"padding_masked": true,
|
| 475 |
+
"probability_sum": 1.0002899169921875,
|
| 476 |
+
"median_ms": 9.886458021355793
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"case": "ja",
|
| 480 |
+
"question": "intent",
|
| 481 |
+
"type": "choice",
|
| 482 |
+
"tokens": 45,
|
| 483 |
+
"options": 3,
|
| 484 |
+
"argmax_agrees": true,
|
| 485 |
+
"coreml_argmax": 0,
|
| 486 |
+
"reference_argmax": 0,
|
| 487 |
+
"max_probability_error": 0.00017899274826049805,
|
| 488 |
+
"max_action_probability_error": 0.0,
|
| 489 |
+
"finite": true,
|
| 490 |
+
"padding_masked": true,
|
| 491 |
+
"probability_sum": 1.0001769065856934,
|
| 492 |
+
"median_ms": 9.888791479170322
|
| 493 |
+
},
|
| 494 |
+
{
|
| 495 |
+
"case": "de",
|
| 496 |
+
"question": "intent",
|
| 497 |
+
"type": "choice",
|
| 498 |
+
"tokens": 44,
|
| 499 |
+
"options": 3,
|
| 500 |
+
"argmax_agrees": true,
|
| 501 |
+
"coreml_argmax": 0,
|
| 502 |
+
"reference_argmax": 0,
|
| 503 |
+
"max_probability_error": 7.051229476928711e-05,
|
| 504 |
+
"max_action_probability_error": 0.0,
|
| 505 |
+
"finite": true,
|
| 506 |
+
"padding_masked": true,
|
| 507 |
+
"probability_sum": 0.9998779296875,
|
| 508 |
+
"median_ms": 9.867791522992775
|
| 509 |
+
},
|
| 510 |
+
{
|
| 511 |
+
"case": "many_options",
|
| 512 |
+
"question": "lane",
|
| 513 |
+
"type": "choice",
|
| 514 |
+
"tokens": 157,
|
| 515 |
+
"options": 20,
|
| 516 |
+
"argmax_agrees": true,
|
| 517 |
+
"coreml_argmax": 3,
|
| 518 |
+
"reference_argmax": 3,
|
| 519 |
+
"max_probability_error": 0.001290060579776764,
|
| 520 |
+
"max_action_probability_error": 0.0,
|
| 521 |
+
"finite": true,
|
| 522 |
+
"padding_masked": true,
|
| 523 |
+
"probability_sum": 0.9998397827148438,
|
| 524 |
+
"median_ms": 9.76256252033636
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"case": "guard",
|
| 528 |
+
"question": "injection",
|
| 529 |
+
"type": "noul",
|
| 530 |
+
"tokens": 48,
|
| 531 |
+
"options": 2,
|
| 532 |
+
"argmax_agrees": true,
|
| 533 |
+
"coreml_argmax": 1,
|
| 534 |
+
"reference_argmax": 1,
|
| 535 |
+
"max_probability_error": 0.012559235095977783,
|
| 536 |
+
"max_action_probability_error": 0.0,
|
| 537 |
+
"finite": true,
|
| 538 |
+
"padding_masked": true,
|
| 539 |
+
"probability_sum": 1.000244140625,
|
| 540 |
+
"median_ms": 9.754520520800725
|
| 541 |
+
},
|
| 542 |
+
{
|
| 543 |
+
"case": "long_state",
|
| 544 |
+
"question": "ready",
|
| 545 |
+
"type": "noul",
|
| 546 |
+
"tokens": 174,
|
| 547 |
+
"options": 2,
|
| 548 |
+
"argmax_agrees": true,
|
| 549 |
+
"coreml_argmax": 1,
|
| 550 |
+
"reference_argmax": 1,
|
| 551 |
+
"max_probability_error": 0.0003440380096435547,
|
| 552 |
+
"max_action_probability_error": 0.0,
|
| 553 |
+
"finite": true,
|
| 554 |
+
"padding_masked": true,
|
| 555 |
+
"probability_sum": 0.999664306640625,
|
| 556 |
+
"median_ms": 9.926104015903547
|
| 557 |
+
},
|
| 558 |
+
{
|
| 559 |
+
"case": "long_state",
|
| 560 |
+
"question": "owner",
|
| 561 |
+
"type": "choice",
|
| 562 |
+
"tokens": 158,
|
| 563 |
+
"options": 3,
|
| 564 |
+
"argmax_agrees": true,
|
| 565 |
+
"coreml_argmax": 2,
|
| 566 |
+
"reference_argmax": 2,
|
| 567 |
+
"max_probability_error": 0.0005267858505249023,
|
| 568 |
+
"max_action_probability_error": 0.0,
|
| 569 |
+
"finite": true,
|
| 570 |
+
"padding_masked": true,
|
| 571 |
+
"probability_sum": 0.999908447265625,
|
| 572 |
+
"median_ms": 9.970562008675188
|
| 573 |
+
}
|
| 574 |
+
],
|
| 575 |
+
"passed": true
|
| 576 |
+
}
|
| 577 |
+
},
|
| 578 |
+
"passed": true
|
| 579 |
+
}
|
reports/verification-multilingual-L512.json
ADDED
|
@@ -0,0 +1,579 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"package": "laya_multilingual_fp16_L512_options32.mlpackage",
|
| 3 |
+
"package_files_sha256": {
|
| 4 |
+
"Data/com.apple.CoreML/model.mlmodel": "a9efa197dc468765e5074399dd9cbad131052e8f241654aefeea390d1faa57ac",
|
| 5 |
+
"Data/com.apple.CoreML/weights/weight.bin": "b232c8f3e17396c1077b42633c3098daa5bf94b27668accd22042ee7a49e0b49",
|
| 6 |
+
"Manifest.json": "b75c4c456a85d54c38eda45ebe88560e208079d0eec16d568a39abebfbc6be04"
|
| 7 |
+
},
|
| 8 |
+
"source_repo": "convaiinnovations/laya",
|
| 9 |
+
"source_revision": "1c5edc17a7acd8701df6fc341c0d179f1c62c982",
|
| 10 |
+
"variant": "multilingual",
|
| 11 |
+
"length": 512,
|
| 12 |
+
"max_options": 32,
|
| 13 |
+
"environment": {
|
| 14 |
+
"chip": "Apple M5 Pro",
|
| 15 |
+
"os": "27.0",
|
| 16 |
+
"python": "3.12.13",
|
| 17 |
+
"torch": "2.7.0",
|
| 18 |
+
"coremltools": "9.0"
|
| 19 |
+
},
|
| 20 |
+
"gates": {
|
| 21 |
+
"argmax_agreement": "100%",
|
| 22 |
+
"max_probability_error": 0.02,
|
| 23 |
+
"max_action_probability_error": 0.02,
|
| 24 |
+
"finite": true,
|
| 25 |
+
"padding_masked": true
|
| 26 |
+
},
|
| 27 |
+
"cases_sha256": "4de4e34309c421913524209b62bb7c8a0a232928bd228ec12e884347ce65eeed",
|
| 28 |
+
"runs": {
|
| 29 |
+
"ALL": {
|
| 30 |
+
"compute_units": "ALL",
|
| 31 |
+
"load_seconds": 2.4458846659981646,
|
| 32 |
+
"questions": 16,
|
| 33 |
+
"argmax_agreements": 16,
|
| 34 |
+
"max_probability_error": 0.0021350979804992676,
|
| 35 |
+
"max_action_probability_error": 0.0,
|
| 36 |
+
"latency_ms": {
|
| 37 |
+
"p50": 9.00429199100472,
|
| 38 |
+
"p95": 9.368602131144144,
|
| 39 |
+
"min": 8.766749990172684,
|
| 40 |
+
"max": 13.464874995406717
|
| 41 |
+
},
|
| 42 |
+
"skipped": [],
|
| 43 |
+
"rows": [
|
| 44 |
+
{
|
| 45 |
+
"case": "tetris_clean",
|
| 46 |
+
"question": "clean",
|
| 47 |
+
"type": "noul",
|
| 48 |
+
"tokens": 80,
|
| 49 |
+
"options": 2,
|
| 50 |
+
"argmax_agrees": true,
|
| 51 |
+
"coreml_argmax": 0,
|
| 52 |
+
"reference_argmax": 0,
|
| 53 |
+
"max_probability_error": 0.0001417025923728943,
|
| 54 |
+
"max_action_probability_error": 0.0,
|
| 55 |
+
"finite": true,
|
| 56 |
+
"padding_masked": true,
|
| 57 |
+
"probability_sum": 1.00006103515625,
|
| 58 |
+
"median_ms": 9.15208351216279
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"case": "tetris_flat",
|
| 62 |
+
"question": "clean",
|
| 63 |
+
"type": "noul",
|
| 64 |
+
"tokens": 84,
|
| 65 |
+
"options": 2,
|
| 66 |
+
"argmax_agrees": true,
|
| 67 |
+
"coreml_argmax": 1,
|
| 68 |
+
"reference_argmax": 1,
|
| 69 |
+
"max_probability_error": 0.00014770030975341797,
|
| 70 |
+
"max_action_probability_error": 0.0,
|
| 71 |
+
"finite": true,
|
| 72 |
+
"padding_masked": true,
|
| 73 |
+
"probability_sum": 0.9998626708984375,
|
| 74 |
+
"median_ms": 9.204000001773238
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"case": "tetris_lane",
|
| 78 |
+
"question": "lane",
|
| 79 |
+
"type": "choice",
|
| 80 |
+
"tokens": 62,
|
| 81 |
+
"options": 4,
|
| 82 |
+
"argmax_agrees": true,
|
| 83 |
+
"coreml_argmax": 1,
|
| 84 |
+
"reference_argmax": 1,
|
| 85 |
+
"max_probability_error": 0.0005466341972351074,
|
| 86 |
+
"max_action_probability_error": 0.0,
|
| 87 |
+
"finite": true,
|
| 88 |
+
"padding_masked": true,
|
| 89 |
+
"probability_sum": 1.0001220703125,
|
| 90 |
+
"median_ms": 9.063625504495576
|
| 91 |
+
},
|
| 92 |
+
{
|
| 93 |
+
"case": "tetris_lane",
|
| 94 |
+
"question": "risk",
|
| 95 |
+
"type": "score",
|
| 96 |
+
"tokens": 64,
|
| 97 |
+
"options": 3,
|
| 98 |
+
"argmax_agrees": true,
|
| 99 |
+
"coreml_argmax": 1,
|
| 100 |
+
"reference_argmax": 1,
|
| 101 |
+
"max_probability_error": 0.00014688074588775635,
|
| 102 |
+
"max_action_probability_error": 0.0,
|
| 103 |
+
"finite": true,
|
| 104 |
+
"padding_masked": true,
|
| 105 |
+
"probability_sum": 1.0,
|
| 106 |
+
"median_ms": 9.09689551917836
|
| 107 |
+
},
|
| 108 |
+
{
|
| 109 |
+
"case": "form_field",
|
| 110 |
+
"question": "action",
|
| 111 |
+
"type": "choice",
|
| 112 |
+
"tokens": 104,
|
| 113 |
+
"options": 3,
|
| 114 |
+
"argmax_agrees": true,
|
| 115 |
+
"coreml_argmax": 0,
|
| 116 |
+
"reference_argmax": 0,
|
| 117 |
+
"max_probability_error": 0.0021350979804992676,
|
| 118 |
+
"max_action_probability_error": 0.0,
|
| 119 |
+
"finite": true,
|
| 120 |
+
"padding_masked": true,
|
| 121 |
+
"probability_sum": 0.9998779296875,
|
| 122 |
+
"median_ms": 9.05550000607036
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"case": "support_triage",
|
| 126 |
+
"question": "intent",
|
| 127 |
+
"type": "choice",
|
| 128 |
+
"tokens": 62,
|
| 129 |
+
"options": 4,
|
| 130 |
+
"argmax_agrees": true,
|
| 131 |
+
"coreml_argmax": 1,
|
| 132 |
+
"reference_argmax": 1,
|
| 133 |
+
"max_probability_error": 5.608797073364258e-05,
|
| 134 |
+
"max_action_probability_error": 0.0,
|
| 135 |
+
"finite": true,
|
| 136 |
+
"padding_masked": true,
|
| 137 |
+
"probability_sum": 1.0000602006912231,
|
| 138 |
+
"median_ms": 8.988271001726389
|
| 139 |
+
},
|
| 140 |
+
{
|
| 141 |
+
"case": "support_triage",
|
| 142 |
+
"question": "frustration",
|
| 143 |
+
"type": "score",
|
| 144 |
+
"tokens": 69,
|
| 145 |
+
"options": 3,
|
| 146 |
+
"argmax_agrees": true,
|
| 147 |
+
"coreml_argmax": 1,
|
| 148 |
+
"reference_argmax": 1,
|
| 149 |
+
"max_probability_error": 0.0008405148983001709,
|
| 150 |
+
"max_action_probability_error": 0.0,
|
| 151 |
+
"finite": true,
|
| 152 |
+
"padding_masked": true,
|
| 153 |
+
"probability_sum": 1.00006103515625,
|
| 154 |
+
"median_ms": 9.089228988159448
|
| 155 |
+
},
|
| 156 |
+
{
|
| 157 |
+
"case": "support_triage",
|
| 158 |
+
"question": "churn",
|
| 159 |
+
"type": "noul",
|
| 160 |
+
"tokens": 71,
|
| 161 |
+
"options": 2,
|
| 162 |
+
"argmax_agrees": true,
|
| 163 |
+
"coreml_argmax": 0,
|
| 164 |
+
"reference_argmax": 0,
|
| 165 |
+
"max_probability_error": 3.820657730102539e-05,
|
| 166 |
+
"max_action_probability_error": 0.0,
|
| 167 |
+
"finite": true,
|
| 168 |
+
"padding_masked": true,
|
| 169 |
+
"probability_sum": 0.9999732971191406,
|
| 170 |
+
"median_ms": 8.93558349343948
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"case": "zh",
|
| 174 |
+
"question": "intent",
|
| 175 |
+
"type": "choice",
|
| 176 |
+
"tokens": 37,
|
| 177 |
+
"options": 3,
|
| 178 |
+
"argmax_agrees": true,
|
| 179 |
+
"coreml_argmax": 0,
|
| 180 |
+
"reference_argmax": 0,
|
| 181 |
+
"max_probability_error": 0.00023287534713745117,
|
| 182 |
+
"max_action_probability_error": 0.0,
|
| 183 |
+
"finite": true,
|
| 184 |
+
"padding_masked": true,
|
| 185 |
+
"probability_sum": 0.9997743368148804,
|
| 186 |
+
"median_ms": 8.88558299629949
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"case": "zh",
|
| 190 |
+
"question": "urgent",
|
| 191 |
+
"type": "noul",
|
| 192 |
+
"tokens": 48,
|
| 193 |
+
"options": 2,
|
| 194 |
+
"argmax_agrees": true,
|
| 195 |
+
"coreml_argmax": 1,
|
| 196 |
+
"reference_argmax": 1,
|
| 197 |
+
"max_probability_error": 0.00011944770812988281,
|
| 198 |
+
"max_action_probability_error": 0.0,
|
| 199 |
+
"finite": true,
|
| 200 |
+
"padding_masked": true,
|
| 201 |
+
"probability_sum": 0.999786376953125,
|
| 202 |
+
"median_ms": 8.887521020369604
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"case": "ja",
|
| 206 |
+
"question": "intent",
|
| 207 |
+
"type": "choice",
|
| 208 |
+
"tokens": 45,
|
| 209 |
+
"options": 3,
|
| 210 |
+
"argmax_agrees": true,
|
| 211 |
+
"coreml_argmax": 0,
|
| 212 |
+
"reference_argmax": 0,
|
| 213 |
+
"max_probability_error": 0.00017899274826049805,
|
| 214 |
+
"max_action_probability_error": 0.0,
|
| 215 |
+
"finite": true,
|
| 216 |
+
"padding_masked": true,
|
| 217 |
+
"probability_sum": 1.000178337097168,
|
| 218 |
+
"median_ms": 9.004458494018763
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"case": "de",
|
| 222 |
+
"question": "intent",
|
| 223 |
+
"type": "choice",
|
| 224 |
+
"tokens": 44,
|
| 225 |
+
"options": 3,
|
| 226 |
+
"argmax_agrees": true,
|
| 227 |
+
"coreml_argmax": 0,
|
| 228 |
+
"reference_argmax": 0,
|
| 229 |
+
"max_probability_error": 7.051229476928711e-05,
|
| 230 |
+
"max_action_probability_error": 0.0,
|
| 231 |
+
"finite": true,
|
| 232 |
+
"padding_masked": true,
|
| 233 |
+
"probability_sum": 0.9998912811279297,
|
| 234 |
+
"median_ms": 8.942167012719437
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"case": "many_options",
|
| 238 |
+
"question": "lane",
|
| 239 |
+
"type": "choice",
|
| 240 |
+
"tokens": 157,
|
| 241 |
+
"options": 20,
|
| 242 |
+
"argmax_agrees": true,
|
| 243 |
+
"coreml_argmax": 3,
|
| 244 |
+
"reference_argmax": 3,
|
| 245 |
+
"max_probability_error": 0.0014334022998809814,
|
| 246 |
+
"max_action_probability_error": 0.0,
|
| 247 |
+
"finite": true,
|
| 248 |
+
"padding_masked": true,
|
| 249 |
+
"probability_sum": 0.9998855590820312,
|
| 250 |
+
"median_ms": 9.010562498588115
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"case": "guard",
|
| 254 |
+
"question": "injection",
|
| 255 |
+
"type": "noul",
|
| 256 |
+
"tokens": 48,
|
| 257 |
+
"options": 2,
|
| 258 |
+
"argmax_agrees": true,
|
| 259 |
+
"coreml_argmax": 1,
|
| 260 |
+
"reference_argmax": 1,
|
| 261 |
+
"max_probability_error": 0.0011126399040222168,
|
| 262 |
+
"max_action_probability_error": 0.0,
|
| 263 |
+
"finite": true,
|
| 264 |
+
"padding_masked": true,
|
| 265 |
+
"probability_sum": 1.0,
|
| 266 |
+
"median_ms": 8.911187498597428
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"case": "long_state",
|
| 270 |
+
"question": "ready",
|
| 271 |
+
"type": "noul",
|
| 272 |
+
"tokens": 174,
|
| 273 |
+
"options": 2,
|
| 274 |
+
"argmax_agrees": true,
|
| 275 |
+
"coreml_argmax": 1,
|
| 276 |
+
"reference_argmax": 1,
|
| 277 |
+
"max_probability_error": 0.0001442432403564453,
|
| 278 |
+
"max_action_probability_error": 0.0,
|
| 279 |
+
"finite": true,
|
| 280 |
+
"padding_masked": true,
|
| 281 |
+
"probability_sum": 1.0000152587890625,
|
| 282 |
+
"median_ms": 8.846083976095542
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"case": "long_state",
|
| 286 |
+
"question": "owner",
|
| 287 |
+
"type": "choice",
|
| 288 |
+
"tokens": 158,
|
| 289 |
+
"options": 3,
|
| 290 |
+
"argmax_agrees": true,
|
| 291 |
+
"coreml_argmax": 2,
|
| 292 |
+
"reference_argmax": 2,
|
| 293 |
+
"max_probability_error": 0.0009380578994750977,
|
| 294 |
+
"max_action_probability_error": 0.0,
|
| 295 |
+
"finite": true,
|
| 296 |
+
"padding_masked": true,
|
| 297 |
+
"probability_sum": 0.999969482421875,
|
| 298 |
+
"median_ms": 9.015603980515152
|
| 299 |
+
}
|
| 300 |
+
],
|
| 301 |
+
"passed": true
|
| 302 |
+
},
|
| 303 |
+
"CPU_AND_NE": {
|
| 304 |
+
"compute_units": "CPU_AND_NE",
|
| 305 |
+
"load_seconds": 4.784726334037259,
|
| 306 |
+
"questions": 16,
|
| 307 |
+
"argmax_agreements": 16,
|
| 308 |
+
"max_probability_error": 0.012559235095977783,
|
| 309 |
+
"max_action_probability_error": 0.0,
|
| 310 |
+
"latency_ms": {
|
| 311 |
+
"p50": 27.451770496554673,
|
| 312 |
+
"p95": 27.900547938770615,
|
| 313 |
+
"min": 27.101749961730093,
|
| 314 |
+
"max": 30.147792014759034
|
| 315 |
+
},
|
| 316 |
+
"skipped": [],
|
| 317 |
+
"rows": [
|
| 318 |
+
{
|
| 319 |
+
"case": "tetris_clean",
|
| 320 |
+
"question": "clean",
|
| 321 |
+
"type": "noul",
|
| 322 |
+
"tokens": 80,
|
| 323 |
+
"options": 2,
|
| 324 |
+
"argmax_agrees": true,
|
| 325 |
+
"coreml_argmax": 0,
|
| 326 |
+
"reference_argmax": 0,
|
| 327 |
+
"max_probability_error": 0.0004076361656188965,
|
| 328 |
+
"max_action_probability_error": 0.0,
|
| 329 |
+
"finite": true,
|
| 330 |
+
"padding_masked": true,
|
| 331 |
+
"probability_sum": 1.00018310546875,
|
| 332 |
+
"median_ms": 27.682187472237274
|
| 333 |
+
},
|
| 334 |
+
{
|
| 335 |
+
"case": "tetris_flat",
|
| 336 |
+
"question": "clean",
|
| 337 |
+
"type": "noul",
|
| 338 |
+
"tokens": 84,
|
| 339 |
+
"options": 2,
|
| 340 |
+
"argmax_agrees": true,
|
| 341 |
+
"coreml_argmax": 1,
|
| 342 |
+
"reference_argmax": 1,
|
| 343 |
+
"max_probability_error": 0.00034058094024658203,
|
| 344 |
+
"max_action_probability_error": 0.0,
|
| 345 |
+
"finite": true,
|
| 346 |
+
"padding_masked": true,
|
| 347 |
+
"probability_sum": 1.0003662109375,
|
| 348 |
+
"median_ms": 27.74883297388442
|
| 349 |
+
},
|
| 350 |
+
{
|
| 351 |
+
"case": "tetris_lane",
|
| 352 |
+
"question": "lane",
|
| 353 |
+
"type": "choice",
|
| 354 |
+
"tokens": 62,
|
| 355 |
+
"options": 4,
|
| 356 |
+
"argmax_agrees": true,
|
| 357 |
+
"coreml_argmax": 1,
|
| 358 |
+
"reference_argmax": 1,
|
| 359 |
+
"max_probability_error": 0.0033596158027648926,
|
| 360 |
+
"max_action_probability_error": 0.0,
|
| 361 |
+
"finite": true,
|
| 362 |
+
"padding_masked": true,
|
| 363 |
+
"probability_sum": 0.9998779296875,
|
| 364 |
+
"median_ms": 27.54443750018254
|
| 365 |
+
},
|
| 366 |
+
{
|
| 367 |
+
"case": "tetris_lane",
|
| 368 |
+
"question": "risk",
|
| 369 |
+
"type": "score",
|
| 370 |
+
"tokens": 64,
|
| 371 |
+
"options": 3,
|
| 372 |
+
"argmax_agrees": true,
|
| 373 |
+
"coreml_argmax": 1,
|
| 374 |
+
"reference_argmax": 1,
|
| 375 |
+
"max_probability_error": 0.0014896541833877563,
|
| 376 |
+
"max_action_probability_error": 0.0,
|
| 377 |
+
"finite": true,
|
| 378 |
+
"padding_masked": true,
|
| 379 |
+
"probability_sum": 0.9996337890625,
|
| 380 |
+
"median_ms": 27.727770997444168
|
| 381 |
+
},
|
| 382 |
+
{
|
| 383 |
+
"case": "form_field",
|
| 384 |
+
"question": "action",
|
| 385 |
+
"type": "choice",
|
| 386 |
+
"tokens": 104,
|
| 387 |
+
"options": 3,
|
| 388 |
+
"argmax_agrees": true,
|
| 389 |
+
"coreml_argmax": 0,
|
| 390 |
+
"reference_argmax": 0,
|
| 391 |
+
"max_probability_error": 0.0034949183464050293,
|
| 392 |
+
"max_action_probability_error": 0.0,
|
| 393 |
+
"finite": true,
|
| 394 |
+
"padding_masked": true,
|
| 395 |
+
"probability_sum": 0.9996337890625,
|
| 396 |
+
"median_ms": 27.766666520619765
|
| 397 |
+
},
|
| 398 |
+
{
|
| 399 |
+
"case": "support_triage",
|
| 400 |
+
"question": "intent",
|
| 401 |
+
"type": "choice",
|
| 402 |
+
"tokens": 62,
|
| 403 |
+
"options": 4,
|
| 404 |
+
"argmax_agrees": true,
|
| 405 |
+
"coreml_argmax": 1,
|
| 406 |
+
"reference_argmax": 1,
|
| 407 |
+
"max_probability_error": 0.0004321932792663574,
|
| 408 |
+
"max_action_probability_error": 0.0,
|
| 409 |
+
"finite": true,
|
| 410 |
+
"padding_masked": true,
|
| 411 |
+
"probability_sum": 0.9995682239532471,
|
| 412 |
+
"median_ms": 27.386936999391764
|
| 413 |
+
},
|
| 414 |
+
{
|
| 415 |
+
"case": "support_triage",
|
| 416 |
+
"question": "frustration",
|
| 417 |
+
"type": "score",
|
| 418 |
+
"tokens": 69,
|
| 419 |
+
"options": 3,
|
| 420 |
+
"argmax_agrees": true,
|
| 421 |
+
"coreml_argmax": 1,
|
| 422 |
+
"reference_argmax": 1,
|
| 423 |
+
"max_probability_error": 0.001600891351699829,
|
| 424 |
+
"max_action_probability_error": 0.0,
|
| 425 |
+
"finite": true,
|
| 426 |
+
"padding_masked": true,
|
| 427 |
+
"probability_sum": 0.99981689453125,
|
| 428 |
+
"median_ms": 27.639291482046247
|
| 429 |
+
},
|
| 430 |
+
{
|
| 431 |
+
"case": "support_triage",
|
| 432 |
+
"question": "churn",
|
| 433 |
+
"type": "noul",
|
| 434 |
+
"tokens": 71,
|
| 435 |
+
"options": 2,
|
| 436 |
+
"argmax_agrees": true,
|
| 437 |
+
"coreml_argmax": 0,
|
| 438 |
+
"reference_argmax": 0,
|
| 439 |
+
"max_probability_error": 3.820657730102539e-05,
|
| 440 |
+
"max_action_probability_error": 0.0,
|
| 441 |
+
"finite": true,
|
| 442 |
+
"padding_masked": true,
|
| 443 |
+
"probability_sum": 0.9999771118164062,
|
| 444 |
+
"median_ms": 27.325041504809633
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"case": "zh",
|
| 448 |
+
"question": "intent",
|
| 449 |
+
"type": "choice",
|
| 450 |
+
"tokens": 37,
|
| 451 |
+
"options": 3,
|
| 452 |
+
"argmax_agrees": true,
|
| 453 |
+
"coreml_argmax": 0,
|
| 454 |
+
"reference_argmax": 0,
|
| 455 |
+
"max_probability_error": 0.00023287534713745117,
|
| 456 |
+
"max_action_probability_error": 0.0,
|
| 457 |
+
"finite": true,
|
| 458 |
+
"padding_masked": true,
|
| 459 |
+
"probability_sum": 0.999773383140564,
|
| 460 |
+
"median_ms": 27.280833513941616
|
| 461 |
+
},
|
| 462 |
+
{
|
| 463 |
+
"case": "zh",
|
| 464 |
+
"question": "urgent",
|
| 465 |
+
"type": "noul",
|
| 466 |
+
"tokens": 48,
|
| 467 |
+
"options": 2,
|
| 468 |
+
"argmax_agrees": true,
|
| 469 |
+
"coreml_argmax": 1,
|
| 470 |
+
"reference_argmax": 1,
|
| 471 |
+
"max_probability_error": 0.0004093870520591736,
|
| 472 |
+
"max_action_probability_error": 0.0,
|
| 473 |
+
"finite": true,
|
| 474 |
+
"padding_masked": true,
|
| 475 |
+
"probability_sum": 1.0002899169921875,
|
| 476 |
+
"median_ms": 27.29837500373833
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"case": "ja",
|
| 480 |
+
"question": "intent",
|
| 481 |
+
"type": "choice",
|
| 482 |
+
"tokens": 45,
|
| 483 |
+
"options": 3,
|
| 484 |
+
"argmax_agrees": true,
|
| 485 |
+
"coreml_argmax": 0,
|
| 486 |
+
"reference_argmax": 0,
|
| 487 |
+
"max_probability_error": 0.00017899274826049805,
|
| 488 |
+
"max_action_probability_error": 0.0,
|
| 489 |
+
"finite": true,
|
| 490 |
+
"padding_masked": true,
|
| 491 |
+
"probability_sum": 1.0001769065856934,
|
| 492 |
+
"median_ms": 27.201541524846107
|
| 493 |
+
},
|
| 494 |
+
{
|
| 495 |
+
"case": "de",
|
| 496 |
+
"question": "intent",
|
| 497 |
+
"type": "choice",
|
| 498 |
+
"tokens": 44,
|
| 499 |
+
"options": 3,
|
| 500 |
+
"argmax_agrees": true,
|
| 501 |
+
"coreml_argmax": 0,
|
| 502 |
+
"reference_argmax": 0,
|
| 503 |
+
"max_probability_error": 7.051229476928711e-05,
|
| 504 |
+
"max_action_probability_error": 0.0,
|
| 505 |
+
"finite": true,
|
| 506 |
+
"padding_masked": true,
|
| 507 |
+
"probability_sum": 0.9998779296875,
|
| 508 |
+
"median_ms": 27.596458472544327
|
| 509 |
+
},
|
| 510 |
+
{
|
| 511 |
+
"case": "many_options",
|
| 512 |
+
"question": "lane",
|
| 513 |
+
"type": "choice",
|
| 514 |
+
"tokens": 157,
|
| 515 |
+
"options": 20,
|
| 516 |
+
"argmax_agrees": true,
|
| 517 |
+
"coreml_argmax": 3,
|
| 518 |
+
"reference_argmax": 3,
|
| 519 |
+
"max_probability_error": 0.001290060579776764,
|
| 520 |
+
"max_action_probability_error": 0.0,
|
| 521 |
+
"finite": true,
|
| 522 |
+
"padding_masked": true,
|
| 523 |
+
"probability_sum": 0.9998397827148438,
|
| 524 |
+
"median_ms": 27.25768749951385
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"case": "guard",
|
| 528 |
+
"question": "injection",
|
| 529 |
+
"type": "noul",
|
| 530 |
+
"tokens": 48,
|
| 531 |
+
"options": 2,
|
| 532 |
+
"argmax_agrees": true,
|
| 533 |
+
"coreml_argmax": 1,
|
| 534 |
+
"reference_argmax": 1,
|
| 535 |
+
"max_probability_error": 0.012559235095977783,
|
| 536 |
+
"max_action_probability_error": 0.0,
|
| 537 |
+
"finite": true,
|
| 538 |
+
"padding_masked": true,
|
| 539 |
+
"probability_sum": 1.000244140625,
|
| 540 |
+
"median_ms": 27.227271028095856
|
| 541 |
+
},
|
| 542 |
+
{
|
| 543 |
+
"case": "long_state",
|
| 544 |
+
"question": "ready",
|
| 545 |
+
"type": "noul",
|
| 546 |
+
"tokens": 174,
|
| 547 |
+
"options": 2,
|
| 548 |
+
"argmax_agrees": true,
|
| 549 |
+
"coreml_argmax": 1,
|
| 550 |
+
"reference_argmax": 1,
|
| 551 |
+
"max_probability_error": 0.0003440380096435547,
|
| 552 |
+
"max_action_probability_error": 0.0,
|
| 553 |
+
"finite": true,
|
| 554 |
+
"padding_masked": true,
|
| 555 |
+
"probability_sum": 0.999664306640625,
|
| 556 |
+
"median_ms": 27.25012498558499
|
| 557 |
+
},
|
| 558 |
+
{
|
| 559 |
+
"case": "long_state",
|
| 560 |
+
"question": "owner",
|
| 561 |
+
"type": "choice",
|
| 562 |
+
"tokens": 158,
|
| 563 |
+
"options": 3,
|
| 564 |
+
"argmax_agrees": true,
|
| 565 |
+
"coreml_argmax": 2,
|
| 566 |
+
"reference_argmax": 2,
|
| 567 |
+
"max_probability_error": 0.0005267858505249023,
|
| 568 |
+
"max_action_probability_error": 0.0,
|
| 569 |
+
"finite": true,
|
| 570 |
+
"padding_masked": true,
|
| 571 |
+
"probability_sum": 0.999908447265625,
|
| 572 |
+
"median_ms": 27.342583489371464
|
| 573 |
+
}
|
| 574 |
+
],
|
| 575 |
+
"passed": true
|
| 576 |
+
}
|
| 577 |
+
},
|
| 578 |
+
"passed": true
|
| 579 |
+
}
|