Unverified Commit dd2d91b0 authored by Daniël de Kok's avatar Daniël de Kok Committed by GitHub
Browse files

Idefics2: sync added image tokens with transformers (#2080)

Before this change, the number of reserved image tokens was not the
same as the number of images. Fixes #2029.

While at it, also remove all the image token handling duplication
in `prepare_input`.
parent b53b21c6
......@@ -3832,6 +3832,7 @@ dependencies = [
"hf-hub",
"image",
"init-tracing-opentelemetry",
"itertools 0.10.5",
"jsonschema",
"metrics 0.21.1",
"metrics-exporter-prometheus",
......
......@@ -37,7 +37,7 @@
},
{
"id": 32001,
"logprob": -19.484375,
"logprob": -19.46875,
"text": "<image>"
},
{
......@@ -57,7 +57,7 @@
},
{
"id": 32001,
"logprob": -20.234375,
"logprob": -20.21875,
"text": "<image>"
},
{
......@@ -67,12 +67,12 @@
},
{
"id": 32001,
"logprob": -19.828125,
"logprob": -19.84375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -23.25,
"logprob": -23.265625,
"text": "<image>"
},
{
......@@ -97,12 +97,12 @@
},
{
"id": 32001,
"logprob": -21.296875,
"logprob": -21.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.015625,
"logprob": -21.03125,
"text": "<image>"
},
{
......@@ -112,12 +112,12 @@
},
{
"id": 32001,
"logprob": -16.015625,
"logprob": -16.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.0625,
"logprob": -19.046875,
"text": "<image>"
},
{
......@@ -127,22 +127,22 @@
},
{
"id": 32001,
"logprob": -23.625,
"logprob": -23.609375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.40625,
"logprob": -20.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.421875,
"logprob": -20.4375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.84375,
"logprob": -20.875,
"text": "<image>"
},
{
......@@ -152,12 +152,12 @@
},
{
"id": 32001,
"logprob": -19.984375,
"logprob": -19.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.21875,
"logprob": -18.234375,
"text": "<image>"
},
{
......@@ -167,7 +167,7 @@
},
{
"id": 32001,
"logprob": -21.203125,
"logprob": -21.1875,
"text": "<image>"
},
{
......@@ -182,7 +182,7 @@
},
{
"id": 32001,
"logprob": -18.984375,
"logprob": -19.03125,
"text": "<image>"
},
{
......@@ -197,17 +197,17 @@
},
{
"id": 32001,
"logprob": -18.0,
"logprob": -17.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.828125,
"logprob": -18.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.9375,
"logprob": -17.921875,
"text": "<image>"
},
{
......@@ -217,12 +217,12 @@
},
{
"id": 32001,
"logprob": -18.640625,
"logprob": -18.65625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.125,
"logprob": -20.140625,
"text": "<image>"
},
{
......@@ -242,12 +242,12 @@
},
{
"id": 32001,
"logprob": -23.21875,
"logprob": -23.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.4375,
"logprob": -17.421875,
"text": "<image>"
},
{
......@@ -257,7 +257,7 @@
},
{
"id": 32001,
"logprob": -23.015625,
"logprob": -23.0,
"text": "<image>"
},
{
......@@ -292,32 +292,32 @@
},
{
"id": 32001,
"logprob": -18.34375,
"logprob": -18.328125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.140625,
"logprob": -17.125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.671875,
"logprob": -18.65625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.875,
"logprob": -18.890625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.0,
"logprob": -15.9921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.1875,
"logprob": -18.15625,
"text": "<image>"
},
{
......@@ -337,47 +337,52 @@
},
{
"id": 32001,
"logprob": -18.71875,
"logprob": -18.703125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.9375,
"logprob": -17.921875,
"text": "<image>"
},
{
"id": 32000,
"logprob": -3.015625,
"text": "<fake_token_around_image>"
},
{
"id": 32001,
"logprob": -17.921875,
"logprob": -22.09375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.125,
"logprob": -18.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.296875,
"logprob": -20.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.25,
"logprob": -17.125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.96875,
"logprob": -17.796875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.546875,
"logprob": -19.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.921875,
"logprob": -19.65625,
"text": "<image>"
},
{
......@@ -387,212 +392,212 @@
},
{
"id": 32001,
"logprob": -19.546875,
"logprob": -19.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.1875,
"logprob": -20.046875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.953125,
"logprob": -20.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.828125,
"logprob": -15.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.375,
"logprob": -16.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.171875,
"logprob": -19.953125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.03125,
"logprob": -18.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.71875,
"logprob": -21.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.65625,
"logprob": -20.515625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.453125,
"logprob": -19.171875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.484375,
"logprob": -19.296875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.875,
"logprob": -16.703125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.65625,
"logprob": -20.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.296875,
"logprob": -21.125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.546875,
"logprob": -20.40625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.78125,
"logprob": -20.71875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.96875,
"logprob": -17.78125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.5,
"logprob": -20.328125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.4140625,
"logprob": -15.2734375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.046875,
"logprob": -18.84375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.109375,
"logprob": -17.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.7265625,
"logprob": -15.578125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.5625,
"logprob": -18.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.734375,
"logprob": -20.625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.359375,
"logprob": -20.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.421875,
"logprob": -18.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.984375,
"logprob": -17.796875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.265625,
"logprob": -17.046875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.78125,
"logprob": -16.5625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.046875,
"logprob": -20.859375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.4375,
"logprob": -18.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.78125,
"logprob": -19.625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.671875,
"logprob": -19.546875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -14.2421875,
"logprob": -14.1953125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.0,
"logprob": -18.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.40625,
"logprob": -18.171875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.59375,
"logprob": -17.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.671875,
"logprob": -20.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.265625,
"logprob": -20.0,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.578125,
"logprob": -20.359375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.234375,
"logprob": -18.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.484375,
"logprob": -17.203125,
"text": "<image>"
},
{
......@@ -602,317 +607,322 @@
},
{
"id": 32001,
"logprob": -18.296875,
"logprob": -18.09375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.8671875,
"logprob": -15.71875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.375,
"logprob": -18.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.765625,
"logprob": -18.4375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.609375,
"logprob": -18.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.515625,
"logprob": -17.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.25,
"logprob": -16.265625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.640625,
"logprob": -19.4375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -14.8515625,
"logprob": -14.734375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.875,
"logprob": -20.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.28125,
"logprob": -20.21875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.453125,
"logprob": -18.359375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.25,
"logprob": -18.1875,
"text": "<image>"
},
{
"id": 32000,
"logprob": -3.2988281,
"text": "<fake_token_around_image>"
},
{
"id": 32001,
"logprob": -19.203125,
"logprob": -25.75,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.71875,
"logprob": -18.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.390625,
"logprob": -19.265625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.984375,
"logprob": -17.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.390625,
"logprob": -17.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.5,
"logprob": -20.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.296875,
"logprob": -17.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.4375,
"logprob": -16.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.015625,
"logprob": -19.65625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.359375,
"logprob": -18.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.375,
"logprob": -16.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.8125,
"logprob": -15.7421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.953125,
"logprob": -16.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.515625,
"logprob": -17.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.875,
"logprob": -16.5625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.109375,
"logprob": -18.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.265625,
"logprob": -18.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.234375,
"logprob": -16.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.28125,
"logprob": -18.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.25,
"logprob": -19.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.25,
"logprob": -19.890625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.765625,
"logprob": -20.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.609375,
"logprob": -19.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.359375,
"logprob": -20.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.90625,
"logprob": -19.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.28125,
"logprob": -19.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.5,
"logprob": -19.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.0625,
"logprob": -19.625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.125,
"logprob": -17.65625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.40625,
"logprob": -16.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.546875,
"logprob": -15.9453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.84375,
"logprob": -19.53125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.484375,
"logprob": -20.078125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.265625,
"logprob": -18.953125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.0625,
"logprob": -15.4375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.234375,
"logprob": -19.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.9453125,
"logprob": -15.4453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.0625,
"logprob": -20.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.515625,
"logprob": -20.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.375,
"logprob": -19.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.796875,
"logprob": -17.40625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.03125,
"logprob": -15.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.671875,
"logprob": -18.21875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.15625,
"logprob": -19.71875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.234375,
"logprob": -19.9375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.84375,
"logprob": -20.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.78125,
"logprob": -19.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.234375,
"logprob": -18.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.078125,
"logprob": -17.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.28125,
"logprob": -18.90625,
"text": "<image>"
},
{
......@@ -922,407 +932,417 @@
},
{
"id": 32001,
"logprob": -16.59375,
"logprob": -16.234375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.65625,
"logprob": -16.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.453125,
"logprob": -18.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.9375,
"logprob": -19.578125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.703125,
"logprob": -16.328125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.15625,
"logprob": -17.703125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.46875,
"logprob": -16.4375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.796875,
"logprob": -19.359375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.34375,
"logprob": -17.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.3125,
"logprob": -17.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.203125,
"logprob": -20.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.921875,
"logprob": -18.5625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.09375,
"logprob": -18.0,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.6875,
"text": "<image>"
"id": 32000,
"logprob": -2.7207031,
"text": "<fake_token_around_image>"
},
{
"id": 32001,
"logprob": -22.625,
"logprob": -23.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.375,
"logprob": -22.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.765625,
"logprob": -21.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.46875,
"logprob": -18.578125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.546875,
"logprob": -18.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.875,
"logprob": -20.296875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.453125,
"logprob": -18.5,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.09375,
"logprob": -16.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.5625,
"logprob": -20.890625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.15625,
"logprob": -19.390625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.171875,
"logprob": -18.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.671875,
"logprob": -16.0625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.84375,
"logprob": -17.265625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.75,
"logprob": -18.53125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.8125,
"logprob": -20.5,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.96875,
"logprob": -21.5,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.046875,
"logprob": -19.671875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.78125,
"logprob": -18.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.421875,
"logprob": -19.671875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.21875,
"logprob": -19.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.515625,
"logprob": -21.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.609375,
"logprob": -21.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.71875,
"logprob": -19.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.046875,
"logprob": -20.609375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.1875,
"logprob": -18.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.1875,
"logprob": -19.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.828125,
"logprob": -20.9375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.359375,
"logprob": -19.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.75,
"logprob": -18.0,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.90625,
"logprob": -17.40625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.765625,
"logprob": -18.640625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.453125,
"logprob": -18.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.890625,
"logprob": -20.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.015625,
"logprob": -19.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.90625,
"logprob": -15.796875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.953125,
"logprob": -18.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.46875,
"logprob": -15.6640625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.984375,
"logprob": -21.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.875,
"logprob": -19.796875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.859375,
"logprob": -19.671875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.046875,
"logprob": -18.640625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.375,
"logprob": -15.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.140625,
"logprob": -20.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.140625,
"logprob": -19.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.6875,
"logprob": -20.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.453125,
"logprob": -21.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.171875,
"logprob": -21.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.78125,
"logprob": -19.09375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.65625,
"logprob": -17.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.078125,
"logprob": -19.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.109375,
"logprob": -17.09375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.171875,
"logprob": -16.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.453125,
"logprob": -18.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.0625,
"logprob": -20.171875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.734375,
"logprob": -20.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.21875,
"logprob": -16.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.421875,
"logprob": -18.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.015625,
"logprob": -16.4375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.796875,
"logprob": -19.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.3125,
"logprob": -17.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.390625,
"logprob": -17.171875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.28125,
"logprob": -20.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.59375,
"logprob": -19.171875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.578125,
"text": "<image>"
},
{
"id": 32000,
"logprob": -3.0917969,
"text": "<fake_token_around_image>"
},
{
"id": 32001,
"logprob": -18.8125,
"logprob": -25.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.09375,
"logprob": -18.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.890625,
"logprob": -20.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.09375,
"logprob": -17.921875,
"text": "<image>"
},
{
......@@ -1332,62 +1352,62 @@
},
{
"id": 32001,
"logprob": -18.90625,
"logprob": -18.71875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.375,
"logprob": -21.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.640625,
"logprob": -16.734375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.859375,
"logprob": -20.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.40625,
"logprob": -21.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.8828125,
"logprob": -15.796875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.1171875,
"logprob": -15.1328125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.0625,
"logprob": -17.125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.921875,
"logprob": -18.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.40625,
"logprob": -21.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.0,
"logprob": -21.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.75,
"logprob": -20.734375,
"text": "<image>"
},
{
......@@ -1397,7 +1417,7 @@
},
{
"id": 32001,
"logprob": -19.46875,
"logprob": -19.5,
"text": "<image>"
},
{
......@@ -1407,37 +1427,37 @@
},
{
"id": 32001,
"logprob": -22.421875,
"logprob": -22.515625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.9375,
"logprob": -20.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.671875,
"logprob": -19.703125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.890625,
"logprob": -21.0,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.921875,
"logprob": -16.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.5,
"logprob": -17.53125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.90625,
"logprob": -17.9375,
"text": "<image>"
},
{
......@@ -1452,37 +1472,37 @@
},
{
"id": 32001,
"logprob": -16.40625,
"logprob": -16.390625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.453125,
"logprob": -18.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.234375,
"logprob": -20.265625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -22.28125,
"logprob": -22.296875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.515625,
"logprob": -18.484375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.4296875,
"logprob": -15.3984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.765625,
"logprob": -19.75,
"text": "<image>"
},
{
......@@ -1492,32 +1512,32 @@
},
{
"id": 32001,
"logprob": -21.46875,
"logprob": -21.609375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.875,
"logprob": -18.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.859375,
"logprob": -20.84375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.078125,
"logprob": -17.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.4375,
"logprob": -16.40625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.015625,
"logprob": -21.046875,
"text": "<image>"
},
{
......@@ -1527,87 +1547,87 @@
},
{
"id": 32001,
"logprob": -17.140625,
"logprob": -17.125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.484375,
"logprob": -21.5,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.015625,
"logprob": -20.0,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.84375,
"logprob": -18.78125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.40625,
"logprob": -16.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.890625,
"logprob": -16.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.65625,
"logprob": -16.703125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -13.6328125,
"logprob": -13.6171875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.4140625,
"logprob": -15.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.546875,
"logprob": -17.5,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.859375,
"logprob": -21.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.65625,
"logprob": -15.640625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.484375,
"logprob": -16.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.359375,
"logprob": -16.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.9375,
"logprob": -19.890625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.875,
"logprob": -17.890625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.453125,
"logprob": -17.40625,
"text": "<image>"
},
{
......@@ -1617,32 +1637,32 @@
},
{
"id": 32001,
"logprob": -19.171875,
"logprob": -19.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.9921875,
"logprob": -15.9609375,
"text": "<image>"
},
{
"id": 32000,
"logprob": -2.0429688,
"logprob": -2.0332031,
"text": "<fake_token_around_image>"
},
{
"id": 12018,
"logprob": -12.03125,
"logprob": -12.078125,
"text": "Write"
},
{
"id": 528,
"logprob": -10.25,
"logprob": -10.09375,
"text": "me"
},
{
"id": 264,
"logprob": -0.10437012,
"logprob": -0.103393555,
"text": "a"
},
{
......@@ -1652,37 +1672,37 @@
},
{
"id": 2838,
"logprob": -0.2277832,
"logprob": -0.23815918,
"text": "story"
},
{
"id": 32002,
"logprob": -10.84375,
"logprob": -10.9765625,
"text": "<end_of_utterance>"
},
{
"id": 259,
"logprob": -20.1875,
"logprob": -20.34375,
"text": " "
},
{
"id": 13,
"logprob": -8.7578125,
"logprob": -8.53125,
"text": "\n"
},
{
"id": 7226,
"logprob": -10.421875,
"logprob": -10.4765625,
"text": "Ass"
},
{
"id": 11143,
"logprob": -13.640625,
"logprob": -13.6015625,
"text": "istant"
},
{
"id": 28747,
"logprob": -0.005619049,
"logprob": -0.008514404,
"text": ":"
}
],
......@@ -1690,61 +1710,61 @@
"tokens": [
{
"id": 330,
"logprob": -0.12939453,
"logprob": -0.09289551,
"special": false,
"text": " A"
},
{
"id": 13088,
"logprob": -0.6660156,
"logprob": -0.6743164,
"special": false,
"text": " chicken"
},
{
"id": 349,
"logprob": -0.29638672,
"logprob": -0.31396484,
"special": false,
"text": " is"
},
{
"id": 6398,
"logprob": -0.05960083,
"logprob": -0.051727295,
"special": false,
"text": " sitting"
},
{
"id": 356,
"logprob": -0.26953125,
"logprob": -0.34448242,
"special": false,
"text": " on"
},
{
"id": 264,
"logprob": -0.1427002,
"logprob": -0.1194458,
"special": false,
"text": " a"
},
{
"id": 17972,
"logprob": -0.040649414,
"logprob": -0.03237915,
"special": false,
"text": " pile"
},
{
"id": 302,
"logprob": -0.0002708435,
"logprob": -0.00018751621,
"special": false,
"text": " of"
},
{
"id": 2445,
"logprob": -0.09429932,
"logprob": -0.07043457,
"special": false,
"text": " money"
},
{
"id": 28723,
"logprob": -0.006931305,
"logprob": -0.00422287,
"special": false,
"text": "."
}
......@@ -1766,12 +1786,12 @@
},
{
"id": 1247,
"logprob": -5.234375,
"logprob": -5.2382812,
"text": "User"
},
{
"id": 28747,
"logprob": -6.9648438,
"logprob": -6.9492188,
"text": ":"
},
{
......@@ -1781,12 +1801,12 @@
},
{
"id": 32001,
"logprob": -18.96875,
"logprob": -18.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.1875,
"logprob": -18.171875,
"text": "<image>"
},
{
......@@ -1811,7 +1831,7 @@
},
{
"id": 32001,
"logprob": -20.234375,
"logprob": -20.21875,
"text": "<image>"
},
{
......@@ -1821,7 +1841,7 @@
},
{
"id": 32001,
"logprob": -19.828125,
"logprob": -19.84375,
"text": "<image>"
},
{
......@@ -1856,22 +1876,22 @@
},
{
"id": 32001,
"logprob": -21.015625,
"logprob": -21.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.4375,
"logprob": -20.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.015625,
"logprob": -16.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.0625,
"logprob": -19.046875,
"text": "<image>"
},
{
......@@ -1886,7 +1906,7 @@
},
{
"id": 32001,
"logprob": -20.40625,
"logprob": -20.421875,
"text": "<image>"
},
{
......@@ -1896,7 +1916,7 @@
},
{
"id": 32001,
"logprob": -20.84375,
"logprob": -20.875,
"text": "<image>"
},
{
......@@ -1906,12 +1926,12 @@
},
{
"id": 32001,
"logprob": -19.984375,
"logprob": -19.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.21875,
"logprob": -18.234375,
"text": "<image>"
},
{
......@@ -1921,7 +1941,7 @@
},
{
"id": 32001,
"logprob": -21.203125,
"logprob": -21.1875,
"text": "<image>"
},
{
......@@ -1936,7 +1956,7 @@
},
{
"id": 32001,
"logprob": -18.984375,
"logprob": -19.03125,
"text": "<image>"
},
{
......@@ -1951,17 +1971,17 @@
},
{
"id": 32001,
"logprob": -18.0,
"logprob": -17.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.828125,
"logprob": -18.859375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.9375,
"logprob": -17.921875,
"text": "<image>"
},
{
......@@ -1971,12 +1991,12 @@
},
{
"id": 32001,
"logprob": -18.640625,
"logprob": -18.65625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.125,
"logprob": -20.140625,
"text": "<image>"
},
{
......@@ -1996,12 +2016,12 @@
},
{
"id": 32001,
"logprob": -23.203125,
"logprob": -23.21875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.4375,
"logprob": -17.421875,
"text": "<image>"
},
{
......@@ -2011,7 +2031,7 @@
},
{
"id": 32001,
"logprob": -23.015625,
"logprob": -23.0,
"text": "<image>"
},
{
......@@ -2036,7 +2056,7 @@
},
{
"id": 32001,
"logprob": -18.40625,
"logprob": -18.421875,
"text": "<image>"
},
{
......@@ -2046,7 +2066,7 @@
},
{
"id": 32001,
"logprob": -18.34375,
"logprob": -18.328125,
"text": "<image>"
},
{
......@@ -2056,7 +2076,7 @@
},
{
"id": 32001,
"logprob": -18.671875,
"logprob": -18.65625,
"text": "<image>"
},
{
......@@ -2071,7 +2091,7 @@
},
{
"id": 32001,
"logprob": -18.1875,
"logprob": -18.15625,
"text": "<image>"
},
{
......@@ -2081,7 +2101,7 @@
},
{
"id": 32001,
"logprob": -20.140625,
"logprob": -20.15625,
"text": "<image>"
},
{
......@@ -2091,47 +2111,52 @@
},
{
"id": 32001,
"logprob": -18.71875,
"logprob": -18.703125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.9375,
"logprob": -17.921875,
"text": "<image>"
},
{
"id": 32000,
"logprob": -3.015625,
"text": "<fake_token_around_image>"
},
{
"id": 32001,
"logprob": -17.90625,
"logprob": -22.109375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.125,
"logprob": -18.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.296875,
"logprob": -20.125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.25,
"logprob": -17.125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.96875,
"logprob": -17.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.546875,
"logprob": -19.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.921875,
"logprob": -19.65625,
"text": "<image>"
},
{
......@@ -2141,212 +2166,212 @@
},
{
"id": 32001,
"logprob": -19.546875,
"logprob": -19.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.1875,
"logprob": -20.046875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.953125,
"logprob": -20.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.828125,
"logprob": -15.8046875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.375,
"logprob": -16.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.171875,
"logprob": -19.953125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.03125,
"logprob": -18.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.71875,
"logprob": -21.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.65625,
"logprob": -20.515625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.453125,
"logprob": -19.171875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.484375,
"logprob": -19.296875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.875,
"logprob": -16.71875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.65625,
"logprob": -20.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.296875,
"logprob": -21.125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.546875,
"logprob": -20.40625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.78125,
"logprob": -20.734375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.96875,
"logprob": -17.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.5,
"logprob": -20.328125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.4140625,
"logprob": -15.2734375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.046875,
"logprob": -18.84375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.109375,
"logprob": -17.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.7265625,
"logprob": -15.578125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.5625,
"logprob": -18.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.734375,
"logprob": -20.625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.359375,
"logprob": -20.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.421875,
"logprob": -18.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.984375,
"logprob": -17.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.265625,
"logprob": -17.046875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.78125,
"logprob": -16.578125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.046875,
"logprob": -20.859375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.4375,
"logprob": -18.15625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.78125,
"logprob": -19.625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.671875,
"logprob": -19.546875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -14.2421875,
"logprob": -14.1953125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.0,
"logprob": -18.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.40625,
"logprob": -18.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.59375,
"logprob": -17.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.671875,
"logprob": -20.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.265625,
"logprob": -20.0,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.578125,
"logprob": -20.359375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.234375,
"logprob": -18.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.484375,
"logprob": -17.203125,
"text": "<image>"
},
{
......@@ -2356,317 +2381,322 @@
},
{
"id": 32001,
"logprob": -18.28125,
"logprob": -18.09375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.8671875,
"logprob": -15.71875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.375,
"logprob": -18.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.765625,
"logprob": -18.4375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.609375,
"logprob": -18.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.515625,
"logprob": -17.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.25,
"logprob": -16.265625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.640625,
"logprob": -19.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -14.8515625,
"logprob": -14.734375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.875,
"logprob": -20.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.28125,
"logprob": -20.21875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.453125,
"logprob": -18.359375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.25,
"logprob": -18.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.203125,
"text": "<image>"
"id": 32000,
"logprob": -3.2988281,
"text": "<fake_token_around_image>"
},
{
"id": 32001,
"logprob": -18.703125,
"logprob": -25.75,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.390625,
"logprob": -18.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.984375,
"logprob": -19.265625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.390625,
"logprob": -17.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.5,
"logprob": -17.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.296875,
"logprob": -20.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.4375,
"logprob": -17.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.015625,
"logprob": -16.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.359375,
"logprob": -19.65625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.375,
"logprob": -18.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.8125,
"logprob": -16.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.953125,
"logprob": -15.75,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.515625,
"logprob": -16.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.875,
"logprob": -17.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.109375,
"logprob": -16.5625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.265625,
"logprob": -18.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.234375,
"logprob": -18.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.28125,
"logprob": -16.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.234375,
"logprob": -18.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.25,
"logprob": -19.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.765625,
"logprob": -19.890625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.609375,
"logprob": -20.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.359375,
"logprob": -19.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.90625,
"logprob": -20.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.28125,
"logprob": -19.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.5,
"logprob": -19.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.0625,
"logprob": -19.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.125,
"logprob": -19.625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.40625,
"logprob": -17.671875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.546875,
"logprob": -16.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.84375,
"logprob": -15.953125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.484375,
"logprob": -19.53125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.265625,
"logprob": -20.078125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.0625,
"logprob": -18.953125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.234375,
"logprob": -15.4453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.9453125,
"logprob": -19.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.0625,
"logprob": -15.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.515625,
"logprob": -20.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.375,
"logprob": -20.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.796875,
"logprob": -19.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.03125,
"logprob": -17.40625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.671875,
"logprob": -15.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.15625,
"logprob": -18.21875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.234375,
"logprob": -19.71875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.84375,
"logprob": -19.9375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.78125,
"logprob": -19.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.234375,
"logprob": -18.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.078125,
"logprob": -17.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.28125,
"logprob": -18.90625,
"text": "<image>"
},
{
......@@ -2676,472 +2706,482 @@
},
{
"id": 32001,
"logprob": -16.59375,
"logprob": -16.234375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.65625,
"logprob": -16.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.4375,
"logprob": -18.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.9375,
"logprob": -19.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.703125,
"logprob": -16.328125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.15625,
"logprob": -17.703125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.46875,
"logprob": -16.4375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.796875,
"logprob": -19.359375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.34375,
"logprob": -17.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.3125,
"logprob": -17.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.203125,
"logprob": -20.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.90625,
"logprob": -18.5625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.09375,
"logprob": -18.0,
"text": "<image>"
},
{
"id": 32000,
"logprob": -2.7207031,
"text": "<fake_token_around_image>"
},
{
"id": 32001,
"logprob": -18.6875,
"logprob": -23.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -22.625,
"logprob": -22.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.375,
"logprob": -21.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.765625,
"logprob": -18.578125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.46875,
"logprob": -18.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.546875,
"logprob": -20.296875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.875,
"logprob": -18.484375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.453125,
"logprob": -16.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.09375,
"logprob": -20.890625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.5625,
"logprob": -19.390625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.15625,
"logprob": -18.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.171875,
"logprob": -16.078125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.671875,
"logprob": -17.28125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.84375,
"logprob": -18.53125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.75,
"logprob": -20.5,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.8125,
"logprob": -21.5,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.96875,
"logprob": -19.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.046875,
"logprob": -18.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.78125,
"logprob": -19.671875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.421875,
"logprob": -19.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.21875,
"logprob": -21.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.515625,
"logprob": -21.328125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.609375,
"logprob": -19.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.71875,
"logprob": -20.609375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.046875,
"logprob": -18.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.171875,
"logprob": -19.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.1875,
"logprob": -20.9375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.828125,
"logprob": -19.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.359375,
"logprob": -18.0,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.75,
"logprob": -17.40625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.90625,
"logprob": -18.640625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.765625,
"logprob": -18.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.453125,
"logprob": -20.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.890625,
"logprob": -19.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.015625,
"logprob": -15.796875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.90625,
"logprob": -18.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.953125,
"logprob": -15.6640625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.46875,
"logprob": -21.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.984375,
"logprob": -19.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.890625,
"logprob": -19.671875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.875,
"logprob": -18.640625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.046875,
"logprob": -15.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.375,
"logprob": -20.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.140625,
"logprob": -19.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.140625,
"logprob": -20.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.703125,
"logprob": -21.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.453125,
"logprob": -21.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.171875,
"logprob": -19.09375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.765625,
"logprob": -17.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.65625,
"logprob": -19.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.078125,
"logprob": -17.09375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.109375,
"logprob": -16.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.171875,
"logprob": -18.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.453125,
"logprob": -20.171875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.0625,
"logprob": -20.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.734375,
"logprob": -16.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.21875,
"logprob": -18.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.421875,
"logprob": -16.4375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.015625,
"logprob": -19.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.796875,
"logprob": -17.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.3125,
"logprob": -17.171875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.390625,
"logprob": -20.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.28125,
"logprob": -19.171875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.59375,
"logprob": -18.578125,
"text": "<image>"
},
{
"id": 32000,
"logprob": -3.0917969,
"text": "<fake_token_around_image>"
},
{
"id": 32001,
"logprob": -18.8125,
"logprob": -25.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.09375,
"logprob": -18.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.890625,
"logprob": -20.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.09375,
"logprob": -17.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.75,
"logprob": -18.734375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.90625,
"logprob": -18.71875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.375,
"logprob": -21.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.640625,
"logprob": -16.734375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.859375,
"logprob": -20.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.40625,
"logprob": -21.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.875,
"logprob": -15.796875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.1171875,
"logprob": -15.1328125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.078125,
"logprob": -17.125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.921875,
"logprob": -18.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.40625,
"logprob": -21.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.0,
"logprob": -21.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.75,
"logprob": -20.734375,
"text": "<image>"
},
{
......@@ -3151,7 +3191,7 @@
},
{
"id": 32001,
"logprob": -19.453125,
"logprob": -19.5,
"text": "<image>"
},
{
......@@ -3161,37 +3201,37 @@
},
{
"id": 32001,
"logprob": -22.421875,
"logprob": -22.515625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.9375,
"logprob": -20.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.671875,
"logprob": -19.703125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.890625,
"logprob": -21.0,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.921875,
"logprob": -16.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.5,
"logprob": -17.53125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.890625,
"logprob": -17.921875,
"text": "<image>"
},
{
......@@ -3206,37 +3246,37 @@
},
{
"id": 32001,
"logprob": -16.40625,
"logprob": -16.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.453125,
"logprob": -18.4375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.21875,
"logprob": -20.265625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -22.28125,
"logprob": -22.296875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.515625,
"logprob": -18.484375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.4296875,
"logprob": -15.390625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.765625,
"logprob": -19.75,
"text": "<image>"
},
{
......@@ -3246,32 +3286,32 @@
},
{
"id": 32001,
"logprob": -21.46875,
"logprob": -21.609375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.875,
"logprob": -18.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.859375,
"logprob": -20.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.078125,
"logprob": -17.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.4375,
"logprob": -16.40625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.015625,
"logprob": -21.046875,
"text": "<image>"
},
{
......@@ -3281,27 +3321,27 @@
},
{
"id": 32001,
"logprob": -17.125,
"logprob": -17.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.484375,
"logprob": -21.515625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.015625,
"logprob": -20.0,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.84375,
"logprob": -18.78125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.421875,
"logprob": -16.375,
"text": "<image>"
},
{
......@@ -3311,57 +3351,57 @@
},
{
"id": 32001,
"logprob": -16.65625,
"logprob": -16.703125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -13.640625,
"logprob": -13.625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.421875,
"logprob": -15.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.546875,
"logprob": -17.515625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.875,
"logprob": -21.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.65625,
"logprob": -15.640625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.484375,
"logprob": -16.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.359375,
"logprob": -16.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.953125,
"logprob": -19.890625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.875,
"logprob": -17.890625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.453125,
"logprob": -17.40625,
"text": "<image>"
},
{
......@@ -3371,32 +3411,32 @@
},
{
"id": 32001,
"logprob": -19.171875,
"logprob": -19.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.9921875,
"logprob": -15.9609375,
"text": "<image>"
},
{
"id": 32000,
"logprob": -2.0429688,
"logprob": -2.0332031,
"text": "<fake_token_around_image>"
},
{
"id": 12018,
"logprob": -12.03125,
"logprob": -12.078125,
"text": "Write"
},
{
"id": 528,
"logprob": -10.2578125,
"logprob": -10.109375,
"text": "me"
},
{
"id": 264,
"logprob": -0.10418701,
"logprob": -0.103515625,
"text": "a"
},
{
......@@ -3406,37 +3446,37 @@
},
{
"id": 2838,
"logprob": -0.22741699,
"logprob": -0.23864746,
"text": "story"
},
{
"id": 32002,
"logprob": -10.8515625,
"logprob": -10.9609375,
"text": "<end_of_utterance>"
},
{
"id": 259,
"logprob": -20.203125,
"logprob": -20.34375,
"text": " "
},
{
"id": 13,
"logprob": -8.7421875,
"logprob": -8.5546875,
"text": "\n"
},
{
"id": 7226,
"logprob": -10.4140625,
"logprob": -10.484375,
"text": "Ass"
},
{
"id": 11143,
"logprob": -13.6328125,
"logprob": -13.6015625,
"text": "istant"
},
{
"id": 28747,
"logprob": -0.005580902,
"logprob": -0.008308411,
"text": ":"
}
],
......@@ -3444,61 +3484,61 @@
"tokens": [
{
"id": 330,
"logprob": -0.1295166,
"logprob": -0.09448242,
"special": false,
"text": " A"
},
{
"id": 13088,
"logprob": -0.6669922,
"logprob": -0.6743164,
"special": false,
"text": " chicken"
},
{
"id": 349,
"logprob": -0.29711914,
"logprob": -0.31201172,
"special": false,
"text": " is"
},
{
"id": 6398,
"logprob": -0.059936523,
"logprob": -0.051635742,
"special": false,
"text": " sitting"
},
{
"id": 356,
"logprob": -0.27124023,
"logprob": -0.34033203,
"special": false,
"text": " on"
},
{
"id": 264,
"logprob": -0.140625,
"logprob": -0.1194458,
"special": false,
"text": " a"
},
{
"id": 17972,
"logprob": -0.04058838,
"logprob": -0.032562256,
"special": false,
"text": " pile"
},
{
"id": 302,
"logprob": -0.00027012825,
"logprob": -0.00018763542,
"special": false,
"text": " of"
},
{
"id": 2445,
"logprob": -0.09503174,
"logprob": -0.07122803,
"special": false,
"text": " money"
},
{
"id": 28723,
"logprob": -0.006942749,
"logprob": -0.0041007996,
"special": false,
"text": "."
}
......@@ -3520,12 +3560,12 @@
},
{
"id": 1247,
"logprob": -5.2460938,
"logprob": -5.2382812,
"text": "User"
},
{
"id": 28747,
"logprob": -6.9570312,
"logprob": -6.9492188,
"text": ":"
},
{
......@@ -3535,12 +3575,12 @@
},
{
"id": 32001,
"logprob": -18.96875,
"logprob": -18.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.1875,
"logprob": -18.171875,
"text": "<image>"
},
{
......@@ -3565,7 +3605,7 @@
},
{
"id": 32001,
"logprob": -20.234375,
"logprob": -20.21875,
"text": "<image>"
},
{
......@@ -3610,22 +3650,22 @@
},
{
"id": 32001,
"logprob": -21.015625,
"logprob": -21.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.4375,
"logprob": -20.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.015625,
"logprob": -16.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.0625,
"logprob": -19.046875,
"text": "<image>"
},
{
......@@ -3635,12 +3675,12 @@
},
{
"id": 32001,
"logprob": -23.625,
"logprob": -23.609375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.40625,
"logprob": -20.421875,
"text": "<image>"
},
{
......@@ -3650,7 +3690,7 @@
},
{
"id": 32001,
"logprob": -20.84375,
"logprob": -20.875,
"text": "<image>"
},
{
......@@ -3660,12 +3700,12 @@
},
{
"id": 32001,
"logprob": -19.984375,
"logprob": -19.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.21875,
"logprob": -18.234375,
"text": "<image>"
},
{
......@@ -3675,7 +3715,7 @@
},
{
"id": 32001,
"logprob": -21.21875,
"logprob": -21.1875,
"text": "<image>"
},
{
......@@ -3690,7 +3730,7 @@
},
{
"id": 32001,
"logprob": -18.984375,
"logprob": -19.03125,
"text": "<image>"
},
{
......@@ -3705,12 +3745,12 @@
},
{
"id": 32001,
"logprob": -18.0,
"logprob": -17.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.84375,
"logprob": -18.859375,
"text": "<image>"
},
{
......@@ -3725,12 +3765,12 @@
},
{
"id": 32001,
"logprob": -18.640625,
"logprob": -18.65625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.125,
"logprob": -20.140625,
"text": "<image>"
},
{
......@@ -3750,12 +3790,12 @@
},
{
"id": 32001,
"logprob": -23.203125,
"logprob": -23.21875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.4375,
"logprob": -17.421875,
"text": "<image>"
},
{
......@@ -3765,7 +3805,7 @@
},
{
"id": 32001,
"logprob": -23.015625,
"logprob": -23.0,
"text": "<image>"
},
{
......@@ -3790,7 +3830,7 @@
},
{
"id": 32001,
"logprob": -18.40625,
"logprob": -18.421875,
"text": "<image>"
},
{
......@@ -3800,7 +3840,7 @@
},
{
"id": 32001,
"logprob": -18.34375,
"logprob": -18.328125,
"text": "<image>"
},
{
......@@ -3810,7 +3850,7 @@
},
{
"id": 32001,
"logprob": -18.671875,
"logprob": -18.65625,
"text": "<image>"
},
{
......@@ -3825,7 +3865,7 @@
},
{
"id": 32001,
"logprob": -18.1875,
"logprob": -18.15625,
"text": "<image>"
},
{
......@@ -3835,7 +3875,7 @@
},
{
"id": 32001,
"logprob": -20.140625,
"logprob": -20.15625,
"text": "<image>"
},
{
......@@ -3845,47 +3885,52 @@
},
{
"id": 32001,
"logprob": -18.71875,
"logprob": -18.703125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.9375,
"logprob": -17.921875,
"text": "<image>"
},
{
"id": 32000,
"logprob": -3.015625,
"text": "<fake_token_around_image>"
},
{
"id": 32001,
"logprob": -17.921875,
"logprob": -22.109375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.125,
"logprob": -18.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.296875,
"logprob": -20.125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.25,
"logprob": -17.125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.96875,
"logprob": -17.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.546875,
"logprob": -19.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.921875,
"logprob": -19.65625,
"text": "<image>"
},
{
......@@ -3895,212 +3940,212 @@
},
{
"id": 32001,
"logprob": -19.546875,
"logprob": -19.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.1875,
"logprob": -20.046875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.953125,
"logprob": -20.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.8359375,
"logprob": -15.8046875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.375,
"logprob": -16.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.1875,
"logprob": -19.953125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.03125,
"logprob": -18.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.71875,
"logprob": -21.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.65625,
"logprob": -20.515625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.453125,
"logprob": -19.171875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.484375,
"logprob": -19.296875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.875,
"logprob": -16.71875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.65625,
"logprob": -20.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.296875,
"logprob": -21.125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.546875,
"logprob": -20.40625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.78125,
"logprob": -20.734375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.96875,
"logprob": -17.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.5,
"logprob": -20.328125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.4140625,
"logprob": -15.2734375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.046875,
"logprob": -18.84375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.109375,
"logprob": -17.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.7265625,
"logprob": -15.578125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.5625,
"logprob": -18.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.734375,
"logprob": -20.625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.359375,
"logprob": -20.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.421875,
"logprob": -18.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.984375,
"logprob": -17.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.265625,
"logprob": -17.046875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.78125,
"logprob": -16.578125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.046875,
"logprob": -20.859375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.4375,
"logprob": -18.15625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.78125,
"logprob": -19.625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.671875,
"logprob": -19.546875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -14.25,
"logprob": -14.1953125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.0,
"logprob": -18.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.40625,
"logprob": -18.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.59375,
"logprob": -17.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.671875,
"logprob": -20.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.265625,
"logprob": -20.0,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.578125,
"logprob": -20.359375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.234375,
"logprob": -18.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.484375,
"logprob": -17.203125,
"text": "<image>"
},
{
......@@ -4110,317 +4155,322 @@
},
{
"id": 32001,
"logprob": -18.28125,
"logprob": -18.09375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.8671875,
"logprob": -15.71875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.375,
"logprob": -18.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.765625,
"logprob": -18.4375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.609375,
"logprob": -18.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.515625,
"logprob": -17.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.25,
"logprob": -16.265625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.640625,
"logprob": -19.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -14.8515625,
"logprob": -14.734375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.875,
"logprob": -20.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.265625,
"logprob": -20.21875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.453125,
"logprob": -18.359375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.25,
"logprob": -18.1875,
"text": "<image>"
},
{
"id": 32000,
"logprob": -3.2988281,
"text": "<fake_token_around_image>"
},
{
"id": 32001,
"logprob": -19.203125,
"logprob": -25.75,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.71875,
"logprob": -18.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.390625,
"logprob": -19.265625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.0,
"logprob": -17.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.390625,
"logprob": -17.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.5,
"logprob": -20.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.296875,
"logprob": -17.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.4375,
"logprob": -16.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.015625,
"logprob": -19.65625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.359375,
"logprob": -18.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.375,
"logprob": -16.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.8125,
"logprob": -15.75,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.953125,
"logprob": -16.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.515625,
"logprob": -17.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.875,
"logprob": -16.5625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.109375,
"logprob": -18.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.265625,
"logprob": -18.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.234375,
"logprob": -16.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.28125,
"logprob": -18.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.234375,
"logprob": -19.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.25,
"logprob": -19.890625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.765625,
"logprob": -20.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.609375,
"logprob": -19.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.34375,
"logprob": -20.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.921875,
"logprob": -19.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.28125,
"logprob": -19.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.515625,
"logprob": -19.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.0625,
"logprob": -19.625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.125,
"logprob": -17.671875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.40625,
"logprob": -16.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.546875,
"logprob": -15.953125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.84375,
"logprob": -19.53125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.484375,
"logprob": -20.078125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.265625,
"logprob": -18.953125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.0625,
"logprob": -15.4453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.234375,
"logprob": -19.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.9453125,
"logprob": -15.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.046875,
"logprob": -20.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.515625,
"logprob": -20.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.375,
"logprob": -19.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.796875,
"logprob": -17.40625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.03125,
"logprob": -15.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.671875,
"logprob": -18.21875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.15625,
"logprob": -19.71875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.234375,
"logprob": -19.9375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.84375,
"logprob": -20.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.78125,
"logprob": -19.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.234375,
"logprob": -18.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.078125,
"logprob": -17.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.28125,
"logprob": -18.90625,
"text": "<image>"
},
{
......@@ -4430,472 +4480,482 @@
},
{
"id": 32001,
"logprob": -16.59375,
"logprob": -16.234375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.65625,
"logprob": -16.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.453125,
"logprob": -18.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.9375,
"logprob": -19.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.703125,
"logprob": -16.328125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.15625,
"logprob": -17.703125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.46875,
"logprob": -16.4375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.796875,
"logprob": -19.359375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.34375,
"logprob": -17.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.3125,
"logprob": -17.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.1875,
"logprob": -20.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.921875,
"logprob": -18.5625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.09375,
"logprob": -18.0,
"text": "<image>"
},
{
"id": 32000,
"logprob": -2.7207031,
"text": "<fake_token_around_image>"
},
{
"id": 32001,
"logprob": -18.6875,
"logprob": -23.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -22.625,
"logprob": -22.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.359375,
"logprob": -21.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.765625,
"logprob": -18.578125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.46875,
"logprob": -18.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.546875,
"logprob": -20.296875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.875,
"logprob": -18.484375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.453125,
"logprob": -16.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.09375,
"logprob": -20.890625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.5625,
"logprob": -19.390625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.15625,
"logprob": -18.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.171875,
"logprob": -16.078125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.671875,
"logprob": -17.28125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.859375,
"logprob": -18.53125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.734375,
"logprob": -20.5,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.8125,
"logprob": -21.5,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.96875,
"logprob": -19.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.046875,
"logprob": -18.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.78125,
"logprob": -19.671875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.421875,
"logprob": -19.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.21875,
"logprob": -21.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.515625,
"logprob": -21.328125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.609375,
"logprob": -19.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.734375,
"logprob": -20.609375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.046875,
"logprob": -18.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.1875,
"logprob": -19.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.1875,
"logprob": -20.9375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.828125,
"logprob": -19.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.359375,
"logprob": -18.0,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.75,
"logprob": -17.40625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.90625,
"logprob": -18.640625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.765625,
"logprob": -18.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.453125,
"logprob": -20.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.890625,
"logprob": -19.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.015625,
"logprob": -15.796875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.90625,
"logprob": -18.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.953125,
"logprob": -15.6640625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.46875,
"logprob": -21.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.984375,
"logprob": -19.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.875,
"logprob": -19.671875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.859375,
"logprob": -18.640625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.046875,
"logprob": -15.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.375,
"logprob": -20.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.140625,
"logprob": -19.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.140625,
"logprob": -20.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.703125,
"logprob": -21.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.453125,
"logprob": -21.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.171875,
"logprob": -19.09375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.765625,
"logprob": -17.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.65625,
"logprob": -19.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.078125,
"logprob": -17.09375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.109375,
"logprob": -16.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.171875,
"logprob": -18.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.453125,
"logprob": -20.171875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.0625,
"logprob": -20.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.734375,
"logprob": -16.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.21875,
"logprob": -18.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.421875,
"logprob": -16.4375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.015625,
"logprob": -19.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.796875,
"logprob": -17.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.3125,
"logprob": -17.171875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.390625,
"logprob": -20.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.28125,
"logprob": -19.171875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.59375,
"logprob": -18.578125,
"text": "<image>"
},
{
"id": 32000,
"logprob": -3.0917969,
"text": "<fake_token_around_image>"
},
{
"id": 32001,
"logprob": -18.8125,
"logprob": -25.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.09375,
"logprob": -18.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.890625,
"logprob": -20.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.09375,
"logprob": -17.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.75,
"logprob": -18.734375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.921875,
"logprob": -18.71875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.375,
"logprob": -21.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.640625,
"logprob": -16.734375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.859375,
"logprob": -20.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.40625,
"logprob": -21.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.875,
"logprob": -15.796875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.1171875,
"logprob": -15.1328125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.078125,
"logprob": -17.125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.921875,
"logprob": -18.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.40625,
"logprob": -21.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.0,
"logprob": -21.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.75,
"logprob": -20.734375,
"text": "<image>"
},
{
......@@ -4905,7 +4965,7 @@
},
{
"id": 32001,
"logprob": -19.453125,
"logprob": -19.5,
"text": "<image>"
},
{
......@@ -4915,37 +4975,37 @@
},
{
"id": 32001,
"logprob": -22.421875,
"logprob": -22.515625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.9375,
"logprob": -20.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.671875,
"logprob": -19.703125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.890625,
"logprob": -21.0,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.921875,
"logprob": -16.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.5,
"logprob": -17.53125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.90625,
"logprob": -17.921875,
"text": "<image>"
},
{
......@@ -4960,37 +5020,37 @@
},
{
"id": 32001,
"logprob": -16.40625,
"logprob": -16.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.453125,
"logprob": -18.4375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.234375,
"logprob": -20.265625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -22.28125,
"logprob": -22.296875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.53125,
"logprob": -18.484375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.4296875,
"logprob": -15.390625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.765625,
"logprob": -19.75,
"text": "<image>"
},
{
......@@ -5000,32 +5060,32 @@
},
{
"id": 32001,
"logprob": -21.46875,
"logprob": -21.609375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.875,
"logprob": -18.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.859375,
"logprob": -20.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.078125,
"logprob": -17.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.421875,
"logprob": -16.40625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.0,
"logprob": -21.046875,
"text": "<image>"
},
{
......@@ -5040,22 +5100,22 @@
},
{
"id": 32001,
"logprob": -21.484375,
"logprob": -21.515625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.015625,
"logprob": -20.0,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.84375,
"logprob": -18.78125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.421875,
"logprob": -16.375,
"text": "<image>"
},
{
......@@ -5065,57 +5125,57 @@
},
{
"id": 32001,
"logprob": -16.65625,
"logprob": -16.703125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -13.640625,
"logprob": -13.625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.4140625,
"logprob": -15.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.546875,
"logprob": -17.515625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.859375,
"logprob": -21.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.65625,
"logprob": -15.640625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.484375,
"logprob": -16.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.359375,
"logprob": -16.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.953125,
"logprob": -19.890625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.875,
"logprob": -17.890625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.4375,
"logprob": -17.40625,
"text": "<image>"
},
{
......@@ -5125,72 +5185,72 @@
},
{
"id": 32001,
"logprob": -19.171875,
"logprob": -19.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.9921875,
"logprob": -15.9609375,
"text": "<image>"
},
{
"id": 32000,
"logprob": -2.0429688,
"logprob": -2.0332031,
"text": "<fake_token_around_image>"
},
{
"id": 12018,
"logprob": -12.0390625,
"logprob": -12.078125,
"text": "Write"
},
{
"id": 528,
"logprob": -10.25,
"logprob": -10.109375,
"text": "me"
},
{
"id": 264,
"logprob": -0.10443115,
"logprob": -0.103515625,
"text": "a"
},
{
"id": 2485,
"logprob": -4.5742188,
"logprob": -4.5664062,
"text": "short"
},
{
"id": 2838,
"logprob": -0.22729492,
"logprob": -0.23864746,
"text": "story"
},
{
"id": 32002,
"logprob": -10.84375,
"logprob": -10.9609375,
"text": "<end_of_utterance>"
},
{
"id": 259,
"logprob": -20.1875,
"logprob": -20.34375,
"text": " "
},
{
"id": 13,
"logprob": -8.7578125,
"logprob": -8.5546875,
"text": "\n"
},
{
"id": 7226,
"logprob": -10.4140625,
"logprob": -10.484375,
"text": "Ass"
},
{
"id": 11143,
"logprob": -13.6328125,
"logprob": -13.6015625,
"text": "istant"
},
{
"id": 28747,
"logprob": -0.0056533813,
"logprob": -0.008308411,
"text": ":"
}
],
......@@ -5198,61 +5258,61 @@
"tokens": [
{
"id": 330,
"logprob": -0.12963867,
"logprob": -0.09448242,
"special": false,
"text": " A"
},
{
"id": 13088,
"logprob": -0.6660156,
"logprob": -0.6743164,
"special": false,
"text": " chicken"
},
{
"id": 349,
"logprob": -0.29516602,
"logprob": -0.31201172,
"special": false,
"text": " is"
},
{
"id": 6398,
"logprob": -0.060028076,
"logprob": -0.051635742,
"special": false,
"text": " sitting"
},
{
"id": 356,
"logprob": -0.27075195,
"logprob": -0.34033203,
"special": false,
"text": " on"
},
{
"id": 264,
"logprob": -0.1427002,
"logprob": -0.1194458,
"special": false,
"text": " a"
},
{
"id": 17972,
"logprob": -0.04067993,
"logprob": -0.032562256,
"special": false,
"text": " pile"
},
{
"id": 302,
"logprob": -0.000269413,
"logprob": -0.00018787384,
"special": false,
"text": " of"
},
{
"id": 2445,
"logprob": -0.09387207,
"logprob": -0.07122803,
"special": false,
"text": " money"
},
{
"id": 28723,
"logprob": -0.0069236755,
"logprob": -0.0041007996,
"special": false,
"text": "."
}
......@@ -5274,12 +5334,12 @@
},
{
"id": 1247,
"logprob": -5.2421875,
"logprob": -5.2382812,
"text": "User"
},
{
"id": 28747,
"logprob": -6.9570312,
"logprob": -6.9492188,
"text": ":"
},
{
......@@ -5289,12 +5349,12 @@
},
{
"id": 32001,
"logprob": -18.96875,
"logprob": -18.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.1875,
"logprob": -18.171875,
"text": "<image>"
},
{
......@@ -5319,7 +5379,7 @@
},
{
"id": 32001,
"logprob": -20.25,
"logprob": -20.21875,
"text": "<image>"
},
{
......@@ -5364,7 +5424,7 @@
},
{
"id": 32001,
"logprob": -21.015625,
"logprob": -21.03125,
"text": "<image>"
},
{
......@@ -5374,12 +5434,12 @@
},
{
"id": 32001,
"logprob": -16.015625,
"logprob": -16.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.078125,
"logprob": -19.046875,
"text": "<image>"
},
{
......@@ -5389,12 +5449,12 @@
},
{
"id": 32001,
"logprob": -23.625,
"logprob": -23.609375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.40625,
"logprob": -20.421875,
"text": "<image>"
},
{
......@@ -5404,7 +5464,7 @@
},
{
"id": 32001,
"logprob": -20.84375,
"logprob": -20.875,
"text": "<image>"
},
{
......@@ -5419,7 +5479,7 @@
},
{
"id": 32001,
"logprob": -18.21875,
"logprob": -18.234375,
"text": "<image>"
},
{
......@@ -5429,7 +5489,7 @@
},
{
"id": 32001,
"logprob": -21.203125,
"logprob": -21.1875,
"text": "<image>"
},
{
......@@ -5444,7 +5504,7 @@
},
{
"id": 32001,
"logprob": -18.984375,
"logprob": -19.03125,
"text": "<image>"
},
{
......@@ -5459,17 +5519,17 @@
},
{
"id": 32001,
"logprob": -18.0,
"logprob": -17.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.84375,
"logprob": -18.859375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.9375,
"logprob": -17.921875,
"text": "<image>"
},
{
......@@ -5479,12 +5539,12 @@
},
{
"id": 32001,
"logprob": -18.640625,
"logprob": -18.65625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.125,
"logprob": -20.140625,
"text": "<image>"
},
{
......@@ -5504,12 +5564,12 @@
},
{
"id": 32001,
"logprob": -23.203125,
"logprob": -23.21875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.4375,
"logprob": -17.421875,
"text": "<image>"
},
{
......@@ -5519,7 +5579,7 @@
},
{
"id": 32001,
"logprob": -23.015625,
"logprob": -23.0,
"text": "<image>"
},
{
......@@ -5544,7 +5604,7 @@
},
{
"id": 32001,
"logprob": -18.40625,
"logprob": -18.421875,
"text": "<image>"
},
{
......@@ -5554,17 +5614,17 @@
},
{
"id": 32001,
"logprob": -18.34375,
"logprob": -18.328125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.140625,
"logprob": -17.125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.671875,
"logprob": -18.65625,
"text": "<image>"
},
{
......@@ -5579,7 +5639,7 @@
},
{
"id": 32001,
"logprob": -18.1875,
"logprob": -18.15625,
"text": "<image>"
},
{
......@@ -5599,47 +5659,52 @@
},
{
"id": 32001,
"logprob": -18.71875,
"logprob": -18.703125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.9375,
"logprob": -17.921875,
"text": "<image>"
},
{
"id": 32000,
"logprob": -3.015625,
"text": "<fake_token_around_image>"
},
{
"id": 32001,
"logprob": -17.90625,
"logprob": -22.109375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.125,
"logprob": -18.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.296875,
"logprob": -20.125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.25,
"logprob": -17.125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.96875,
"logprob": -17.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.546875,
"logprob": -19.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.921875,
"logprob": -19.65625,
"text": "<image>"
},
{
......@@ -5649,212 +5714,212 @@
},
{
"id": 32001,
"logprob": -19.546875,
"logprob": -19.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.1875,
"logprob": -20.046875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.953125,
"logprob": -20.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.8359375,
"logprob": -15.8046875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.375,
"logprob": -16.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.171875,
"logprob": -19.953125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.03125,
"logprob": -18.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.71875,
"logprob": -21.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.65625,
"logprob": -20.515625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.453125,
"logprob": -19.171875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.484375,
"logprob": -19.296875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.875,
"logprob": -16.71875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.65625,
"logprob": -20.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.296875,
"logprob": -21.125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.546875,
"logprob": -20.40625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.78125,
"logprob": -20.734375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.96875,
"logprob": -17.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.5,
"logprob": -20.328125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.4140625,
"logprob": -15.2734375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.046875,
"logprob": -18.84375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.109375,
"logprob": -17.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.7265625,
"logprob": -15.578125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.5625,
"logprob": -18.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.734375,
"logprob": -20.625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.359375,
"logprob": -20.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.421875,
"logprob": -18.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.984375,
"logprob": -17.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.265625,
"logprob": -17.046875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.78125,
"logprob": -16.578125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.046875,
"logprob": -20.859375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.4375,
"logprob": -18.15625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.78125,
"logprob": -19.625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.671875,
"logprob": -19.546875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -14.2421875,
"logprob": -14.1953125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.015625,
"logprob": -18.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.40625,
"logprob": -18.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.59375,
"logprob": -17.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.671875,
"logprob": -20.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.265625,
"logprob": -20.0,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.578125,
"logprob": -20.359375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.25,
"logprob": -18.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.46875,
"logprob": -17.203125,
"text": "<image>"
},
{
......@@ -5864,317 +5929,322 @@
},
{
"id": 32001,
"logprob": -18.296875,
"logprob": -18.09375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.8671875,
"logprob": -15.71875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.375,
"logprob": -18.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.765625,
"logprob": -18.4375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.609375,
"logprob": -18.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.515625,
"logprob": -17.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.25,
"logprob": -16.265625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.640625,
"logprob": -19.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -14.8515625,
"logprob": -14.734375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.875,
"logprob": -20.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.28125,
"logprob": -20.21875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.453125,
"logprob": -18.359375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.25,
"logprob": -18.1875,
"text": "<image>"
},
{
"id": 32000,
"logprob": -3.2988281,
"text": "<fake_token_around_image>"
},
{
"id": 32001,
"logprob": -19.203125,
"logprob": -25.75,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.71875,
"logprob": -18.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.390625,
"logprob": -19.265625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.984375,
"logprob": -17.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.390625,
"logprob": -17.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.5,
"logprob": -20.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.296875,
"logprob": -17.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.4375,
"logprob": -16.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.015625,
"logprob": -19.65625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.359375,
"logprob": -18.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.375,
"logprob": -16.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.8125,
"logprob": -15.75,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.9375,
"logprob": -16.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.515625,
"logprob": -17.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.875,
"logprob": -16.5625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.09375,
"logprob": -18.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.265625,
"logprob": -18.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.234375,
"logprob": -16.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.28125,
"logprob": -18.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.25,
"logprob": -19.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.25,
"logprob": -19.890625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.765625,
"logprob": -20.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.609375,
"logprob": -19.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.359375,
"logprob": -20.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.90625,
"logprob": -19.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.28125,
"logprob": -19.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.5,
"logprob": -19.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.0625,
"logprob": -19.625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.125,
"logprob": -17.671875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.40625,
"logprob": -16.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.546875,
"logprob": -15.953125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.84375,
"logprob": -19.53125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.484375,
"logprob": -20.078125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.265625,
"logprob": -18.953125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.0625,
"logprob": -15.4453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.234375,
"logprob": -19.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.9453125,
"logprob": -15.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.046875,
"logprob": -20.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.515625,
"logprob": -20.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.375,
"logprob": -19.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.796875,
"logprob": -17.40625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.03125,
"logprob": -15.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.671875,
"logprob": -18.21875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.15625,
"logprob": -19.71875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.234375,
"logprob": -19.9375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.84375,
"logprob": -20.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.78125,
"logprob": -19.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.234375,
"logprob": -18.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.078125,
"logprob": -17.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.28125,
"logprob": -18.90625,
"text": "<image>"
},
{
......@@ -6184,472 +6254,482 @@
},
{
"id": 32001,
"logprob": -16.59375,
"logprob": -16.234375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.65625,
"logprob": -16.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.4375,
"logprob": -18.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.9375,
"logprob": -19.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.703125,
"logprob": -16.328125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.15625,
"logprob": -17.703125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.46875,
"logprob": -16.4375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.796875,
"logprob": -19.359375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.359375,
"logprob": -17.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.3125,
"logprob": -17.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.203125,
"logprob": -20.140625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.921875,
"logprob": -18.5625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.09375,
"logprob": -18.0,
"text": "<image>"
},
{
"id": 32000,
"logprob": -2.7207031,
"text": "<fake_token_around_image>"
},
{
"id": 32001,
"logprob": -18.6875,
"logprob": -23.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -22.625,
"logprob": -22.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.375,
"logprob": -21.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.765625,
"logprob": -18.578125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.46875,
"logprob": -18.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.546875,
"logprob": -20.296875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.875,
"logprob": -18.484375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.453125,
"logprob": -16.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.09375,
"logprob": -20.890625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.5625,
"logprob": -19.390625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.15625,
"logprob": -18.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.171875,
"logprob": -16.078125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.671875,
"logprob": -17.28125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.859375,
"logprob": -18.53125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.734375,
"logprob": -20.5,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.8125,
"logprob": -21.5,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.96875,
"logprob": -19.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.046875,
"logprob": -18.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.78125,
"logprob": -19.671875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.421875,
"logprob": -19.203125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.21875,
"logprob": -21.03125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.515625,
"logprob": -21.328125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.625,
"logprob": -19.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.734375,
"logprob": -20.609375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.046875,
"logprob": -18.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.1875,
"logprob": -19.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.1875,
"logprob": -20.9375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.828125,
"logprob": -19.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.359375,
"logprob": -18.0,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.75,
"logprob": -17.40625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.90625,
"logprob": -18.640625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.765625,
"logprob": -18.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.453125,
"logprob": -20.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.890625,
"logprob": -19.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.015625,
"logprob": -15.796875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.90625,
"logprob": -18.765625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.953125,
"logprob": -15.6640625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.46875,
"logprob": -21.3125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.984375,
"logprob": -19.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.875,
"logprob": -19.671875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.859375,
"logprob": -18.640625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.046875,
"logprob": -15.96875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.375,
"logprob": -20.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.140625,
"logprob": -19.8125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.140625,
"logprob": -20.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.6875,
"logprob": -21.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.453125,
"logprob": -21.25,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.1875,
"logprob": -19.09375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.765625,
"logprob": -17.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.65625,
"logprob": -19.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.078125,
"logprob": -17.09375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.109375,
"logprob": -16.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.171875,
"logprob": -18.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.4375,
"logprob": -20.171875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.0625,
"logprob": -20.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.734375,
"logprob": -16.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.21875,
"logprob": -18.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.421875,
"logprob": -16.4375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.015625,
"logprob": -19.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.796875,
"logprob": -17.59375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.3125,
"logprob": -17.171875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.390625,
"logprob": -20.34375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.28125,
"logprob": -19.171875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.59375,
"logprob": -18.578125,
"text": "<image>"
},
{
"id": 32000,
"logprob": -3.0917969,
"text": "<fake_token_around_image>"
},
{
"id": 32001,
"logprob": -18.8125,
"logprob": -25.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.09375,
"logprob": -18.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.890625,
"logprob": -20.6875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.09375,
"logprob": -17.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.75,
"logprob": -18.734375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.90625,
"logprob": -18.71875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.375,
"logprob": -21.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.640625,
"logprob": -16.734375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.859375,
"logprob": -20.875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.40625,
"logprob": -21.453125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.875,
"logprob": -15.796875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.1171875,
"logprob": -15.1328125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.078125,
"logprob": -17.125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.921875,
"logprob": -18.90625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.40625,
"logprob": -21.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.0,
"logprob": -21.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.75,
"logprob": -20.734375,
"text": "<image>"
},
{
......@@ -6659,7 +6739,7 @@
},
{
"id": 32001,
"logprob": -19.453125,
"logprob": -19.5,
"text": "<image>"
},
{
......@@ -6669,37 +6749,37 @@
},
{
"id": 32001,
"logprob": -22.421875,
"logprob": -22.515625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.9375,
"logprob": -20.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.671875,
"logprob": -19.703125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.890625,
"logprob": -21.0,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.921875,
"logprob": -16.984375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.5,
"logprob": -17.53125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.90625,
"logprob": -17.921875,
"text": "<image>"
},
{
......@@ -6709,42 +6789,42 @@
},
{
"id": 32001,
"logprob": -18.734375,
"logprob": -18.75,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.40625,
"logprob": -16.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.453125,
"logprob": -18.4375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.234375,
"logprob": -20.265625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -22.28125,
"logprob": -22.296875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.515625,
"logprob": -18.484375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.4296875,
"logprob": -15.390625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.765625,
"logprob": -19.75,
"text": "<image>"
},
{
......@@ -6754,32 +6834,32 @@
},
{
"id": 32001,
"logprob": -21.46875,
"logprob": -21.609375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.875,
"logprob": -18.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.859375,
"logprob": -20.828125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.078125,
"logprob": -17.015625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.4375,
"logprob": -16.40625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.015625,
"logprob": -21.046875,
"text": "<image>"
},
{
......@@ -6794,22 +6874,22 @@
},
{
"id": 32001,
"logprob": -21.484375,
"logprob": -21.515625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -20.015625,
"logprob": -20.0,
"text": "<image>"
},
{
"id": 32001,
"logprob": -18.84375,
"logprob": -18.78125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.421875,
"logprob": -16.375,
"text": "<image>"
},
{
......@@ -6819,57 +6899,57 @@
},
{
"id": 32001,
"logprob": -16.65625,
"logprob": -16.703125,
"text": "<image>"
},
{
"id": 32001,
"logprob": -13.640625,
"logprob": -13.625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.4140625,
"logprob": -15.375,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.546875,
"logprob": -17.515625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -21.859375,
"logprob": -21.921875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.65625,
"logprob": -15.640625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.484375,
"logprob": -16.46875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -16.359375,
"logprob": -16.421875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -19.9375,
"logprob": -19.890625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.875,
"logprob": -17.890625,
"text": "<image>"
},
{
"id": 32001,
"logprob": -17.453125,
"logprob": -17.40625,
"text": "<image>"
},
{
......@@ -6879,72 +6959,72 @@
},
{
"id": 32001,
"logprob": -19.171875,
"logprob": -19.1875,
"text": "<image>"
},
{
"id": 32001,
"logprob": -15.9921875,
"logprob": -15.9609375,
"text": "<image>"
},
{
"id": 32000,
"logprob": -2.0429688,
"logprob": -2.0332031,
"text": "<fake_token_around_image>"
},
{
"id": 12018,
"logprob": -12.03125,
"logprob": -12.078125,
"text": "Write"
},
{
"id": 528,
"logprob": -10.25,
"logprob": -10.109375,
"text": "me"
},
{
"id": 264,
"logprob": -0.10437012,
"logprob": -0.103515625,
"text": "a"
},
{
"id": 2485,
"logprob": -4.578125,
"logprob": -4.5664062,
"text": "short"
},
{
"id": 2838,
"logprob": -0.22924805,
"logprob": -0.23864746,
"text": "story"
},
{
"id": 32002,
"logprob": -10.84375,
"logprob": -10.9609375,
"text": "<end_of_utterance>"
},
{
"id": 259,
"logprob": -20.171875,
"logprob": -20.34375,
"text": " "
},
{
"id": 13,
"logprob": -8.765625,
"logprob": -8.5546875,
"text": "\n"
},
{
"id": 7226,
"logprob": -10.4140625,
"logprob": -10.484375,
"text": "Ass"
},
{
"id": 11143,
"logprob": -13.640625,
"logprob": -13.6015625,
"text": "istant"
},
{
"id": 28747,
"logprob": -0.005744934,
"logprob": -0.008308411,
"text": ":"
}
],
......@@ -6952,61 +7032,61 @@
"tokens": [
{
"id": 330,
"logprob": -0.12976074,
"logprob": -0.09448242,
"special": false,
"text": " A"
},
{
"id": 13088,
"logprob": -0.66308594,
"logprob": -0.6743164,
"special": false,
"text": " chicken"
},
{
"id": 349,
"logprob": -0.29541016,
"logprob": -0.31201172,
"special": false,
"text": " is"
},
{
"id": 6398,
"logprob": -0.05996704,
"logprob": -0.051635742,
"special": false,
"text": " sitting"
},
{
"id": 356,
"logprob": -0.27075195,
"logprob": -0.34033203,
"special": false,
"text": " on"
},
{
"id": 264,
"logprob": -0.14160156,
"logprob": -0.1194458,
"special": false,
"text": " a"
},
{
"id": 17972,
"logprob": -0.040863037,
"logprob": -0.032562256,
"special": false,
"text": " pile"
},
{
"id": 302,
"logprob": -0.00027036667,
"logprob": -0.00018763542,
"special": false,
"text": " of"
},
{
"id": 2445,
"logprob": -0.093322754,
"logprob": -0.07122803,
"special": false,
"text": " money"
},
{
"id": 28723,
"logprob": -0.006931305,
"logprob": -0.0041007996,
"special": false,
"text": "."
}
......
......@@ -8,61 +8,61 @@
"tokens": [
{
"id": 330,
"logprob": -0.13000488,
"logprob": -0.08660889,
"special": false,
"text": " A"
},
{
"id": 13088,
"logprob": -0.6713867,
"logprob": -0.7089844,
"special": false,
"text": " chicken"
},
{
"id": 349,
"logprob": -0.2980957,
"logprob": -0.32885742,
"special": false,
"text": " is"
},
{
"id": 6398,
"logprob": -0.060638428,
"logprob": -0.05126953,
"special": false,
"text": " sitting"
},
{
"id": 356,
"logprob": -0.27319336,
"logprob": -0.35229492,
"special": false,
"text": " on"
},
{
"id": 264,
"logprob": -0.140625,
"logprob": -0.12561035,
"special": false,
"text": " a"
},
{
"id": 17972,
"logprob": -0.040405273,
"logprob": -0.038085938,
"special": false,
"text": " pile"
},
{
"id": 302,
"logprob": -0.0002708435,
"logprob": -0.00018656254,
"special": false,
"text": " of"
},
{
"id": 2445,
"logprob": -0.095336914,
"logprob": -0.07293701,
"special": false,
"text": " money"
},
{
"id": 28723,
"logprob": -0.0068359375,
"logprob": -0.004852295,
"special": false,
"text": "."
}
......
......@@ -8,115 +8,115 @@
"tokens": [
{
"id": 415,
"logprob": -0.04421997,
"logprob": -0.039886475,
"special": false,
"text": " The"
},
{
"id": 12072,
"logprob": -0.13500977,
"logprob": -0.1430664,
"special": false,
"text": " cow"
},
{
"id": 349,
"logprob": -0.06750488,
"logprob": -0.056488037,
"special": false,
"text": " is"
},
{
"id": 6328,
"logprob": -0.6352539,
"logprob": -0.6855469,
"special": false,
"text": " standing"
},
{
"id": 356,
"logprob": -0.16186523,
"logprob": -0.1685791,
"special": false,
"text": " on"
},
{
"id": 272,
"logprob": -0.5078125,
"logprob": -0.50097656,
"special": false,
"text": " the"
},
{
"id": 10305,
"logprob": -0.017913818,
"logprob": -0.017303467,
"special": false,
"text": " beach"
},
{
"id": 304,
"logprob": -1.5205078,
"logprob": -1.3564453,
"special": false,
"text": " and"
},
{
"id": 272,
"logprob": -0.029174805,
"logprob": -0.017868042,
"special": false,
"text": " the"
},
{
"id": 13088,
"logprob": -0.003479004,
"logprob": -0.0027103424,
"special": false,
"text": " chicken"
},
{
"id": 349,
"logprob": -0.0035095215,
"logprob": -0.003156662,
"special": false,
"text": " is"
},
{
"id": 6398,
"logprob": -0.3088379,
"logprob": -0.37304688,
"special": false,
"text": " sitting"
},
{
"id": 356,
"logprob": -0.027755737,
"logprob": -0.034576416,
"special": false,
"text": " on"
},
{
"id": 264,
"logprob": -0.31884766,
"logprob": -0.29418945,
"special": false,
"text": " a"
},
{
"id": 17972,
"logprob": -0.047943115,
"logprob": -0.042877197,
"special": false,
"text": " pile"
},
{
"id": 302,
"logprob": -0.0002925396,
"logprob": -0.00028443336,
"special": false,
"text": " of"
},
{
"id": 2445,
"logprob": -0.02935791,
"logprob": -0.023223877,
"special": false,
"text": " money"
},
{
"id": 28723,
"logprob": -0.031219482,
"logprob": -0.018157959,
"special": false,
"text": "."
},
{
"id": 32002,
"logprob": -0.00034475327,
"logprob": -0.00018393993,
"special": true,
"text": "<end_of_utterance>"
},
......
......@@ -22,6 +22,7 @@ text-generation-client = { path = "client" }
clap = { version = "4.4.5", features = ["derive", "env"] }
futures = "0.3.28"
hf-hub = { workspace = true }
itertools = "0.10"
jsonschema = { version = "0.17.1", features = ["draft202012"] }
metrics = "0.21.1"
metrics-exporter-prometheus = { version = "0.15.1", features = [] }
......
......@@ -71,10 +71,12 @@ fn get_unpadded_features(
let current_aspect_ratio: f64 = current_width as f64 / current_height as f64;
let (current_height, current_width) = if aspect_ratio > current_aspect_ratio {
let new_height = (height * current_width) / width;
(new_height, current_width)
let padding = (current_height - new_height) / 2;
(current_height - (2 * padding), current_width)
} else {
let new_width = (width * current_height) / height;
(current_height, new_width)
let padding = (current_width - new_width) / 2;
(current_height, current_width - (2 * padding))
};
let unpadded_features = current_height * current_width;
......@@ -88,7 +90,9 @@ impl LlavaNext {
let patch_size = self.vision_config.patch_size;
assert!(image_size % patch_size == 0);
let npatches = image_size / patch_size;
let (num_patch_height, num_patch_width) =
// Dimensions are intentionally swapped to be bug-compatible with
// upstream: https://github.com/LLaVA-VL/LLaVA-NeXT/issues/59
let (num_patch_width, num_patch_height) =
get_anyres_image_grid_shape(height, width, &self.image_grid_pinpoints, image_size);
let (unpadded_features, newline_features) =
......@@ -112,7 +116,7 @@ pub struct Idefics2 {}
impl Idefics2 {
pub fn get_number_of_features(&self, _height: usize, _width: usize) -> usize {
320
64
}
}
......
......@@ -70,6 +70,25 @@ impl HubTokenizerConfig {
}
}
#[derive(Debug, Clone, Serialize, Deserialize)]
#[serde(tag = "processor_class")]
pub enum HubPreprocessorConfig {
Idefics2Processor(Idefics2Preprocessor),
}
impl HubPreprocessorConfig {
pub fn from_file<P: AsRef<std::path::Path>>(filename: P) -> Option<Self> {
let content = std::fs::read_to_string(filename).ok()?;
serde_json::from_str(&content).ok()
}
}
#[derive(Clone, Debug, Serialize, Deserialize)]
pub struct Idefics2Preprocessor {
#[serde(default)]
do_image_splitting: bool,
}
#[derive(Debug, Clone, Deserialize, Default)]
pub struct HubProcessorConfig {
pub chat_template: Option<ChatTemplateVersions>,
......
......@@ -13,7 +13,9 @@ use std::io::BufReader;
use std::net::{IpAddr, Ipv4Addr, SocketAddr};
use std::path::{Path, PathBuf};
use text_generation_router::config::Config;
use text_generation_router::{server, HubModelInfo, HubProcessorConfig, HubTokenizerConfig};
use text_generation_router::{
server, HubModelInfo, HubPreprocessorConfig, HubProcessorConfig, HubTokenizerConfig,
};
use thiserror::Error;
use tokenizers::Tokenizer;
use tower_http::cors::AllowOrigin;
......@@ -214,6 +216,7 @@ async fn main() -> Result<(), RouterError> {
tokenizer_filename,
config_filename,
tokenizer_config_filename,
preprocessor_config_filename,
processor_config_filename,
model_info,
) = match api {
......@@ -221,6 +224,7 @@ async fn main() -> Result<(), RouterError> {
Some(local_path.join("tokenizer.json")),
Some(local_path.join("config.json")),
Some(local_path.join("tokenizer_config.json")),
Some(local_path.join("preprocessor_config.json")),
Some(local_path.join("processor_config.json")),
None,
),
......@@ -237,6 +241,7 @@ async fn main() -> Result<(), RouterError> {
};
let config_filename = api_repo.get("config.json").await.ok();
let tokenizer_config_filename = api_repo.get("tokenizer_config.json").await.ok();
let preprocessor_config_filename = api_repo.get("preprocessor_config.json").await.ok();
let processor_config_filename = api_repo.get("processor_config.json").await.ok();
let model_info = if let Some(model_info) = get_model_info(&api_repo).await {
......@@ -249,6 +254,7 @@ async fn main() -> Result<(), RouterError> {
tokenizer_filename,
config_filename,
tokenizer_config_filename,
preprocessor_config_filename,
processor_config_filename,
model_info,
)
......@@ -263,6 +269,7 @@ async fn main() -> Result<(), RouterError> {
repo.get("tokenizer.json"),
repo.get("config.json"),
repo.get("tokenizer_config.json"),
repo.get("preprocessor_config.json"),
repo.get("processor_config.json"),
None,
)
......@@ -300,6 +307,8 @@ async fn main() -> Result<(), RouterError> {
HubTokenizerConfig::default()
});
let preprocessor_config =
preprocessor_config_filename.and_then(HubPreprocessorConfig::from_file);
let processor_config = processor_config_filename
.and_then(HubProcessorConfig::from_file)
.unwrap_or_default();
......@@ -361,6 +370,7 @@ async fn main() -> Result<(), RouterError> {
ngrok_authtoken,
ngrok_edge,
tokenizer_config,
preprocessor_config,
processor_config,
messages_api_enabled,
disable_grammar_support,
......
......@@ -12,9 +12,9 @@ use crate::kserve::{
use crate::validation::ValidationError;
use crate::{
BestOfSequence, Details, ErrorResponse, FinishReason, GenerateParameters, GenerateRequest,
GenerateResponse, GrammarType, HubModelInfo, HubProcessorConfig, HubTokenizerConfig, Info,
Message, PrefillToken, SimpleToken, StreamDetails, StreamResponse, Token, TokenizeResponse,
Usage, Validation,
GenerateResponse, GrammarType, HubModelInfo, HubPreprocessorConfig, HubProcessorConfig,
HubTokenizerConfig, Info, Message, PrefillToken, SimpleToken, StreamDetails, StreamResponse,
Token, TokenizeResponse, Usage, Validation,
};
use crate::{
ChatCompletion, ChatCompletionChoice, ChatCompletionChunk, ChatCompletionComplete,
......@@ -1423,6 +1423,7 @@ pub async fn run(
_ngrok_authtoken: Option<String>,
_ngrok_edge: Option<String>,
tokenizer_config: HubTokenizerConfig,
preprocessor_config: Option<HubPreprocessorConfig>,
processor_config: HubProcessorConfig,
messages_api_enabled: bool,
grammar_support: bool,
......@@ -1636,6 +1637,7 @@ pub async fn run(
validation_workers,
tokenizer,
config,
preprocessor_config,
max_best_of,
max_stop_sequences,
max_top_n_tokens,
......
/// Payload validation logic
use crate::config::Config;
use crate::validation::ValidationError::{BestOfSampling, BestOfSeed, EmptyInput};
use crate::{GenerateParameters, GenerateRequest, GrammarType};
use crate::{
GenerateParameters, GenerateRequest, GrammarType, HubPreprocessorConfig, Idefics2Preprocessor,
};
use base64::{engine::general_purpose::STANDARD, Engine};
use image::{io::Reader as ImageReader, ImageFormat};
use jsonschema::{Draft, JSONSchema};
use rand::{thread_rng, Rng};
use serde_json::Value;
use std::io::Cursor;
use std::iter;
use text_generation_client::{Chunk, Image, InputChunk};
use thiserror::Error;
use tokenizers::tokenizer::Tokenizer;
......@@ -36,6 +39,7 @@ impl Validation {
workers: usize,
tokenizer: Option<Tokenizer>,
config: Option<Config>,
preprocessor_config: Option<HubPreprocessorConfig>,
max_best_of: usize,
max_stop_sequences: usize,
max_top_n_tokens: u32,
......@@ -53,12 +57,18 @@ impl Validation {
for _ in 0..workers {
let tokenizer_clone = tokenizer.clone();
let config_clone = config.clone();
let preprocessor_config_clone = preprocessor_config.clone();
let (tokenizer_sender, tokenizer_receiver) = mpsc::unbounded_channel();
senders.push(tokenizer_sender);
// Spawn worker
tokio::task::spawn_blocking(move || {
tokenizer_worker(tokenizer_clone, config_clone, tokenizer_receiver)
tokenizer_worker(
tokenizer_clone,
config_clone,
preprocessor_config_clone,
tokenizer_receiver,
)
});
}
......@@ -422,13 +432,20 @@ async fn round_robin_task(
fn tokenizer_worker(
tokenizer: Tokenizer,
config: Option<Config>,
preprocessor_config: Option<HubPreprocessorConfig>,
mut receiver: mpsc::UnboundedReceiver<TokenizerRequest>,
) {
// Loop over requests
while let Some(((inputs, truncate), response_tx, parent_span)) = receiver.blocking_recv() {
parent_span.in_scope(|| {
response_tx
.send(prepare_input(inputs, truncate, &tokenizer, &config))
.send(prepare_input(
inputs,
truncate,
&tokenizer,
config.as_ref(),
preprocessor_config.as_ref(),
))
.unwrap_or(())
})
}
......@@ -508,16 +525,67 @@ fn fetch_image(input: &str) -> Result<(Vec<u8>, String, usize, usize), Validatio
}
}
fn image_tokens(
config: &Config,
preprocessor_config: Option<&HubPreprocessorConfig>,
height: usize,
width: usize,
) -> String {
use Config::*;
use HubPreprocessorConfig::*;
match config {
Idefics => "<image>".to_string(),
Idefics2(config) => {
const FAKE: &str = "<fake_token_around_image>";
const IMAGE: &str = "<image>";
let slots = config.get_number_of_features(height, width);
let mut image_string = String::with_capacity(2 * FAKE.len() + slots * IMAGE.len());
image_string.push_str(FAKE);
image_string.extend(iter::repeat(IMAGE).take(slots));
image_string.push_str(FAKE);
if matches!(
preprocessor_config,
Some(Idefics2Processor(Idefics2Preprocessor {
do_image_splitting: true,
..
}))
) {
image_string = image_string.repeat(5);
};
image_string
}
Paligemma(config) => "<image>".repeat(config.get_number_of_features(height, width)),
LlavaNext(config) => "<image>".repeat(config.get_number_of_features(height, width)),
_ => unimplemented!("Images tokens are not supported for this model configuration"),
}
}
fn image_tokens_fixup(config: &Config, text: String) -> String {
match config {
Config::Idefics2(_) => {
const FAKE: &str = "<fake_token_around_image>";
text.replace(&format!("{FAKE}{FAKE}"), FAKE)
}
_ => text,
}
}
/// Get input length and optionally truncate it
fn prepare_input(
inputs: String,
_truncate: Option<usize>,
tokenizer: &Tokenizer,
config: &Option<Config>,
config: Option<&Config>,
preprocessor_config: Option<&HubPreprocessorConfig>,
) -> Result<(tokenizers::Encoding, Vec<InputChunk>), ValidationError> {
use Config::*;
static RE: Lazy<Regex> = Lazy::new(|| Regex::new(r"!\[\]\([^\)]*\)").unwrap());
let (tokenizer_query, input_chunks) = match config {
Some(Config::LlavaNext(config)) => {
Some(config @ (Idefics | Idefics2(_) | Paligemma(_) | LlavaNext(_))) => {
let mut input_chunks = Vec::new();
let mut tokenizer_query = String::with_capacity(inputs.len());
let mut start = 0;
......@@ -529,88 +597,17 @@ fn prepare_input(
tokenizer_query.push_str(&inputs[start..chunk_start]);
}
let (data, mimetype, height, width) = fetch_image(&inputs[chunk_start..chunk_end])?;
let slots = config.get_number_of_features(height, width);
input_chunks.push(Chunk::Image(Image { data, mimetype }).into());
tokenizer_query.push_str(&"<image>".repeat(slots));
tokenizer_query.push_str(&image_tokens(config, preprocessor_config, height, width));
start = chunk_end;
}
if start != inputs.len() {
input_chunks.push(Chunk::Text(inputs[start..].to_string()).into());
tokenizer_query.push_str(&inputs[start..]);
}
(tokenizer_query, input_chunks)
}
Some(Config::Paligemma(config)) => {
let mut input_chunks = Vec::new();
let mut tokenizer_query = String::with_capacity(inputs.len());
let mut start = 0;
for chunk in RE.find_iter(&inputs) {
let chunk_start = chunk.start();
let chunk_end = chunk.end();
if chunk_start != start {
input_chunks.push(Chunk::Text(inputs[start..chunk_start].to_string()).into());
tokenizer_query.push_str(&inputs[start..chunk_start]);
}
let (data, mimetype, height, width) = fetch_image(&inputs[chunk_start..chunk_end])?;
let slots = config.get_number_of_features(height, width);
input_chunks.push(Chunk::Image(Image { data, mimetype }).into());
tokenizer_query.push_str(&"<image>".repeat(slots));
start = chunk_end;
}
if start != inputs.len() {
input_chunks.push(Chunk::Text(inputs[start..].to_string()).into());
tokenizer_query.push_str(&inputs[start..]);
}
(tokenizer_query, input_chunks)
}
Some(Config::Idefics2(config)) => {
let mut input_chunks = Vec::new();
let mut tokenizer_query = String::with_capacity(inputs.len());
let mut start = 0;
for chunk in RE.find_iter(&inputs) {
let chunk_start = chunk.start();
let chunk_end = chunk.end();
if chunk_start != start {
input_chunks.push(Chunk::Text(inputs[start..chunk_start].to_string()).into());
tokenizer_query.push_str(&inputs[start..chunk_start]);
}
let (data, mimetype, height, width) = fetch_image(&inputs[chunk_start..chunk_end])?;
let slots = config.get_number_of_features(height, width);
tokenizer_query.push_str("<fake_token_around_image>");
tokenizer_query.push_str(&"<image>".repeat(slots));
tokenizer_query.push_str("<fake_token_around_image>");
input_chunks.push(Chunk::Image(Image { data, mimetype }).into());
start = chunk_end;
}
if start != inputs.len() {
input_chunks.push(Chunk::Text(inputs[start..].to_string()).into());
tokenizer_query.push_str(&inputs[start..]);
}
(tokenizer_query, input_chunks)
}
Some(Config::Idefics) => {
let mut input_chunks = Vec::new();
let mut tokenizer_query = String::with_capacity(inputs.len());
let mut start = 0;
for chunk in RE.find_iter(&inputs) {
let chunk_start = chunk.start();
let chunk_end = chunk.end();
if chunk_start != start {
input_chunks.push(Chunk::Text(inputs[start..chunk_start].to_string()).into());
tokenizer_query.push_str(&inputs[start..chunk_start]);
}
let (data, mimetype, _height, _width) =
fetch_image(&inputs[chunk_start..chunk_end])?;
let slots = 1;
tokenizer_query.push_str(&"<image>".repeat(slots));
input_chunks.push(Chunk::Image(Image { data, mimetype }).into());
start = chunk_end;
}
if start != inputs.len() {
input_chunks.push(Chunk::Text(inputs[start..].to_string()).into());
tokenizer_query.push_str(&inputs[start..]);
}
tokenizer_query = image_tokens_fixup(config, tokenizer_query);
(tokenizer_query, input_chunks)
}
_ => (inputs.clone(), vec![Chunk::Text(inputs).into()]),
......@@ -750,7 +747,7 @@ pub enum ValidationError {
#[cfg(test)]
mod tests {
use super::*;
use crate::config::{PaliTextConfig, Paligemma};
use crate::config::{Idefics2, PaliTextConfig, Paligemma};
use crate::default_parameters;
use crate::tests::get_tokenizer;
......@@ -769,6 +766,7 @@ mod tests {
workers,
tokenizer,
config,
None,
max_best_of,
max_stop_sequence,
max_top_n_tokens,
......@@ -803,6 +801,7 @@ mod tests {
workers,
tokenizer,
config,
None,
max_best_of,
max_stop_sequence,
max_top_n_tokens,
......@@ -836,6 +835,7 @@ mod tests {
workers,
tokenizer,
config,
None,
max_best_of,
max_stop_sequence,
max_top_n_tokens,
......@@ -874,6 +874,7 @@ mod tests {
workers,
tokenizer,
config,
None,
max_best_of,
max_stop_sequence,
max_top_n_tokens,
......@@ -941,6 +942,7 @@ mod tests {
workers,
tokenizer,
config,
None,
max_best_of,
max_stop_sequences,
max_top_n_tokens,
......@@ -1026,6 +1028,7 @@ mod tests {
workers,
tokenizer,
Some(config),
None,
max_best_of,
max_stop_sequence,
max_top_n_tokens,
......@@ -1058,4 +1061,83 @@ mod tests {
"Failed to process images",
);
}
#[tokio::test]
async fn test_idefics2_correct_n_fake_tokens() {
let pixel_data = STANDARD.decode(PIXEL_GIF).unwrap();
let tokenizer = Some(get_tokenizer().await);
let max_best_of = 2;
let max_stop_sequence = 3;
let max_top_n_tokens = 4;
let max_input_length = 5;
let max_total_tokens = 6;
let disable_grammar_support = true;
let workers = 1;
let config = Config::Idefics2(Idefics2 {});
let validation = Validation::new(
workers,
tokenizer,
Some(config),
Some(HubPreprocessorConfig::Idefics2Processor(
Idefics2Preprocessor {
do_image_splitting: true,
},
)),
max_best_of,
max_stop_sequence,
max_top_n_tokens,
max_input_length,
max_total_tokens,
disable_grammar_support,
);
let (encoding, chunks) = match validation
.tokenize(
format!(
"test![](data:image/gif;base64,{})![](data:image/gif;base64,{})",
PIXEL_GIF, PIXEL_GIF
),
None,
)
.await
{
Ok(Some((encoding, chunks))) => (encoding, chunks),
_ => panic!("Unexpected tokenization failure"),
};
assert!(
chunks
== vec![
Chunk::Text("test".to_string()).into(),
Chunk::Image(Image {
data: pixel_data.clone(),
mimetype: "image/gif".to_string()
})
.into(),
Chunk::Image(Image {
data: pixel_data.clone(),
mimetype: "image/gif".to_string()
})
.into()
],
"Failed to process images",
);
// Verify the number of fake tokens:
//
// - Two images surrounded/separated by a fake token = 3.
// - Both are split in 5 subimages, separated by a fake token: 2 * 4
//
// Fake tokens get split up by the testing tokenizer, but we don't care.
assert_eq!(
encoding
.get_tokens()
.iter()
.filter(|t| *t == "fake")
.count(),
11
);
}
}
......@@ -39,7 +39,7 @@ def get_anyres_image_grid_shape(image_size, grid_pinpoints, patch_size):
Args:
image_size (`tuple`):
The size of the input image in the format (width, height).
The size of the input image in the format (height, width).
grid_pinpoints (`List`):
A list containing possible resolutions. Each item in the list should be a tuple or list
of the form `(height, width)`.
......@@ -47,7 +47,7 @@ def get_anyres_image_grid_shape(image_size, grid_pinpoints, patch_size):
The size of each image patch.
Returns:
tuple: The shape of the image patch grid in the format (width, height).
tuple: The shape of the image patch grid in the format (height, width).
"""
if not isinstance(grid_pinpoints, list):
raise ValueError("grid_pinpoints should be a list of tuples or lists")
......@@ -230,7 +230,10 @@ class LlavaNextForConditionalGeneration(nn.Module):
raise ValueError(
"The number of patches is not consistent with the image size."
)
num_patch_height, num_patch_width = get_anyres_image_grid_shape(
# Dimensions are intentionally swapped to be bug-compatible with
# upstream: https://github.com/LLaVA-VL/LLaVA-NeXT/issues/59
num_patch_width, num_patch_height = get_anyres_image_grid_shape(
image_sizes[image_idx],
self.config.image_grid_pinpoints,
self.config.vision_config.image_size,
......
......@@ -39,7 +39,9 @@ class PaliGemmaBatch(VlmCausalLMBatch):
# TODO do_convert_RGB should be on by default ?
image = image.convert("RGB")
image_input = processor.image_processor(image, return_tensors="pt")
full_text += image_text_replacement(image_input, config, image_id)
full_text += image_text_replacement(
processor, image_input, config, image_id
)
image_inputs.append(image_input)
else:
raise RuntimeError(f"Invalid chunk type {chunk_type}")
......
from itertools import repeat
import torch
from PIL import Image
from io import BytesIO
......@@ -15,6 +16,9 @@ from text_generation_server.models.flash_mistral import (
tracer = trace.get_tracer(__name__)
IDEFICS2_FAKE_TOKEN = "<fake_token_around_image>"
IDEFICS2_IMAGE_TOKEN = "<image>"
def get_anyres_image_grid_shape(image_size, grid_pinpoints, patch_size):
"""
......@@ -22,7 +26,7 @@ def get_anyres_image_grid_shape(image_size, grid_pinpoints, patch_size):
Args:
image_size (`tuple`):
The size of the input image in the format (width, height).
The size of the input image in the format (height, width).
grid_pinpoints (`List`):
A list containing possible resolutions. Each item in the list should be a tuple or list
of the form `(height, width)`.
......@@ -39,15 +43,13 @@ def get_anyres_image_grid_shape(image_size, grid_pinpoints, patch_size):
return height // patch_size, width // patch_size
def image_text_replacement(image_input, config, image_id) -> str:
def image_text_replacement(processor, image_input, config, image_id: int) -> str:
if config.model_type == "idefics2":
# TODO technically depends on image splitting which is not implemented.
num_features = 320
return (
"<fake_token_around_image>"
+ "<image>" * num_features
+ "<fake_token_around_image>"
)
image_seq_len = 64
image_str = f"{IDEFICS2_FAKE_TOKEN}{IDEFICS2_IMAGE_TOKEN * image_seq_len}{IDEFICS2_FAKE_TOKEN}"
if processor.image_processor.do_image_splitting:
image_str *= 5
return image_str
elif config.model_type == "llava_next":
height, width = image_input["image_sizes"][image_id]
num_features = get_number_of_features(height, width, config)
......@@ -64,20 +66,35 @@ def image_text_replacement(image_input, config, image_id) -> str:
raise RuntimeError(f"Unknown config {config.model_type} for multimodal")
def image_text_replacement_fixup(config, text: str) -> str:
if config.model_type == "idefics2":
return text.replace(
f"{IDEFICS2_FAKE_TOKEN}{IDEFICS2_FAKE_TOKEN}", IDEFICS2_FAKE_TOKEN
)
return text
def get_unpadded_features(
height: int, width: int, npatches: int, num_patch_height: int, num_patch_width: int
original_height: int,
original_width: int,
npatches: int,
num_patch_height: int,
num_patch_width: int,
) -> Tuple[int, int]:
current_height = npatches * num_patch_height
current_width = npatches * num_patch_width
aspect_ratio: float = width / height
aspect_ratio: float = original_width / original_height
current_aspect_ratio: float = current_width / current_height
if aspect_ratio > current_aspect_ratio:
new_height = (height * current_width) // width
current_height = new_height
new_height = (original_height * current_width) // original_width
padding = (current_height - new_height) // 2
current_height = current_height - (2 * padding)
else:
new_width = (width * current_height) // height
current_width = new_width
new_width = (original_width * current_height) // original_height
padding = (current_width - new_width) // 2
current_width = current_width - (2 * padding)
unpadded_features = current_height * current_width
newline_features = current_height
......@@ -96,7 +113,9 @@ def get_number_of_features(height: int, width: int, config) -> int:
npatches = image_size // patch_size
num_patch_height, num_patch_width = get_anyres_image_grid_shape(
# Dimensions are intentionally swapped to be bug-compatible with
# upstream: https://github.com/LLaVA-VL/LLaVA-NeXT/issues/59
num_patch_width, num_patch_height = get_anyres_image_grid_shape(
[height, width],
image_grid_pinpoints,
image_size,
......@@ -168,9 +187,13 @@ class VlmCausalLMBatch(FlashCausalLMBatch):
if chunk_type == "text":
full_text += chunk.text
elif chunk_type == "image":
full_text += image_text_replacement(image_inputs, config, image_id)
full_text += image_text_replacement(
processor, image_inputs, config, image_id
)
image_id += 1
full_text = image_text_replacement_fixup(config, full_text)
batch_inputs.append(full_text)
max_truncation = max(max_truncation, r.truncate)
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment