Niko1221 828a2559a9 vision: the image-token limit has a default and is enforced (I7, #3690) ...

The reader was clamped only when the config carried vision.max_tokens; an old, hand-edited or migrated config left it unclamped and
one picture became 93,700 image tokens. Vision.token_limit gives every reader a limit (GPU 1024, CPU 300, the numbers setup writes)
unless the config names one (a value the person set is kept, 0 still means none). strata-vision refuses a picture that comes back
over the limit plus an eighth with a plain ERR instead of passing 90 times the budget on, and prints a WARNING when its own warm-up
picture is over it. The server logs a picture that is more than 4 times the limit with the numbers to send with a report.

2026-10-22 13:10:12 +02:00

..

CMakeLists.txt

Fix #3359: the HIP picture reader is built position independent so it links on Linux

2026-10-20 14:13:34 +02:00

gguf_meta_test.cpp

strata-vision reads only the vocabulary and header of the text model; Gyro reads pictures

2026-10-18 12:14:19 +02:00

gguf_meta.h

strata-vision reads only the vocabulary and header of the text model; Gyro reads pictures

2026-10-18 12:14:19 +02:00

gpu_arch_test.cpp

strata-vision: HIP build (strata-vision-hip) for AMD GPUs; the reader refuses a GPU that is not in its build; RUNPATH $ORIGIN

2026-10-18 21:31:30 +02:00

gpu_arch.h

strata-vision: HIP build (strata-vision-hip) for AMD GPUs; the reader refuses a GPU that is not in its build; RUNPATH $ORIGIN

2026-10-18 21:31:30 +02:00

strata_vision.cpp

vision: the image-token limit has a default and is enforced (I7, #3690)

2026-10-22 13:10:12 +02:00

strata-utf8.manifest

Windows with a non-UTF-8 code page (Russian): PowerShell output is read as UTF-8 (Documents folder, save dialog, fonts, GPU names), Python runs in UTF-8 mode from the launchers and the host's children, and the engine and picture reader get a UTF-8 manifest so a Cyrillic model folder arrives intact (#3202)

2026-10-19 23:38:49 +02:00