Jesse Gross c2f5d6662b ollamarunner: Re-enable worst case graph preallocation. ...

Worst case graph preallocation was disabled by a27462b
"ollamarunner: Temporarily disable worst case graph preallocation"
since it caused crashes with large batches when not using the GPU.

This backports upstream llama.cpp commit f057808
"ggml: Don't assert fail when tensor data changes (#13222)", which
fixes the underlying bug and allows reverting the previous workaround.

2025-05-02 12:22:47 -07:00

..

0001-ggml-backend-malloc-and-free-using-the-same-compiler.patch

…

0002-pretokenizer.patch

…

0003-embeddings.patch

…

0004-clip-unicode.patch

…

0005-solar-pro.patch

…

0006-add-mllama-support.patch

…

0007-add-unpad-operator.patch

…

0008-fix-deepseek-deseret-regex.patch

…

0009-maintain-ordering-for-rules-for-grammar.patch

…

0010-ensure-KV-cache-is-fully-defragmented.patch

…

0011-sort-devices-by-score.patch

…

0012-add-phony-target-ggml-cpu-for-all-cpu-variants.patch

…

0013-remove-amx.patch

…

0014-fix-string-arr-kv-loading.patch

…

0015-ollama-debug-tensor.patch

…

0016-add-model-quantizations.patch

…

0017-add-ollama-vocab-for-grammar-support.patch

…

0018-ggml-Don-t-assert-fail-when-tensor-data-changes-1322.patch

…