llama: update vendored code to commit 46e3556 (#8308)

1deafd82 · Jeffrey Morgan · GitHub · 57f038ec · 1deafd82 · 1deafd82
Unverified Commit 1deafd82 authored Jan 08, 2025 by Jeffrey Morgan Committed by GitHub Jan 08, 2025
20 changed files
--- a/llama/ggml-cuda/common.cuh
+++ b/llama/ggml-cuda/common.cuh
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *

--- a/llama/ggml-cuda/concat.cu
+++ b/llama/ggml-cuda/concat.cu
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *

--- a/llama/ggml-cuda/concat.cuh
+++ b/llama/ggml-cuda/concat.cuh
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *

--- a/llama/ggml-cuda/conv-transpose-1d.cu
+++ b/llama/ggml-cuda/conv-transpose-1d.cu
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *

--- a/llama/ggml-cuda/conv-transpose-1d.cuh
+++ b/llama/ggml-cuda/conv-transpose-1d.cuh
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *

--- a/llama/ggml-cuda/convert.cu
+++ b/llama/ggml-cuda/convert.cu
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *
@@ -706,6 +706,8 @@ to_fp32_cuda_t ggml_get_to_fp32_cuda(ggml_type type) {
            return dequantize_row_iq3_s_cuda;
        case GGML_TYPE_F16:
            return convert_unary_cuda<half>;
+        case GGML_TYPE_BF16:
+            return convert_unary_cuda<nv_bfloat16>;
        default:
            return nullptr;
    }

--- a/llama/ggml-cuda/convert.cuh
+++ b/llama/ggml-cuda/convert.cuh
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *

--- a/llama/ggml-cuda/count-equal.cu
+++ b/llama/ggml-cuda/count-equal.cu
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *

--- a/llama/ggml-cuda/count-equal.cuh
+++ b/llama/ggml-cuda/count-equal.cuh
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *

--- a/llama/ggml-cuda/cpy.cu
+++ b/llama/ggml-cuda/cpy.cu
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *

--- a/llama/ggml-cuda/cpy.cuh
+++ b/llama/ggml-cuda/cpy.cuh
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *

--- a/llama/ggml-cuda/cross-entropy-loss.cu
+++ b/llama/ggml-cuda/cross-entropy-loss.cu
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *

--- a/llama/ggml-cuda/cross-entropy-loss.cuh
+++ b/llama/ggml-cuda/cross-entropy-loss.cuh
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *

--- a/llama/ggml-cuda/dequantize.cuh
+++ b/llama/ggml-cuda/dequantize.cuh
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *

--- a/llama/ggml-cuda/diagmask.cu
+++ b/llama/ggml-cuda/diagmask.cu
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *

--- a/llama/ggml-cuda/diagmask.cuh
+++ b/llama/ggml-cuda/diagmask.cuh
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *

--- a/llama/ggml-cuda/fattn-common.cuh
+++ b/llama/ggml-cuda/fattn-common.cuh
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *

--- a/llama/ggml-cuda/fattn-tile-f16.cu
+++ b/llama/ggml-cuda/fattn-tile-f16.cu
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *

--- a/llama/ggml-cuda/fattn-tile-f16.cuh
+++ b/llama/ggml-cuda/fattn-tile-f16.cuh
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *

--- a/llama/ggml-cuda/fattn-tile-f32.cu
+++ b/llama/ggml-cuda/fattn-tile-f32.cu
 /**
- * llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
+ * llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
 *
 * MIT License
 *