From 5dc9dd7152dedc6046b646855585bd070c91e8c8 Mon Sep 17 00:00:00 2001 From: Carolinabanana <140120812+Carolinabanana@users.noreply.github.com> Date: Tue, 9 Apr 2024 09:16:13 +0100 Subject: llama : add Command R Plus support (#6491) * Add Command R Plus GGUF * Add Command R Plus GGUF * Loading works up to LayerNorm2D * Export new tensors in 1D so they are not quantized. * Fix embedding layer based on Noeda's example * Whitespace * Add line * Fix unexpected tokens on MPS. Re-add F16 fix. ((Noeda) * dranger003: Fix block index overflow in CUDA dequantizing. * Reverted blocked multiplication code as it still has issues and could affect other Llama arches * export norms as f32 * fix overflow issues during quant and other cleanup * Type convention Co-authored-by: Georgi Gerganov * dranger003: Fix more int overflow during quant. --------- Co-authored-by: S Co-authored-by: S Co-authored-by: slaren Co-authored-by: Georgi Gerganov --- ggml-cuda/quantize.cuh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'ggml-cuda/quantize.cuh') diff --git a/ggml-cuda/quantize.cuh b/ggml-cuda/quantize.cuh index adb89c83..b37a4752 100644 --- a/ggml-cuda/quantize.cuh +++ b/ggml-cuda/quantize.cuh @@ -2,4 +2,4 @@ #define CUDA_QUANTIZE_BLOCK_SIZE 256 -void quantize_row_q8_1_cuda(const float * x, void * vy, const int kx, const int ky, const int kx_padded, cudaStream_t stream); +void quantize_row_q8_1_cuda(const float * x, void * vy, const int64_t kx, const int64_t ky, const int64_t kx_padded, cudaStream_t stream); -- cgit v1.2.3