Download gptq_marlin/kernel.h from RedHatAI/quantization: direct link, hf CLI and curl.
- Browser
- Download file 1.93 kB
-
https://huggingface.co/RedHatAI/quantization/resolve/main/gptq_marlin/kernel.h
- Command line
-
hf download hf://RedHatAI/quantization/gptq_marlin/kernel.h
-
curl -L -o kernel.h https://huggingface.co/RedHatAI/quantization/resolve/main/gptq_marlin/kernel.h
1.93 kB
| namespace MARLIN_NAMESPACE_NAME { | |
| template <typename scalar_t, // compute dtype, half or nv_float16 | |
| const vllm::ScalarTypeId w_type_id, // weight ScalarType id | |
| const int threads, // number of threads in a threadblock | |
| const int thread_m_blocks, // number of 16x16 blocks in the m | |
| // dimension (batchsize) of the | |
| // threadblock | |
| const int thread_n_blocks, // same for n dimension (output) | |
| const int thread_k_blocks, // same for k dimension (reduction) | |
| const bool m_block_size_8, // whether m_block_size == 8 | |
| // only works when thread_m_blocks == 1 | |
| const int stages, // number of stages for the async global->shared | |
| // fetch pipeline | |
| const int group_blocks, // number of consecutive 16x16 blocks | |
| // with a separate quantization scale | |
| const bool is_zp_float // is zero point of float16 type? | |
| > | |
| __global__ void Marlin(MARLIN_KERNEL_PARAMS); | |
| } | |