mirror of
https://github.com/LostRuins/koboldcpp.git
synced 2025-09-11 01:24:36 +00:00
Merge branch 'master' into concedo_experimental
# Conflicts: # .github/workflows/build.yml # .github/workflows/editorconfig.yml # .gitignore # CMakeLists.txt # README.md
This commit is contained in:
Concedo
commit
15deabd200
15 changed files with 1060 additions and 1371 deletions
|
|
@ -8,10 +8,14 @@
|
|||
|
||||
[Linux](#linux)
|
||||
|
||||
[Windows](#windows)
|
||||
|
||||
[Environment Variable](#environment-variable)
|
||||
|
||||
[Known Issue](#known-issue)
|
||||
|
||||
[Q&A](#q&a)
|
||||
|
||||
[Todo](#todo)
|
||||
|
||||
## Background
|
||||
|
|
@ -33,7 +37,7 @@ For Intel CPU, recommend to use llama.cpp for X86 (Intel MKL building).
|
|||
|OS|Status|Verified|
|
||||
|-|-|-|
|
||||
|Linux|Support|Ubuntu 22.04|
|
||||
|Windows|Ongoing| |
|
||||
|Windows|Support|Windows 11|
|
||||
|
||||
|
||||
## Intel GPU
|
||||
|
|
@ -42,7 +46,7 @@ For Intel CPU, recommend to use llama.cpp for X86 (Intel MKL building).
|
|||
|-|-|-|
|
||||
|Intel Data Center Max Series| Support| Max 1550|
|
||||
|Intel Data Center Flex Series| Support| Flex 170|
|
||||
|Intel Arc Series| Support| Arc 770|
|
||||
|Intel Arc Series| Support| Arc 770, 730M|
|
||||
|Intel built-in Arc GPU| Support| built-in Arc GPU in Meteor Lake|
|
||||
|Intel iGPU| Support| iGPU in i5-1250P, i7-1165G7|
|
||||
|
||||
|
|
@ -131,6 +135,7 @@ cmake .. -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx
|
|||
#build all binary
|
||||
cmake --build . --config Release -v
|
||||
|
||||
cd ..
|
||||
```
|
||||
|
||||
or
|
||||
|
|
@ -195,7 +200,7 @@ GGML_SYCL_DEVICE=0 ./build/bin/main -m models/llama-2-7b.Q4_0.gguf -p "Building
|
|||
or run by script:
|
||||
|
||||
```
|
||||
./examples/sycl/run_llama2.sh
|
||||
./examples/sycl/run-llama2.sh
|
||||
```
|
||||
|
||||
Note:
|
||||
|
|
@ -205,11 +210,175 @@ Note:
|
|||
|
||||
5. Check the device ID in output
|
||||
|
||||
Like：
|
||||
Like:
|
||||
```
|
||||
Using device **0** (Intel(R) Arc(TM) A770 Graphics) as main device
|
||||
```
|
||||
|
||||
## Windows
|
||||
|
||||
### Setup Environment
|
||||
|
||||
1. Install Intel GPU driver.
|
||||
|
||||
Please install Intel GPU driver by official guide: [Install GPU Drivers](https://www.intel.com/content/www/us/en/products/docs/discrete-gpus/arc/software/drivers.html).
|
||||
|
||||
2. Install Intel® oneAPI Base toolkit.
|
||||
|
||||
a. Please follow the procedure in [Get the Intel® oneAPI Base Toolkit ](https://www.intel.com/content/www/us/en/developer/tools/oneapi/base-toolkit.html).
|
||||
|
||||
Recommend to install to default folder: **/opt/intel/oneapi**.
|
||||
|
||||
Following guide uses the default folder as example. If you use other folder, please modify the following guide info with your folder.
|
||||
|
||||
b. Enable oneAPI running environment:
|
||||
|
||||
- In Search, input 'oneAPI'.
|
||||
|
||||
Search & open "Intel oneAPI command prompt for Intel 64 for Visual Studio 2022"
|
||||
|
||||
- In Run:
|
||||
|
||||
In CMD:
|
||||
```
|
||||
"C:\Program Files (x86)\Intel\oneAPI\setvars.bat" intel64
|
||||
```
|
||||
|
||||
c. Check GPU
|
||||
|
||||
In oneAPI command line:
|
||||
|
||||
```
|
||||
sycl-ls
|
||||
```
|
||||
|
||||
There should be one or more level-zero devices. Like **[ext_oneapi_level_zero:gpu:0]**.
|
||||
|
||||
Output (example):
|
||||
```
|
||||
[opencl:acc:0] Intel(R) FPGA Emulation Platform for OpenCL(TM), Intel(R) FPGA Emulation Device OpenCL 1.2 [2023.16.10.0.17_160000]
|
||||
[opencl:cpu:1] Intel(R) OpenCL, 11th Gen Intel(R) Core(TM) i7-1185G7 @ 3.00GHz OpenCL 3.0 (Build 0) [2023.16.10.0.17_160000]
|
||||
[opencl:gpu:2] Intel(R) OpenCL Graphics, Intel(R) Iris(R) Xe Graphics OpenCL 3.0 NEO [31.0.101.5186]
|
||||
[ext_oneapi_level_zero:gpu:0] Intel(R) Level-Zero, Intel(R) Iris(R) Xe Graphics 1.3 [1.3.28044]
|
||||
|
||||
```
|
||||
|
||||
3. Install cmake & make
|
||||
|
||||
a. Download & install cmake for windows: https://cmake.org/download/
|
||||
|
||||
b. Download & install make for windows provided by mingw-w64: https://www.mingw-w64.org/downloads/
|
||||
|
||||
|
||||
### Build locally:
|
||||
|
||||
In oneAPI command line window:
|
||||
|
||||
```
|
||||
mkdir -p build
|
||||
cd build
|
||||
@call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat" intel64 --force
|
||||
|
||||
:: for FP16
|
||||
:: faster for long-prompt inference
|
||||
:: cmake -G "MinGW Makefiles" .. -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icx -DCMAKE_BUILD_TYPE=Release -DLLAMA_SYCL_F16=ON
|
||||
|
||||
:: for FP32
|
||||
cmake -G "MinGW Makefiles" .. -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icx -DCMAKE_BUILD_TYPE=Release
|
||||
|
||||
|
||||
:: build example/main only
|
||||
:: make main
|
||||
|
||||
:: build all binary
|
||||
make -j
|
||||
cd ..
|
||||
```
|
||||
|
||||
or
|
||||
|
||||
```
|
||||
.\examples\sycl\win-build-sycl.bat
|
||||
```
|
||||
|
||||
Note:
|
||||
|
||||
- By default, it will build for all binary files. It will take more time. To reduce the time, we recommend to build for **example/main** only.
|
||||
|
||||
### Run
|
||||
|
||||
1. Put model file to folder **models**
|
||||
|
||||
2. Enable oneAPI running environment
|
||||
|
||||
- In Search, input 'oneAPI'.
|
||||
|
||||
Search & open "Intel oneAPI command prompt for Intel 64 for Visual Studio 2022"
|
||||
|
||||
- In Run:
|
||||
|
||||
In CMD:
|
||||
```
|
||||
"C:\Program Files (x86)\Intel\oneAPI\setvars.bat" intel64
|
||||
```
|
||||
|
||||
3. List device ID
|
||||
|
||||
Run without parameter:
|
||||
|
||||
```
|
||||
build\bin\ls-sycl-device.exe
|
||||
|
||||
or
|
||||
|
||||
build\bin\main.exe
|
||||
```
|
||||
|
||||
Check the ID in startup log, like:
|
||||
|
||||
```
|
||||
found 4 SYCL devices:
|
||||
Device 0: Intel(R) Arc(TM) A770 Graphics, compute capability 1.3,
|
||||
max compute_units 512, max work group size 1024, max sub group size 32, global mem size 16225243136
|
||||
Device 1: Intel(R) FPGA Emulation Device, compute capability 1.2,
|
||||
max compute_units 24, max work group size 67108864, max sub group size 64, global mem size 67065057280
|
||||
Device 2: 13th Gen Intel(R) Core(TM) i7-13700K, compute capability 3.0,
|
||||
max compute_units 24, max work group size 8192, max sub group size 64, global mem size 67065057280
|
||||
Device 3: Intel(R) Arc(TM) A770 Graphics, compute capability 3.0,
|
||||
max compute_units 512, max work group size 1024, max sub group size 32, global mem size 16225243136
|
||||
|
||||
```
|
||||
|
||||
|Attribute|Note|
|
||||
|-|-|
|
||||
|compute capability 1.3|Level-zero running time, recommended |
|
||||
|compute capability 3.0|OpenCL running time, slower than level-zero in most cases|
|
||||
|
||||
4. Set device ID and execute llama.cpp
|
||||
|
||||
Set device ID = 0 by **set GGML_SYCL_DEVICE=0**
|
||||
|
||||
```
|
||||
set GGML_SYCL_DEVICE=0
|
||||
build\bin\main.exe -m models\llama-2-7b.Q4_0.gguf -p "Building a website can be done in 10 simple steps:\nStep 1:" -n 400 -e -ngl 33 -s 0
|
||||
```
|
||||
or run by script:
|
||||
|
||||
```
|
||||
.\examples\sycl\win-run-llama2.bat
|
||||
```
|
||||
|
||||
Note:
|
||||
|
||||
- By default, mmap is used to read model file. In some cases, it leads to the hang issue. Recommend to use parameter **--no-mmap** to disable mmap() to skip this issue.
|
||||
|
||||
|
||||
5. Check the device ID in output
|
||||
|
||||
Like:
|
||||
```
|
||||
Using device **0** (Intel(R) Arc(TM) A770 Graphics) as main device
|
||||
```
|
||||
|
||||
## Environment Variable
|
||||
|
||||
|
|
@ -220,7 +389,7 @@ Using device **0** (Intel(R) Arc(TM) A770 Graphics) as main device
|
|||
|LLAMA_SYCL|ON (mandatory)|Enable build with SYCL code path. <br>For FP32/FP16, LLAMA_SYCL=ON is mandatory.|
|
||||
|LLAMA_SYCL_F16|ON (optional)|Enable FP16 build with SYCL code path. Faster for long-prompt inference. <br>For FP32, not set it.|
|
||||
|CMAKE_C_COMPILER|icx|Use icx compiler for SYCL code path|
|
||||
|CMAKE_CXX_COMPILER|icpx|use icpx for SYCL code path|
|
||||
|CMAKE_CXX_COMPILER|icpx (Linux), icx (Windows)|use icpx/icx for SYCL code path|
|
||||
|
||||
#### Running
|
||||
|
||||
|
|
@ -232,19 +401,24 @@ Using device **0** (Intel(R) Arc(TM) A770 Graphics) as main device
|
|||
|
||||
## Known Issue
|
||||
|
||||
- Error: `error while loading shared libraries: libsycl.so.7: cannot open shared object file: No such file or directory`.
|
||||
|
||||
Miss to enable oneAPI running environment.
|
||||
|
||||
Install oneAPI base toolkit and enable it by: `source /opt/intel/oneapi/setvars.sh`.
|
||||
|
||||
|
||||
- Hang during startup
|
||||
|
||||
llama.cpp use mmap as default way to read model file and copy to GPU. In some system, memcpy will be abnormal and block.
|
||||
|
||||
Solution: add **--no-mmap**.
|
||||
|
||||
## Q&A
|
||||
|
||||
- Error: `error while loading shared libraries: libsycl.so.7: cannot open shared object file: No such file or directory`.
|
||||
|
||||
Miss to enable oneAPI running environment.
|
||||
|
||||
Install oneAPI base toolkit and enable it by: `source /opt/intel/oneapi/setvars.sh`.
|
||||
|
||||
- In Windows, no result, not error.
|
||||
|
||||
Miss to enable oneAPI running environment.
|
||||
|
||||
## Todo
|
||||
|
||||
- Support to build in Windows.
|
||||
|
|
@ -1521,7 +1521,9 @@ void dump_non_result_info_yaml(FILE * stream, const gpt_params & params, const l
|
|||
fprintf(stream, "cpu_has_avx512_vbmi: %s\n", ggml_cpu_has_avx512_vbmi() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_avx512_vnni: %s\n", ggml_cpu_has_avx512_vnni() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_cublas: %s\n", ggml_cpu_has_cublas() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_vulkan: %s\n", ggml_cpu_has_vulkan() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_clblast: %s\n", ggml_cpu_has_clblast() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_kompute: %s\n", ggml_cpu_has_kompute() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_fma: %s\n", ggml_cpu_has_fma() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_gpublas: %s\n", ggml_cpu_has_gpublas() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_neon: %s\n", ggml_cpu_has_neon() ? "true" : "false");
|
||||
|
|
|
|||
|
|
@ -564,6 +564,7 @@ struct test {
|
|||
static const bool cuda;
|
||||
static const bool opencl;
|
||||
static const bool vulkan;
|
||||
static const bool kompute;
|
||||
static const bool metal;
|
||||
static const bool gpu_blas;
|
||||
static const bool blas;
|
||||
|
|
@ -648,6 +649,9 @@ struct test {
|
|||
if (vulkan) {
|
||||
return "Vulkan";
|
||||
}
|
||||
if (kompute) {
|
||||
return "Kompute";
|
||||
}
|
||||
if (metal) {
|
||||
return "Metal";
|
||||
}
|
||||
|
|
@ -663,7 +667,7 @@ struct test {
|
|||
static const std::vector<std::string> & get_fields() {
|
||||
static const std::vector<std::string> fields = {
|
||||
"build_commit", "build_number",
|
||||
"cuda", "opencl", "vulkan", "metal", "gpu_blas", "blas",
|
||||
"cuda", "opencl", "vulkan", "kompute", "metal", "gpu_blas", "blas",
|
||||
"cpu_info", "gpu_info",
|
||||
"model_filename", "model_type", "model_size", "model_n_params",
|
||||
"n_batch", "n_threads", "type_k", "type_v",
|
||||
|
|
@ -687,8 +691,9 @@ struct test {
|
|||
field == "avg_ns" || field == "stddev_ns") {
|
||||
return INT;
|
||||
}
|
||||
if (field == "cuda" || field == "opencl" || field == "vulkan"|| field == "metal" || field == "gpu_blas" || field == "blas" ||
|
||||
field == "f16_kv" || field == "no_kv_offload" || field == "mul_mat_q") {
|
||||
if (field == "cuda" || field == "opencl" || field == "vulkan" || field == "kompute" || field == "metal" ||
|
||||
field == "gpu_blas" || field == "blas" || field == "f16_kv" || field == "no_kv_offload" ||
|
||||
field == "mul_mat_q") {
|
||||
return BOOL;
|
||||
}
|
||||
if (field == "avg_ts" || field == "stddev_ts") {
|
||||
|
|
@ -715,7 +720,8 @@ struct test {
|
|||
}
|
||||
std::vector<std::string> values = {
|
||||
build_commit, std::to_string(build_number),
|
||||
std::to_string(cuda), std::to_string(opencl), std::to_string(vulkan), std::to_string(metal), std::to_string(gpu_blas), std::to_string(blas),
|
||||
std::to_string(cuda), std::to_string(opencl), std::to_string(vulkan), std::to_string(vulkan),
|
||||
std::to_string(metal), std::to_string(gpu_blas), std::to_string(blas),
|
||||
cpu_info, gpu_info,
|
||||
model_filename, model_type, std::to_string(model_size), std::to_string(model_n_params),
|
||||
std::to_string(n_batch), std::to_string(n_threads), ggml_type_name(type_k), ggml_type_name(type_v),
|
||||
|
|
@ -744,6 +750,7 @@ const int test::build_number = LLAMA_BUILD_NUMBER;
|
|||
const bool test::cuda = !!ggml_cpu_has_cublas();
|
||||
const bool test::opencl = !!ggml_cpu_has_clblast();
|
||||
const bool test::vulkan = !!ggml_cpu_has_vulkan();
|
||||
const bool test::kompute = !!ggml_cpu_has_kompute();
|
||||
const bool test::metal = !!ggml_cpu_has_metal();
|
||||
const bool test::gpu_blas = !!ggml_cpu_has_gpublas();
|
||||
const bool test::blas = !!ggml_cpu_has_blas();
|
||||
|
|
|
|||
|
|
@ -48,6 +48,7 @@ chat_completion() {
|
|||
top_p: 0.9,
|
||||
n_keep: $n_keep,
|
||||
n_predict: 256,
|
||||
cache_prompt: true,
|
||||
stop: ["\n### Human:"],
|
||||
stream: true
|
||||
}')"
|
||||
|
|
|
|||
|
|
@ -221,6 +221,7 @@ struct llama_client_slot
|
|||
infill = false;
|
||||
ga_i = 0;
|
||||
n_past_se = 0;
|
||||
|
||||
generated_token_probs.clear();
|
||||
|
||||
for (slot_image & img : images)
|
||||
|
|
@ -1228,7 +1229,7 @@ struct llama_server_context
|
|||
std::vector<llama_token> append_tokens = tokenize(json_prompt, false); // has next image
|
||||
for (int i = 0; i < (int) append_tokens.size(); ++i)
|
||||
{
|
||||
llama_batch_add(batch, append_tokens[i], slot.n_past, { slot.id }, true);
|
||||
llama_batch_add(batch, append_tokens[i], system_tokens.size() + slot.n_past, { slot.id }, true);
|
||||
slot.n_past += 1;
|
||||
}
|
||||
}
|
||||
|
|
@ -1296,6 +1297,8 @@ struct llama_server_context
|
|||
for (llama_client_slot &slot : slots)
|
||||
{
|
||||
slot.cache_tokens.clear();
|
||||
slot.n_past = 0;
|
||||
slot.n_past_se = 0;
|
||||
}
|
||||
}
|
||||
|
||||
|
|
@ -1365,26 +1368,26 @@ struct llama_server_context
|
|||
kv_cache_clear();
|
||||
}
|
||||
return true;
|
||||
} else {
|
||||
}
|
||||
|
||||
task_server task;
|
||||
task.type = TASK_TYPE_NEXT_RESPONSE;
|
||||
task.target_id = -1;
|
||||
queue_tasks.post(task);
|
||||
}
|
||||
|
||||
for (llama_client_slot &slot : slots)
|
||||
{
|
||||
if (slot.ga_n == 1)
|
||||
{
|
||||
if (slot.is_processing() && slot.cache_tokens.size() >= (size_t) slot.n_ctx)
|
||||
if (slot.is_processing() && system_tokens.size() + slot.cache_tokens.size() >= (size_t) slot.n_ctx)
|
||||
{
|
||||
// Shift context
|
||||
const int n_left = slot.n_past - slot.params.n_keep - 1;
|
||||
const int n_left = system_tokens.size() + slot.n_past - slot.params.n_keep - 1;
|
||||
const int n_discard = n_left / 2;
|
||||
|
||||
LOG_TEE("slot %d: context shift - n_keep = %d, n_left = %d, n_discard = %d\n", slot.id, slot.params.n_keep, n_left, n_discard);
|
||||
llama_kv_cache_seq_rm (ctx, slot.id, slot.params.n_keep + 1 , slot.params.n_keep + n_discard + 1);
|
||||
llama_kv_cache_seq_shift(ctx, slot.id, slot.params.n_keep + 1 + n_discard, slot.n_past, -n_discard);
|
||||
llama_kv_cache_seq_shift(ctx, slot.id, slot.params.n_keep + 1 + n_discard, system_tokens.size() + slot.n_past, -n_discard);
|
||||
|
||||
for (size_t i = slot.params.n_keep + 1 + n_discard; i < slot.cache_tokens.size(); i++)
|
||||
{
|
||||
|
|
@ -1430,8 +1433,10 @@ struct llama_server_context
|
|||
slot.i_batch = batch.n_tokens;
|
||||
|
||||
const int32_t slot_npast = slot.n_past_se > 0 ? slot.n_past_se : slot.n_past;
|
||||
llama_batch_add(batch, slot.sampled, system_tokens.size() + slot_npast, { slot.id }, true);
|
||||
|
||||
// TODO: we always have to take into account the "system_tokens"
|
||||
// this is not great and needs to be improved somehow
|
||||
llama_batch_add(batch, slot.sampled, system_tokens.size() + slot_npast, { slot.id }, true);
|
||||
slot.n_past += 1;
|
||||
}
|
||||
|
||||
|
|
@ -1592,10 +1597,13 @@ struct llama_server_context
|
|||
|
||||
// process the prefix of first image
|
||||
std::vector<llama_token> prefix_tokens = has_images ? tokenize(slot.images[0].prefix_prompt, add_bos_token) : prompt_tokens;
|
||||
|
||||
int32_t slot_npast = slot.n_past_se > 0 ? slot.n_past_se : slot.n_past;
|
||||
int ga_i = slot.ga_i;
|
||||
|
||||
int32_t ga_i = slot.ga_i;
|
||||
int32_t ga_n = slot.ga_n;
|
||||
int32_t ga_w = slot.ga_w;
|
||||
|
||||
for (; slot.n_past < (int) prefix_tokens.size(); ++slot.n_past)
|
||||
{
|
||||
if (slot.ga_n != 1)
|
||||
|
|
@ -1607,7 +1615,7 @@ struct llama_server_context
|
|||
}
|
||||
}
|
||||
llama_batch_add(batch, prefix_tokens[slot.n_past], system_tokens.size() + slot_npast, {slot.id }, false);
|
||||
slot_npast += 1;
|
||||
slot_npast++;
|
||||
}
|
||||
|
||||
if (has_images && !ingest_images(slot, n_batch))
|
||||
|
|
@ -1667,6 +1675,7 @@ struct llama_server_context
|
|||
slot.n_past_se += n_tokens;
|
||||
}
|
||||
}
|
||||
|
||||
llama_batch batch_view =
|
||||
{
|
||||
n_tokens,
|
||||
|
|
@ -1819,15 +1828,15 @@ static void server_print_usage(const char *argv0, const gpt_params ¶ms,
|
|||
printf(" -np N, --parallel N number of slots for process requests (default: %d)\n", params.n_parallel);
|
||||
printf(" -cb, --cont-batching enable continuous batching (a.k.a dynamic batching) (default: disabled)\n");
|
||||
printf(" -spf FNAME, --system-prompt-file FNAME\n");
|
||||
printf(" Set a file to load a system prompt (initial prompt of all slots), this is useful for chat applications.\n");
|
||||
printf(" set a file to load a system prompt (initial prompt of all slots), this is useful for chat applications.\n");
|
||||
printf(" --mmproj MMPROJ_FILE path to a multimodal projector file for LLaVA.\n");
|
||||
printf(" --log-disable disables logging to a file.\n");
|
||||
printf("\n");
|
||||
printf(" --override-kv KEY=TYPE:VALUE\n");
|
||||
printf(" advanced option to override model metadata by key. may be specified multiple times.\n");
|
||||
printf(" types: int, float, bool. example: --override-kv tokenizer.ggml.add_bos_token=bool:false\n");
|
||||
printf(" -gan N, --grp-attn-n N Set the group attention factor to extend context size through self-extend(default: 1=disabled), used together with group attention width `--grp-attn-w`");
|
||||
printf(" -gaw N, --grp-attn-w N Set the group attention width to extend context size through self-extend(default: 512), used together with group attention factor `--grp-attn-n`");
|
||||
printf(" -gan N, --grp-attn-n N set the group attention factor to extend context size through self-extend(default: 1=disabled), used together with group attention width `--grp-attn-w`");
|
||||
printf(" -gaw N, --grp-attn-w N set the group attention width to extend context size through self-extend(default: 512), used together with group attention factor `--grp-attn-n`");
|
||||
printf("\n");
|
||||
}
|
||||
|
||||
|
|
|
|||
23
examples/sycl/win-build-sycl.bat
Normal file
23
examples/sycl/win-build-sycl.bat
Normal file
|
|
@ -0,0 +1,23 @@
|
|||
|
||||
:: MIT license
|
||||
:: Copyright (C) 2024 Intel Corporation
|
||||
:: SPDX-License-Identifier: MIT
|
||||
|
||||
mkdir -p build
|
||||
cd build
|
||||
@call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat" intel64 --force
|
||||
|
||||
:: for FP16
|
||||
:: faster for long-prompt inference
|
||||
:: cmake -G "MinGW Makefiles" .. -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icx -DCMAKE_BUILD_TYPE=Release -DLLAMA_SYCL_F16=ON
|
||||
|
||||
:: for FP32
|
||||
cmake -G "MinGW Makefiles" .. -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icx -DCMAKE_BUILD_TYPE=Release
|
||||
|
||||
|
||||
:: build example/main only
|
||||
:: make main
|
||||
|
||||
:: build all binary
|
||||
make -j
|
||||
cd ..
|
||||
13
examples/sycl/win-run-llama2.bat
Normal file
13
examples/sycl/win-run-llama2.bat
Normal file
|
|
@ -0,0 +1,13 @@
|
|||
:: MIT license
|
||||
:: Copyright (C) 2024 Intel Corporation
|
||||
:: SPDX-License-Identifier: MIT
|
||||
|
||||
INPUT2="Building a website can be done in 10 simple steps:\nStep 1:"
|
||||
@call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat" intel64 --force
|
||||
|
||||
|
||||
set GGML_SYCL_DEVICE=0
|
||||
rem set GGML_SYCL_DEBUG=1
|
||||
.\build\bin\main.exe -m models\llama-2-7b.Q4_0.gguf -p %INPUT2% -n 400 -e -ngl 33 -s 0
|
||||
|
||||
|
||||
|
|
@ -3688,38 +3688,38 @@ constexpr constant static uint64_t iq2xs_grid[512] = {
|
|||
};
|
||||
|
||||
constexpr constant static uint32_t iq3xxs_grid[256] = {
|
||||
0x04040404, 0x04040414, 0x04040424, 0x04040c0c, 0x04040c1c, 0x04040c3c, 0x04041404, 0x04041414,
|
||||
0x04041c0c, 0x04042414, 0x04043c1c, 0x04043c2c, 0x040c040c, 0x040c041c, 0x040c0c04, 0x040c0c14,
|
||||
0x040c140c, 0x040c142c, 0x040c1c04, 0x040c1c14, 0x040c240c, 0x040c2c24, 0x040c3c04, 0x04140404,
|
||||
0x04140414, 0x04140424, 0x04140c0c, 0x04141404, 0x04141414, 0x04141c0c, 0x04141c1c, 0x04141c3c,
|
||||
0x04142c0c, 0x04142c3c, 0x04143c2c, 0x041c040c, 0x041c043c, 0x041c0c04, 0x041c0c14, 0x041c142c,
|
||||
0x041c3c04, 0x04240c1c, 0x04241c3c, 0x04242424, 0x04242c3c, 0x04243c1c, 0x04243c2c, 0x042c040c,
|
||||
0x042c043c, 0x042c1c14, 0x042c2c14, 0x04341c2c, 0x04343424, 0x043c0c04, 0x043c0c24, 0x043c0c34,
|
||||
0x043c241c, 0x043c340c, 0x0c04040c, 0x0c04041c, 0x0c040c04, 0x0c040c14, 0x0c04140c, 0x0c04141c,
|
||||
0x0c041c04, 0x0c041c14, 0x0c041c24, 0x0c04243c, 0x0c042c04, 0x0c0c0404, 0x0c0c0414, 0x0c0c0c0c,
|
||||
0x04040404, 0x04040414, 0x04040424, 0x04040c0c, 0x04040c1c, 0x04040c3e, 0x04041404, 0x04041414,
|
||||
0x04041c0c, 0x04042414, 0x04043e1c, 0x04043e2c, 0x040c040c, 0x040c041c, 0x040c0c04, 0x040c0c14,
|
||||
0x040c140c, 0x040c142c, 0x040c1c04, 0x040c1c14, 0x040c240c, 0x040c2c24, 0x040c3e04, 0x04140404,
|
||||
0x04140414, 0x04140424, 0x04140c0c, 0x04141404, 0x04141414, 0x04141c0c, 0x04141c1c, 0x04141c3e,
|
||||
0x04142c0c, 0x04142c3e, 0x04143e2c, 0x041c040c, 0x041c043e, 0x041c0c04, 0x041c0c14, 0x041c142c,
|
||||
0x041c3e04, 0x04240c1c, 0x04241c3e, 0x04242424, 0x04242c3e, 0x04243e1c, 0x04243e2c, 0x042c040c,
|
||||
0x042c043e, 0x042c1c14, 0x042c2c14, 0x04341c2c, 0x04343424, 0x043e0c04, 0x043e0c24, 0x043e0c34,
|
||||
0x043e241c, 0x043e340c, 0x0c04040c, 0x0c04041c, 0x0c040c04, 0x0c040c14, 0x0c04140c, 0x0c04141c,
|
||||
0x0c041c04, 0x0c041c14, 0x0c041c24, 0x0c04243e, 0x0c042c04, 0x0c0c0404, 0x0c0c0414, 0x0c0c0c0c,
|
||||
0x0c0c1404, 0x0c0c1414, 0x0c14040c, 0x0c14041c, 0x0c140c04, 0x0c140c14, 0x0c14140c, 0x0c141c04,
|
||||
0x0c143c14, 0x0c1c0404, 0x0c1c0414, 0x0c1c1404, 0x0c1c1c0c, 0x0c1c2434, 0x0c1c3434, 0x0c24040c,
|
||||
0x0c24042c, 0x0c242c04, 0x0c2c1404, 0x0c2c1424, 0x0c2c2434, 0x0c2c3c0c, 0x0c34042c, 0x0c3c1414,
|
||||
0x0c3c2404, 0x14040404, 0x14040414, 0x14040c0c, 0x14040c1c, 0x14041404, 0x14041414, 0x14041434,
|
||||
0x0c143e14, 0x0c1c0404, 0x0c1c0414, 0x0c1c1404, 0x0c1c1c0c, 0x0c1c2434, 0x0c1c3434, 0x0c24040c,
|
||||
0x0c24042c, 0x0c242c04, 0x0c2c1404, 0x0c2c1424, 0x0c2c2434, 0x0c2c3e0c, 0x0c34042c, 0x0c3e1414,
|
||||
0x0c3e2404, 0x14040404, 0x14040414, 0x14040c0c, 0x14040c1c, 0x14041404, 0x14041414, 0x14041434,
|
||||
0x14041c0c, 0x14042414, 0x140c040c, 0x140c041c, 0x140c042c, 0x140c0c04, 0x140c0c14, 0x140c140c,
|
||||
0x140c1c04, 0x140c341c, 0x140c343c, 0x140c3c04, 0x14140404, 0x14140414, 0x14140c0c, 0x14140c3c,
|
||||
0x14141404, 0x14141414, 0x14141c3c, 0x14142404, 0x14142c2c, 0x141c040c, 0x141c0c04, 0x141c0c24,
|
||||
0x141c3c04, 0x141c3c24, 0x14241c2c, 0x14242c1c, 0x142c041c, 0x142c143c, 0x142c240c, 0x142c3c24,
|
||||
0x143c040c, 0x143c041c, 0x143c0c34, 0x143c242c, 0x1c04040c, 0x1c040c04, 0x1c040c14, 0x1c04140c,
|
||||
0x1c04141c, 0x1c042c04, 0x1c04342c, 0x1c043c14, 0x1c0c0404, 0x1c0c0414, 0x1c0c1404, 0x1c0c1c0c,
|
||||
0x1c0c2424, 0x1c0c2434, 0x1c14040c, 0x1c14041c, 0x1c140c04, 0x1c14142c, 0x1c142c14, 0x1c143c14,
|
||||
0x1c1c0c0c, 0x1c1c1c1c, 0x1c241c04, 0x1c24243c, 0x1c243c14, 0x1c2c0404, 0x1c2c0434, 0x1c2c1414,
|
||||
0x1c2c2c2c, 0x1c340c24, 0x1c341c34, 0x1c34341c, 0x1c3c1c1c, 0x1c3c3404, 0x24040424, 0x24040c3c,
|
||||
0x24041c2c, 0x24041c3c, 0x24042c1c, 0x24042c3c, 0x240c3c24, 0x24141404, 0x24141c3c, 0x24142404,
|
||||
0x24143404, 0x24143434, 0x241c043c, 0x241c242c, 0x24240424, 0x24242c0c, 0x24243424, 0x242c142c,
|
||||
0x242c241c, 0x242c3c04, 0x243c042c, 0x243c0c04, 0x243c0c14, 0x243c1c04, 0x2c040c14, 0x2c04240c,
|
||||
0x2c043c04, 0x2c0c0404, 0x2c0c0434, 0x2c0c1434, 0x2c0c2c2c, 0x2c140c24, 0x2c141c14, 0x2c143c14,
|
||||
0x2c1c0414, 0x2c1c2c1c, 0x2c240c04, 0x2c24141c, 0x2c24143c, 0x2c243c14, 0x2c2c0414, 0x2c2c1c0c,
|
||||
0x2c342c04, 0x2c3c1424, 0x2c3c2414, 0x34041424, 0x34042424, 0x34042434, 0x34043424, 0x340c140c,
|
||||
0x340c340c, 0x34140c3c, 0x34143424, 0x341c1c04, 0x341c1c34, 0x34242424, 0x342c042c, 0x342c2c14,
|
||||
0x34341c1c, 0x343c041c, 0x343c140c, 0x3c04041c, 0x3c04042c, 0x3c04043c, 0x3c040c04, 0x3c041c14,
|
||||
0x3c042c14, 0x3c0c1434, 0x3c0c2404, 0x3c140c14, 0x3c14242c, 0x3c142c14, 0x3c1c0404, 0x3c1c0c2c,
|
||||
0x3c1c1c1c, 0x3c1c3404, 0x3c24140c, 0x3c24240c, 0x3c2c0404, 0x3c2c0414, 0x3c2c1424, 0x3c341c04,
|
||||
0x140c1c04, 0x140c341c, 0x140c343e, 0x140c3e04, 0x14140404, 0x14140414, 0x14140c0c, 0x14140c3e,
|
||||
0x14141404, 0x14141414, 0x14141c3e, 0x14142404, 0x14142c2c, 0x141c040c, 0x141c0c04, 0x141c0c24,
|
||||
0x141c3e04, 0x141c3e24, 0x14241c2c, 0x14242c1c, 0x142c041c, 0x142c143e, 0x142c240c, 0x142c3e24,
|
||||
0x143e040c, 0x143e041c, 0x143e0c34, 0x143e242c, 0x1c04040c, 0x1c040c04, 0x1c040c14, 0x1c04140c,
|
||||
0x1c04141c, 0x1c042c04, 0x1c04342c, 0x1c043e14, 0x1c0c0404, 0x1c0c0414, 0x1c0c1404, 0x1c0c1c0c,
|
||||
0x1c0c2424, 0x1c0c2434, 0x1c14040c, 0x1c14041c, 0x1c140c04, 0x1c14142c, 0x1c142c14, 0x1c143e14,
|
||||
0x1c1c0c0c, 0x1c1c1c1c, 0x1c241c04, 0x1c24243e, 0x1c243e14, 0x1c2c0404, 0x1c2c0434, 0x1c2c1414,
|
||||
0x1c2c2c2c, 0x1c340c24, 0x1c341c34, 0x1c34341c, 0x1c3e1c1c, 0x1c3e3404, 0x24040424, 0x24040c3e,
|
||||
0x24041c2c, 0x24041c3e, 0x24042c1c, 0x24042c3e, 0x240c3e24, 0x24141404, 0x24141c3e, 0x24142404,
|
||||
0x24143404, 0x24143434, 0x241c043e, 0x241c242c, 0x24240424, 0x24242c0c, 0x24243424, 0x242c142c,
|
||||
0x242c241c, 0x242c3e04, 0x243e042c, 0x243e0c04, 0x243e0c14, 0x243e1c04, 0x2c040c14, 0x2c04240c,
|
||||
0x2c043e04, 0x2c0c0404, 0x2c0c0434, 0x2c0c1434, 0x2c0c2c2c, 0x2c140c24, 0x2c141c14, 0x2c143e14,
|
||||
0x2c1c0414, 0x2c1c2c1c, 0x2c240c04, 0x2c24141c, 0x2c24143e, 0x2c243e14, 0x2c2c0414, 0x2c2c1c0c,
|
||||
0x2c342c04, 0x2c3e1424, 0x2c3e2414, 0x34041424, 0x34042424, 0x34042434, 0x34043424, 0x340c140c,
|
||||
0x340c340c, 0x34140c3e, 0x34143424, 0x341c1c04, 0x341c1c34, 0x34242424, 0x342c042c, 0x342c2c14,
|
||||
0x34341c1c, 0x343e041c, 0x343e140c, 0x3e04041c, 0x3e04042c, 0x3e04043e, 0x3e040c04, 0x3e041c14,
|
||||
0x3e042c14, 0x3e0c1434, 0x3e0c2404, 0x3e140c14, 0x3e14242c, 0x3e142c14, 0x3e1c0404, 0x3e1c0c2c,
|
||||
0x3e1c1c1c, 0x3e1c3404, 0x3e24140c, 0x3e24240c, 0x3e2c0404, 0x3e2c0414, 0x3e2c1424, 0x3e341c04,
|
||||
};
|
||||
|
||||
|
||||
|
|
|
|||
File diff suppressed because it is too large
Load diff
|
|
@ -817,7 +817,7 @@ static void ggml_vk_load_shaders() {
|
|||
// mulmat
|
||||
std::initializer_list<uint32_t> warptile_l = { 128, 128, 128, 16, vk_device.subgroup_size * 2, 64, 2, 4, 4, vk_device.subgroup_size };
|
||||
std::initializer_list<uint32_t> warptile_m = { 128, 64, 64, 16, vk_device.subgroup_size, 32, 2, 4, 2, vk_device.subgroup_size };
|
||||
std::initializer_list<uint32_t> warptile_s = { vk_device.subgroup_size, 32, 32, 8, 32, 32, 2, 2, 2, vk_device.subgroup_size };
|
||||
std::initializer_list<uint32_t> warptile_s = { vk_device.subgroup_size, 32, 32, 16, 32, 32, 2, 2, 2, vk_device.subgroup_size };
|
||||
|
||||
std::array<uint32_t, 3> l_wg_denoms = {128, 128, 1 };
|
||||
std::array<uint32_t, 3> m_wg_denoms = { 64, 64, 1 };
|
||||
|
|
@ -2873,7 +2873,8 @@ static void ggml_vk_op_f32(vk_context * ctx, const ggml_tensor * src0, const ggm
|
|||
if (op == GGML_OP_CPY) {
|
||||
GGML_ASSERT(!transfer_src0);
|
||||
GGML_ASSERT(!transfer_src1);
|
||||
d_sz = dst->ne[1] * dst->nb[1];
|
||||
x_sz = ggml_nbytes(src0);
|
||||
d_sz = ggml_nbytes(dst);
|
||||
|
||||
if (extra->offset + d_sz >= d_D->size) {
|
||||
d_sz = VK_WHOLE_SIZE;
|
||||
|
|
@ -4556,8 +4557,15 @@ GGML_CALL static bool ggml_backend_vk_graph_compute(ggml_backend_t backend, ggml
|
|||
}
|
||||
ggml_vk_preallocate_buffers();
|
||||
|
||||
int last_node = cgraph->n_nodes - 1;
|
||||
|
||||
// If the last op in the cgraph isn't backend GPU, the command buffer doesn't get closed properly
|
||||
while (last_node > 0 && cgraph->nodes[last_node]->backend != GGML_BACKEND_GPU) {
|
||||
last_node -= 1;
|
||||
}
|
||||
|
||||
for (int i = 0; i < cgraph->n_nodes; i++) {
|
||||
ggml_vk_build_graph(cgraph->nodes[i], i == cgraph->n_nodes - 1);
|
||||
ggml_vk_build_graph(cgraph->nodes[i], i == last_node);
|
||||
}
|
||||
|
||||
ggml_compute_params params = {};
|
||||
|
|
|
|||
11
ggml.c
11
ggml.c
|
|
@ -20510,6 +20510,14 @@ int ggml_cpu_has_vulkan(void) {
|
|||
#endif
|
||||
}
|
||||
|
||||
int ggml_cpu_has_kompute(void) {
|
||||
#if defined(GGML_USE_KOMPUTE)
|
||||
return 1;
|
||||
#else
|
||||
return 0;
|
||||
#endif
|
||||
}
|
||||
|
||||
int ggml_cpu_has_sycl(void) {
|
||||
#if defined(GGML_USE_SYCL)
|
||||
return 1;
|
||||
|
|
@ -20519,7 +20527,8 @@ int ggml_cpu_has_sycl(void) {
|
|||
}
|
||||
|
||||
int ggml_cpu_has_gpublas(void) {
|
||||
return ggml_cpu_has_cublas() || ggml_cpu_has_clblast() || ggml_cpu_has_vulkan() || ggml_cpu_has_sycl();
|
||||
return ggml_cpu_has_cublas() || ggml_cpu_has_clblast() || ggml_cpu_has_vulkan() || ggml_cpu_has_kompute() ||
|
||||
ggml_cpu_has_sycl();
|
||||
}
|
||||
|
||||
int ggml_cpu_has_sse3(void) {
|
||||
|
|
|
|||
1
ggml.h
1
ggml.h
|
|
@ -2273,6 +2273,7 @@ extern "C" {
|
|||
GGML_API int ggml_cpu_has_cublas (void);
|
||||
GGML_API int ggml_cpu_has_clblast (void);
|
||||
GGML_API int ggml_cpu_has_vulkan (void);
|
||||
GGML_API int ggml_cpu_has_kompute (void);
|
||||
GGML_API int ggml_cpu_has_gpublas (void);
|
||||
GGML_API int ggml_cpu_has_sse3 (void);
|
||||
GGML_API int ggml_cpu_has_ssse3 (void);
|
||||
|
|
|
|||
|
|
@ -19,8 +19,8 @@ shader_int8_ext = """
|
|||
|
||||
# Type-specific defines
|
||||
shader_f16_defines = """
|
||||
#define QUANT_K 32
|
||||
#define QUANT_R 2
|
||||
#define QUANT_K 1
|
||||
#define QUANT_R 1
|
||||
|
||||
#define A_TYPE float16_t
|
||||
"""
|
||||
|
|
|
|||
|
|
@ -2760,10 +2760,10 @@ static std::string llama_model_ftype_name(llama_ftype ftype) {
|
|||
case LLAMA_FTYPE_MOSTLY_Q5_K_S: return "Q5_K - Small";
|
||||
case LLAMA_FTYPE_MOSTLY_Q5_K_M: return "Q5_K - Medium";
|
||||
case LLAMA_FTYPE_MOSTLY_Q6_K: return "Q6_K";
|
||||
case LLAMA_FTYPE_MOSTLY_IQ2_XXS:return "IQ2_XSS - 2.0625 bpw";
|
||||
case LLAMA_FTYPE_MOSTLY_IQ2_XXS:return "IQ2_XXS - 2.0625 bpw";
|
||||
case LLAMA_FTYPE_MOSTLY_IQ2_XS: return "IQ2_XS - 2.3125 bpw";
|
||||
case LLAMA_FTYPE_MOSTLY_Q3_K_XS:return "Q3_K - Extra small";
|
||||
case LLAMA_FTYPE_MOSTLY_IQ3_XXS:return "IQ3_XSS - 3.0625 bpw";
|
||||
case LLAMA_FTYPE_MOSTLY_IQ3_XXS:return "IQ3_XXS - 3.0625 bpw";
|
||||
|
||||
default: return "unknown, may not work";
|
||||
}
|
||||
|
|
@ -6954,11 +6954,6 @@ static int llama_decode_internal(
|
|||
n_threads = std::min(4, n_threads);
|
||||
}
|
||||
|
||||
const bool fully_offloaded = model.n_gpu_layers >= (int) hparams.n_layer + 1;
|
||||
if ((ggml_cpu_has_cublas() || ggml_cpu_has_vulkan()) && fully_offloaded) {
|
||||
n_threads = 1;
|
||||
}
|
||||
|
||||
#ifdef GGML_USE_MPI
|
||||
const int64_t n_layer = hparams.n_layer;
|
||||
ggml_mpi_graph_compute_pre(lctx.ctx_mpi, gf, n_layer);
|
||||
|
|
|
|||
19
scripts/install-oneapi.bat
Normal file
19
scripts/install-oneapi.bat
Normal file
|
|
@ -0,0 +1,19 @@
|
|||
:: MIT license
|
||||
:: Copyright (C) 2024 Intel Corporation
|
||||
:: SPDX-License-Identifier: MIT
|
||||
|
||||
|
||||
set URL=%1
|
||||
set COMPONENTS=%2
|
||||
|
||||
curl.exe --output %TEMP%\webimage.exe --url %URL% --retry 5 --retry-delay 5
|
||||
start /b /wait %TEMP%\webimage.exe -s -x -f webimage_extracted --log extract.log
|
||||
del %TEMP%\webimage.exe
|
||||
if "%COMPONENTS%"=="" (
|
||||
webimage_extracted\bootstrapper.exe -s --action install --eula=accept -p=NEED_VS2017_INTEGRATION=0 -p=NEED_VS2019_INTEGRATION=0 -p=NEED_VS2022_INTEGRATION=0 --log-dir=.
|
||||
) else (
|
||||
webimage_extracted\bootstrapper.exe -s --action install --components=%COMPONENTS% --eula=accept -p=NEED_VS2017_INTEGRATION=0 -p=NEED_VS2019_INTEGRATION=0 -p=NEED_VS2022_INTEGRATION=0 --log-dir=.
|
||||
)
|
||||
set installer_exit_code=%ERRORLEVEL%
|
||||
rd /s/q "webimage_extracted"
|
||||
exit /b %installer_exit_code%
|
||||
Loading…
Add table
Add a link
Reference in a new issue