Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Sign up
Appearance settings

Commit abb115c

Browse files
fix: clarify lora quant support and small fixes (#792)
1 parent c648001 commit abb115c

File tree

5 files changed

+49
-4
lines changed

5 files changed

+49
-4
lines changed

‎.gitignore‎

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -4,10 +4,11 @@ test/
44
.cache/
55
*.swp
66
.vscode/
7+
.idea/
78
*.bat
89
*.bin
910
*.exe
1011
*.gguf
1112
output*.png
1213
models*
13-
*.log
14+
*.log

‎README.md‎

Lines changed: 3 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -137,7 +137,9 @@ This provides BLAS acceleration using the ROCm cores of your AMD GPU. Make sure
137137
Windows User Refer to [docs/hipBLAS_on_Windows.md](docs%2FhipBLAS_on_Windows.md) for a comprehensive guide.
138138
139139
```
140-
cmake .. -G "Ninja" -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DSD_HIPBLAS=ON -DCMAKE_BUILD_TYPE=Release -DAMDGPU_TARGETS=gfx1100 -DCMAKE_BUILD_WITH_INSTALL_RPATH=ON
140+
export GFX_NAME=$(rocminfo | grep -m 1 -E "gfx[^0]{1}" | sed -e 's/ *Name: *//' | awk '{1ドル=1ドル; print}' || echo "rocminfo missing")
141+
echo $GFX_NAME
142+
cmake .. -G "Ninja" -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DSD_HIPBLAS=ON -DCMAKE_BUILD_TYPE=Release -DGPU_TARGETS=$GFX_NAME -DCMAKE_BUILD_WITH_INSTALL_RPATH=ON
141143
cmake --build . --config Release
142144
```
143145

‎docs/lora.md‎

Lines changed: 27 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -10,4 +10,30 @@ Here's a simple example:
1010
./bin/sd -m ../models/v1-5-pruned-emaonly.safetensors -p "a lovely cat<lora:marblesh:1>" --lora-model-dir ../models
1111
```
1212

13-
`../models/marblesh.safetensors` or `../models/marblesh.ckpt` will be applied to the model
13+
`../models/marblesh.safetensors` or `../models/marblesh.ckpt` will be applied to the model
14+
15+
# Support matrix
16+
17+
> i️ CUDA `get_rows` support is defined here:
18+
> [ggml-org/ggml/src/ggml-cuda/getrows.cu#L156](https://github.com/ggml-org/ggml/blob/7dee1d6a1e7611f238d09be96738388da97c88ed/src/ggml-cuda/getrows.cu#L156)
19+
> Currently only the basic types + Q4/Q5/Q8 are implemented. K-quants are **not** supported.
20+
21+
NOTE: The other backends may have different support.
22+
23+
| Quant / Type | CUDA |
24+
|--------------|------|
25+
| F32 | ✔️ |
26+
| F16 | ✔️ |
27+
| BF16 | ✔️ |
28+
| I32 | ✔️ |
29+
| Q4_0 | ✔️ |
30+
| Q4_1 | ✔️ |
31+
| Q5_0 | ✔️ |
32+
| Q5_1 | ✔️ |
33+
| Q8_0 | ✔️ |
34+
| Q2_K ||
35+
| Q3_K ||
36+
| Q4_K ||
37+
| Q5_K ||
38+
| Q6_K ||
39+
| Q8_K ||

‎examples/cli/main.cpp‎

Lines changed: 16 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,7 @@
11
#include <stdio.h>
22
#include <string.h>
33
#include <time.h>
4+
#include <filesystem>
45
#include <functional>
56
#include <iostream>
67
#include <map>
@@ -1283,6 +1284,21 @@ int main(int argc, const char* argv[]) {
12831284
}
12841285
}
12851286

1287+
// create directory if not exists
1288+
{
1289+
namespace fs = std::filesystem;
1290+
const fs::path out_path = params.output_path;
1291+
if (const fs::path out_dir = out_path.parent_path(); !out_dir.empty()) {
1292+
std::error_code ec;
1293+
fs::create_directories(out_dir, ec); // OK if already exists
1294+
if (ec) {
1295+
fprintf(stderr, "failed to create directory '%s': %s\n",
1296+
out_dir.string().c_str(), ec.message().c_str());
1297+
return 1;
1298+
}
1299+
}
1300+
}
1301+
12861302
std::string base_path;
12871303
std::string file_ext;
12881304
std::string file_ext_lower;

‎stable-diffusion.cpp‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -330,7 +330,7 @@ class StableDiffusionGGML {
330330
if (sd_version_is_dit(version)) {
331331
use_t5xxl = true;
332332
}
333-
if (!ggml_backend_is_cpu(backend) && use_t5xxl) {
333+
if (!clip_on_cpu && !ggml_backend_is_cpu(backend) && use_t5xxl) {
334334
LOG_WARN(
335335
"!!!It appears that you are using the T5 model. Some backends may encounter issues with it."
336336
"If you notice that the generated images are completely black,"

0 commit comments

Comments
(0)

AltStyle によって変換されたページ (->オリジナル) /