Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

b3639 #310

Merged
merged 5 commits into from
Aug 28, 2024
Merged

b3639 #310

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
a77feb5
server : add some missing env variables (#9116)
ngxson Aug 27, 2024
78eb487
llama : fix qs.n_attention_wv for DeepSeek-V2 (#9156)
compilade Aug 27, 2024
3246fe8
Fix minicpm example directory (#9111)
xyb Aug 27, 2024
231cff5
sync : ggml
ggerganov Aug 27, 2024
20f1789
vulkan : fix build (#0)
ggerganov Aug 27, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions .devops/llama-server-cuda.Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,6 +24,8 @@ ENV CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH}
ENV GGML_CUDA=1
# Enable cURL
ENV LLAMA_CURL=1
# Must be set to 0.0.0.0 so it can listen to requests from host machine
ENV LLAMA_ARG_HOST=0.0.0.0

RUN make -j$(nproc) llama-server

Expand Down
2 changes: 2 additions & 0 deletions .devops/llama-server-intel.Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,6 +26,8 @@ RUN apt-get update && \
COPY --from=build /app/build/bin/llama-server /llama-server

ENV LC_ALL=C.utf8
# Must be set to 0.0.0.0 so it can listen to requests from host machine
ENV LLAMA_ARG_HOST=0.0.0.0

HEALTHCHECK CMD [ "curl", "-f", "http://localhost:8080/health" ]

Expand Down
2 changes: 2 additions & 0 deletions .devops/llama-server-rocm.Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,6 +39,8 @@ ENV GPU_TARGETS=${ROCM_DOCKER_ARCH}
ENV GGML_HIPBLAS=1
ENV CC=/opt/rocm/llvm/bin/clang
ENV CXX=/opt/rocm/llvm/bin/clang++
# Must be set to 0.0.0.0 so it can listen to requests from host machine
ENV LLAMA_ARG_HOST=0.0.0.0

# Enable cURL
ENV LLAMA_CURL=1
Expand Down
2 changes: 2 additions & 0 deletions .devops/llama-server-vulkan.Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,8 @@ RUN cp /app/build/bin/llama-server /llama-server && \
rm -rf /app

ENV LC_ALL=C.utf8
# Must be set to 0.0.0.0 so it can listen to requests from host machine
ENV LLAMA_ARG_HOST=0.0.0.0

HEALTHCHECK CMD [ "curl", "-f", "http://localhost:8080/health" ]

Expand Down
2 changes: 2 additions & 0 deletions .devops/llama-server.Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,8 @@ RUN apt-get update && \
COPY --from=build /app/llama-server /llama-server

ENV LC_ALL=C.utf8
# Must be set to 0.0.0.0 so it can listen to requests from host machine
ENV LLAMA_ARG_HOST=0.0.0.0

HEALTHCHECK CMD [ "curl", "-f", "http://localhost:8080/health" ]

Expand Down
7 changes: 7 additions & 0 deletions common/common.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -327,6 +327,10 @@ bool gpt_params_parse_ex(int argc, char ** argv, gpt_params & params) {
void gpt_params_parse_from_env(gpt_params & params) {
// we only care about server-related params for now
get_env("LLAMA_ARG_MODEL", params.model);
get_env("LLAMA_ARG_MODEL_URL", params.model_url);
get_env("LLAMA_ARG_MODEL_ALIAS", params.model_alias);
get_env("LLAMA_ARG_HF_REPO", params.hf_repo);
get_env("LLAMA_ARG_HF_FILE", params.hf_file);
get_env("LLAMA_ARG_THREADS", params.n_threads);
get_env("LLAMA_ARG_CTX_SIZE", params.n_ctx);
get_env("LLAMA_ARG_N_PARALLEL", params.n_parallel);
Expand All @@ -341,6 +345,9 @@ void gpt_params_parse_from_env(gpt_params & params) {
get_env("LLAMA_ARG_EMBEDDINGS", params.embedding);
get_env("LLAMA_ARG_FLASH_ATTN", params.flash_attn);
get_env("LLAMA_ARG_DEFRAG_THOLD", params.defrag_thold);
get_env("LLAMA_ARG_CONT_BATCHING", params.cont_batching);
get_env("LLAMA_ARG_HOST", params.hostname);
get_env("LLAMA_ARG_PORT", params.port);
}

bool gpt_params_parse(int argc, char ** argv, gpt_params & params) {
Expand Down
4 changes: 2 additions & 2 deletions examples/llava/README-minicpmv2.5.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,8 +15,8 @@ cd llama.cpp
Convert PyTorch model to gguf files (You can also download the converted [gguf](https://huggingface.co/openbmb/MiniCPM-Llama3-V-2_5-gguf) by us)

```bash
python ./examples/minicpmv/minicpmv-surgery.py -m ../MiniCPM-Llama3-V-2_5
python ./examples/minicpmv/minicpmv-convert-image-encoder-to-gguf.py -m ../MiniCPM-Llama3-V-2_5 --minicpmv-projector ../MiniCPM-Llama3-V-2_5/minicpmv.projector --output-dir ../MiniCPM-Llama3-V-2_5/ --image-mean 0.5 0.5 0.5 --image-std 0.5 0.5 0.5 --minicpmv_version 2
python ./examples/llava/minicpmv-surgery.py -m ../MiniCPM-Llama3-V-2_5
python ./examples/llava/minicpmv-convert-image-encoder-to-gguf.py -m ../MiniCPM-Llama3-V-2_5 --minicpmv-projector ../MiniCPM-Llama3-V-2_5/minicpmv.projector --output-dir ../MiniCPM-Llama3-V-2_5/ --image-mean 0.5 0.5 0.5 --image-std 0.5 0.5 0.5 --minicpmv_version 2
python ./convert_hf_to_gguf.py ../MiniCPM-Llama3-V-2_5/model

# quantize int4 version
Expand Down
60 changes: 43 additions & 17 deletions examples/server/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -249,23 +249,49 @@ logging:

Available environment variables (if specified, these variables will override parameters specified in arguments):

- `LLAMA_CACHE` (cache directory, used by `--hf-repo`)
- `HF_TOKEN` (Hugging Face access token, used when accessing a gated model with `--hf-repo`)
- `LLAMA_ARG_MODEL`
- `LLAMA_ARG_THREADS`
- `LLAMA_ARG_CTX_SIZE`
- `LLAMA_ARG_N_PARALLEL`
- `LLAMA_ARG_BATCH`
- `LLAMA_ARG_UBATCH`
- `LLAMA_ARG_N_GPU_LAYERS`
- `LLAMA_ARG_THREADS_HTTP`
- `LLAMA_ARG_CHAT_TEMPLATE`
- `LLAMA_ARG_N_PREDICT`
- `LLAMA_ARG_ENDPOINT_METRICS`
- `LLAMA_ARG_ENDPOINT_SLOTS`
- `LLAMA_ARG_EMBEDDINGS`
- `LLAMA_ARG_FLASH_ATTN`
- `LLAMA_ARG_DEFRAG_THOLD`
- `LLAMA_CACHE`: cache directory, used by `--hf-repo`
- `HF_TOKEN`: Hugging Face access token, used when accessing a gated model with `--hf-repo`
- `LLAMA_ARG_MODEL`: equivalent to `-m`
- `LLAMA_ARG_MODEL_URL`: equivalent to `-mu`
- `LLAMA_ARG_MODEL_ALIAS`: equivalent to `-a`
- `LLAMA_ARG_HF_REPO`: equivalent to `--hf-repo`
- `LLAMA_ARG_HF_FILE`: equivalent to `--hf-file`
- `LLAMA_ARG_THREADS`: equivalent to `-t`
- `LLAMA_ARG_CTX_SIZE`: equivalent to `-c`
- `LLAMA_ARG_N_PARALLEL`: equivalent to `-np`
- `LLAMA_ARG_BATCH`: equivalent to `-b`
- `LLAMA_ARG_UBATCH`: equivalent to `-ub`
- `LLAMA_ARG_N_GPU_LAYERS`: equivalent to `-ngl`
- `LLAMA_ARG_THREADS_HTTP`: equivalent to `--threads-http`
- `LLAMA_ARG_CHAT_TEMPLATE`: equivalent to `--chat-template`
- `LLAMA_ARG_N_PREDICT`: equivalent to `-n`
- `LLAMA_ARG_ENDPOINT_METRICS`: if set to `1`, it will enable metrics endpoint (equivalent to `--metrics`)
- `LLAMA_ARG_ENDPOINT_SLOTS`: if set to `0`, it will **disable** slots endpoint (equivalent to `--no-slots`). This feature is enabled by default.
- `LLAMA_ARG_EMBEDDINGS`: if set to `1`, it will enable embeddings endpoint (equivalent to `--embeddings`)
- `LLAMA_ARG_FLASH_ATTN`: if set to `1`, it will enable flash attention (equivalent to `-fa`)
- `LLAMA_ARG_CONT_BATCHING`: if set to `0`, it will **disable** continuous batching (equivalent to `--no-cont-batching`). This feature is enabled by default.
- `LLAMA_ARG_DEFRAG_THOLD`: equivalent to `-dt`
- `LLAMA_ARG_HOST`: equivalent to `--host`
- `LLAMA_ARG_PORT`: equivalent to `--port`

Example usage of docker compose with environment variables:

```yml
services:
llamacpp-server:
image: ghcr.io/ggerganov/llama.cpp:server
ports:
- 8080:8080
volumes:
- ./models:/models
environment:
# alternatively, you can use "LLAMA_ARG_MODEL_URL" to download the model
LLAMA_ARG_MODEL: /models/my_model.gguf
LLAMA_ARG_CTX_SIZE: 4096
LLAMA_ARG_N_PARALLEL: 2
LLAMA_ARG_ENDPOINT_METRICS: 1 # to disable, either remove or set to 0
LLAMA_ARG_PORT: 8080
```

## Build

Expand Down
1 change: 1 addition & 0 deletions ggml/include/ggml-backend.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,6 +63,7 @@ extern "C" {
GGML_API void ggml_backend_tensor_set_async(ggml_backend_t backend, struct ggml_tensor * tensor, const void * data, size_t offset, size_t size);
GGML_API void ggml_backend_tensor_get_async(ggml_backend_t backend, const struct ggml_tensor * tensor, void * data, size_t offset, size_t size);

// "offset" refers to the offset of the tensor data for setting/getting data
GGML_API GGML_CALL void ggml_backend_tensor_set( struct ggml_tensor * tensor, const void * data, size_t offset, size_t size);
GGML_API GGML_CALL void ggml_backend_tensor_get(const struct ggml_tensor * tensor, void * data, size_t offset, size_t size);

Expand Down
123 changes: 85 additions & 38 deletions ggml/include/ggml.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -220,7 +220,7 @@
#include <stdio.h>

#define GGML_FILE_MAGIC 0x67676d6c // "ggml"
#define GGML_FILE_VERSION 1
#define GGML_FILE_VERSION 2

#define GGML_QNT_VERSION 2 // bump this on quantization format changes
#define GGML_QNT_VERSION_FACTOR 1000 // do not change this
Expand Down Expand Up @@ -453,6 +453,8 @@ extern "C" {
GGML_OP_SQR,
GGML_OP_SQRT,
GGML_OP_LOG,
GGML_OP_SIN,
GGML_OP_COS,
GGML_OP_SUM,
GGML_OP_SUM_ROWS,
GGML_OP_MEAN,
Expand Down Expand Up @@ -490,9 +492,11 @@ extern "C" {
GGML_OP_CLAMP,
GGML_OP_CONV_TRANSPOSE_1D,
GGML_OP_IM2COL,
GGML_OP_IM2COL_BACK,
GGML_OP_CONV_TRANSPOSE_2D,
GGML_OP_POOL_1D,
GGML_OP_POOL_2D,
GGML_OP_POOL_2D_BACK,
GGML_OP_UPSCALE, // nearest interpolate
GGML_OP_PAD,
GGML_OP_ARANGE,
Expand Down Expand Up @@ -969,6 +973,22 @@ extern "C" {
struct ggml_context * ctx,
struct ggml_tensor * a);

GGML_API struct ggml_tensor * ggml_sin(
struct ggml_context * ctx,
struct ggml_tensor * a);

GGML_API struct ggml_tensor * ggml_sin_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a);

GGML_API struct ggml_tensor * ggml_cos(
struct ggml_context * ctx,
struct ggml_tensor * a);

GGML_API struct ggml_tensor * ggml_cos_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a);

// return scalar
GGML_API struct ggml_tensor * ggml_sum(
struct ggml_context * ctx,
Expand Down Expand Up @@ -1566,34 +1586,49 @@ extern "C" {
float min,
float max);

// im2col
// converts data into a format that effectively results in a convolution when combined with matrix multiplication
GGML_API struct ggml_tensor * ggml_im2col(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b,
int s0,
int s1,
int p0,
int p1,
int d0,
int d1,
bool is_2D,
enum ggml_type dst_type);
struct ggml_tensor * a, // convolution kernel
struct ggml_tensor * b, // data
int s0, // stride dimension 0
int s1, // stride dimension 1
int p0, // padding dimension 0
int p1, // padding dimension 1
int d0, // dilation dimension 0
int d1, // dilation dimension 1
bool is_2D,
enum ggml_type dst_type);

GGML_API struct ggml_tensor * ggml_im2col_back(
struct ggml_context * ctx,
struct ggml_tensor * a, // convolution kernel
struct ggml_tensor * b, // gradient of im2col output
int64_t * ne, // shape of im2col input
int s0, // stride dimension 0
int s1, // stride dimension 1
int p0, // padding dimension 0
int p1, // padding dimension 1
int d0, // dilation dimension 0
int d1, // dilation dimension 1
bool is_2D);

GGML_API struct ggml_tensor * ggml_conv_depthwise_2d(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b,
int s0,
int s1,
int p0,
int p1,
int d0,
int d1);
struct ggml_tensor * a, // convolution kernel
struct ggml_tensor * b, // data
int s0, // stride dimension 0
int s1, // stride dimension 1
int p0, // padding dimension 0
int p1, // padding dimension 1
int d0, // dilation dimension 0
int d1); // dilation dimension 1

GGML_API struct ggml_tensor * ggml_conv_1d(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b,
struct ggml_tensor * a, // convolution kernel
struct ggml_tensor * b, // data
int s0, // stride
int p0, // padding
int d0); // dilation
Expand All @@ -1602,29 +1637,29 @@ extern "C" {
// alias for ggml_conv_1d(a, b, s, a->ne[0]/2, d)
GGML_API struct ggml_tensor* ggml_conv_1d_ph(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b,
int s,
int d);
struct ggml_tensor * a, // convolution kernel
struct ggml_tensor * b, // data
int s, // stride
int d); // dilation

GGML_API struct ggml_tensor * ggml_conv_transpose_1d(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b,
int s0,
int p0,
int d0);
struct ggml_tensor * a, // convolution kernel
struct ggml_tensor * b, // data
int s0, // stride
int p0, // padding
int d0); // dilation

GGML_API struct ggml_tensor * ggml_conv_2d(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b,
int s0,
int s1,
int p0,
int p1,
int d0,
int d1);
struct ggml_tensor * a, // convolution kernel
struct ggml_tensor * b, // data
int s0, // stride dimension 0
int s1, // stride dimension 1
int p0, // padding dimension 0
int p1, // padding dimension 1
int d0, // dilation dimension 0
int d1); // dilation dimension 1


// kernel size is a->ne[0] x a->ne[1]
Expand Down Expand Up @@ -1686,6 +1721,18 @@ extern "C" {
float p0,
float p1);

GGML_API struct ggml_tensor * ggml_pool_2d_back(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * af, // "a"/input used in forward pass
enum ggml_op_pool op,
int k0,
int k1,
int s0,
int s1,
float p0,
float p1);

// nearest interpolate
// multiplies ne0 and ne1 by scale factor
// used in stable-diffusion
Expand Down
Loading
Loading