You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: README.md
+33-36Lines changed: 33 additions & 36 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -51,42 +51,6 @@ GPULlama3ChatModel model = GPULlama3ChatModel.builder()
51
51
#### **[Interactive-mode]** Running on a RTX 5090 with nvtop on bottom to track GPU utilization and memory usage.
52
52
53
53
54
-
-----------
55
-
#### **[Instruct-mode]** Running on a RTX 5090
56
-
57
-
58
-
----------
59
-
60
-
### TornadoVM-Accelerated Inference Performance and Optimization Status
61
-
62
-
We are at the early stages of Java entering the AI world with features added to the JVM that enable faster execution such as GPU acceleration, Vector acceleration, high-performance access to off-heap memory and others.
63
-
<br><br>This repository provides the first Java-native implementation of Llama3 that automatically compiles and executes Java code on GPUs via TornadoVM.
64
-
The baseline numbers presented below provide a solid starting point for achieving more competitive performance compared to llama.cpp or native CUDA implementations.
65
-
[Our roadmap](https://github.com/beehive-lab/GPULlama3.java/blob/main/docs/GPULlama3_ROADMAP.md) provides the upcoming set of features that will dramatically improve the numbers below with the clear target being to achieve performance parity with the fastest implementations.
66
-
<br><br>
67
-
If you achieve additional performance data points (e.g. new hardware or platforms) please let us know to add them below.
68
-
<br><br>
69
-
In addition, if you are interested to learn more about the challenges of managed programming languages and GPU acceleration, you can read [our book](https://link.springer.com/book/10.1007/978-3-031-49559-5) or consult the [TornadoVM educational pages](https://www.tornadovm.org/resources).
TornadoVM currently runs on Apple Silicon via [OpenCL](https://developer.apple.com/opencl/), which has been officially deprecated since macOS 10.14.
87
-
88
-
Despite being deprecated, OpenCL can still run on Apple Silicon; albeit, with older drivers which do not support all optimizations of TornadoVM. Therefore, the performance is not optimal since TornadoVM does not have a Metal backend yet (it currently has OpenCL, PTX, and SPIR-V backends). We recommend using Apple silicon for development and for performance testing to use OpenCL/PTX compatible Nvidia GPUs for the time being (until we add a Metal backend to TornadoVM and start optimizing it).
89
-
90
54
91
55
-----------
92
56
@@ -159,6 +123,39 @@ make
159
123
./llama-tornado --gpu --verbose-init --opencl --model beehive-llama-3.2-1b-instruct-fp16.gguf --prompt "tell me a joke"
160
124
```
161
125
126
+
127
+
----------
128
+
129
+
### TornadoVM-Accelerated Inference Performance and Optimization Status
130
+
131
+
We are at the early stages of Java entering the AI world with features added to the JVM that enable faster execution such as GPU acceleration, Vector acceleration, high-performance access to off-heap memory and others.
132
+
<br><br>This repository provides the first Java-native implementation of Llama3 that automatically compiles and executes Java code on GPUs via TornadoVM.
133
+
The baseline numbers presented below provide a solid starting point for achieving more competitive performance compared to llama.cpp or native CUDA implementations.
134
+
[Our roadmap](https://github.com/beehive-lab/GPULlama3.java/blob/main/docs/GPULlama3_ROADMAP.md) provides the upcoming set of features that will dramatically improve the numbers below with the clear target being to achieve performance parity with the fastest implementations.
135
+
<br><br>
136
+
If you achieve additional performance data points (e.g. new hardware or platforms) please let us know to add them below.
137
+
<br><br>
138
+
In addition, if you are interested to learn more about the challenges of managed programming languages and GPU acceleration, you can read [our book](https://link.springer.com/book/10.1007/978-3-031-49559-5) or consult the [TornadoVM educational pages](https://www.tornadovm.org/resources).
TornadoVM currently runs on Apple Silicon via [OpenCL](https://developer.apple.com/opencl/), which has been officially deprecated since macOS 10.14.
156
+
157
+
Despite being deprecated, OpenCL can still run on Apple Silicon; albeit, with older drivers which do not support all optimizations of TornadoVM. Therefore, the performance is not optimal since TornadoVM does not have a Metal backend yet (it currently has OpenCL, PTX, and SPIR-V backends). We recommend using Apple silicon for development and for performance testing to use OpenCL/PTX compatible Nvidia GPUs for the time being (until we add a Metal backend to TornadoVM and start optimizing it).
0 commit comments