mirror of
https://github.com/m5stack/StackFlow.git
synced 2026-05-20 11:32:11 -07:00
650
This commit is contained in:
@@ -36,4 +36,69 @@
|
||||
| whisper-base | 660.31 | 51.11 | v0.4 | v1.7 |
|
||||
| whisper-small | 1606.08 | 148.92 | v0.4 | v1.7 |
|
||||
|
||||
`The STT test uses a 30-second wav English audio`
|
||||
|
||||
## Pyramid (AX650C)
|
||||
|
||||
### LLM
|
||||
| model | ttft (ms) | avg-token/s | model version | llm version |
|
||||
|---------------------------------|------------|-------------|---------------|-------------|
|
||||
| qwen2.5-0.5B-Int4-ax650 | 136.65 | 40.01 | v0.4 | v1.8 |
|
||||
| qwen2.5-1.5B-Int4-ax650 | 285.93 | 17.62 | v0.4 | v1.8 |
|
||||
| qwen2.5-3B-Int4-ax650 | 546.34 | 9.79 | v0.4 | v1.8 |
|
||||
| qwen2.5-7B-Int4-ax650 | - | - | v0.4 | v1.8 |
|
||||
| deepseek-r1-1.5B-Int4-ax650 | 281.88 | 18.31 | v0.4 | v1.8 |
|
||||
| deepseek-r1-7B-Int4-ax650 | - | - | v0.4 | v1.8 |
|
||||
|
||||
`The input text used by the llm test is "hello!“`
|
||||
|
||||
### VLM
|
||||
| model | ttft (ms) | avg-token/s | image encode (ms) | model version | vlm version |
|
||||
|---------------------------------|------------|-------------|-------------------|---------------|-------------|
|
||||
| internvl2.5-1B-448-ax650 | 430.86 | 29.76 | 370.82 | v0.4 | v1.7 |
|
||||
| smolvlm-256M-ax650 | 57.79 | 74.22 | 104.16 | v0.4 | v1.7 |
|
||||
| smolvlm-500M-ax650 | - | - | - | v0.4 | v1.7 |
|
||||
|
||||
`The image encoding test uses a jpg image with a size of 810*1080`
|
||||
|
||||
|
||||
### STT
|
||||
| model | encode (ms) | avg-decode (ms) | model version | whisper version |
|
||||
|--------------------|-------------|-----------------|---------------|-----------------|
|
||||
| whisper-tiny-ax650 | 21.19 | 5.72 | v0.4 | v1.7 |
|
||||
| whisper-base-ax650 | 50.1 | 9.93 | v0.4 | v1.7 |
|
||||
| whisper-small-ax650| 188.98 | 29.77 | v0.4 | v1.7 |
|
||||
|
||||
`The STT test uses a 30-second wav English audio`
|
||||
|
||||
## Sipeed (AX650N)
|
||||
|
||||
### LLM
|
||||
| model | ttft (ms) | avg-token/s | model version | llm version |
|
||||
|---------------------------------|------------|-------------|---------------|-------------|
|
||||
| qwen2.5-0.5B-Int4-ax650 | 134.49 | 42.34 | v0.4 | v1.8 |
|
||||
| qwen2.5-1.5B-Int4-ax650 | 282.07 | 18.31 | v0.4 | v1.8 |
|
||||
| qwen2.5-3B-Int4-ax650 | 543.67 | 9.73 | v0.4 | v1.8 |
|
||||
| qwen2.5-7B-Int4-ax650 | 1136.33 | 4.54 | v0.4 | v1.8 | 4335MB CMM
|
||||
| deepseek-r1-1.5B-Int4-ax650 | 281.80 | 18.37 | v0.4 | v1.8 |
|
||||
| deepseek-r1-7B-Int4-ax650 | - | - | v0.4 | v1.8 |
|
||||
|
||||
`The input text used by the llm test is "hello!“`
|
||||
|
||||
### VLM
|
||||
| model | ttft (ms) | avg-token/s | image encode (ms) | model version | vlm version |
|
||||
|---------------------------------|------------|-------------|-------------------|---------------|-------------|
|
||||
| internvl2.5-1B-448-ax650 | 427.35 | 30.44 | 371.7 | v0.4 | v1.7 |
|
||||
| smolvlm-256M-ax650 | 57.31 | 74.16 | 106.44 | v0.4 | v1.7 |
|
||||
|
||||
`The image encoding test uses a jpg image with a size of 810*1080`
|
||||
|
||||
|
||||
### STT
|
||||
| model | encode (ms) | avg-decode (ms) | model version | whisper version |
|
||||
|--------------------|-------------|-----------------|---------------|-----------------|
|
||||
| whisper-tiny-ax650 | 21.16 | 5.62 | v0.4 | v1.7 |
|
||||
| whisper-base-ax650 | 50.21 | 9.68 | v0.4 | v1.7 |
|
||||
| whisper-small-ax650| 188.69 | 28.34 | v0.4 | v1.7 |
|
||||
|
||||
`The STT test uses a 30-second wav English audio`
|
||||
@@ -54,3 +54,59 @@ if 'CONFIG_AX_620E_MSP_ENABLED' in os.environ:
|
||||
'REGISTER':'static'
|
||||
})
|
||||
|
||||
if 'CONFIG_AX_650N_MSP_ENABLED' in os.environ:
|
||||
MSP_PATH = ''
|
||||
if 'CONFIG_MSP_V3_0_0_ENABLED' in os.environ:
|
||||
MSP_PATH = os.environ.get('EXT_MSP_PATH', check_wget_down("https://m5stack.oss-cn-shenzhen.aliyuncs.com/resource/linux/llm/m5stack_ax650n_msp.tar.gz", 'm5stack_ax650n_msp.tar.gz'))
|
||||
|
||||
SRCS=[]
|
||||
INCLUDE=[]
|
||||
PRIVATE_INCLUDE=[]
|
||||
REQUIREMENTS=[]
|
||||
STATIC_LIB=[]
|
||||
DYNAMIC_LIB=[]
|
||||
DEFINITIONS=[]
|
||||
DEFINITIONS_PRIVATE=['']
|
||||
LDFLAGS=[]
|
||||
LINK_SEARCH_PATH=[]
|
||||
|
||||
env["MSP_PATH"] = MSP_PATH
|
||||
# , os.path.join(MSP_PATH, 'sample/common')
|
||||
INCLUDE += [os.path.join(MSP_PATH, 'out/include'), os.path.join(MSP_PATH, 'sample'), os.path.join(MSP_PATH, 'sample/rtsp')]
|
||||
LINK_SEARCH_PATH += [os.path.join(MSP_PATH, 'out/lib')]
|
||||
REQUIREMENTS += ['dl', 'm', 'pthread']
|
||||
third_party = ['drm','faac','fdk-aac','ffmpeg','libsamplerate','mp4','opencv','opus','tinyalsa']
|
||||
for dirn in third_party:
|
||||
INCLUDE.append(os.path.join(MSP_PATH,'third-party',dirn,'include'))
|
||||
LINK_SEARCH_PATH.append(os.path.join(MSP_PATH,'third-party',dirn,'lib/arm64/glibc'))
|
||||
INCLUDE.append(os.path.join(MSP_PATH,'third-party/live/out/include/liveMedia'))
|
||||
INCLUDE.append(os.path.join(MSP_PATH,'third-party/live/out/include/groupsock'))
|
||||
INCLUDE.append(os.path.join(MSP_PATH,'third-party/live/out/include/UsageEnvironment'))
|
||||
INCLUDE.append(os.path.join(MSP_PATH,'third-party/live/out/include/BasicUsageEnvironment'))
|
||||
INCLUDE.append(os.path.join(MSP_PATH,'component/isp_proton/sensor/platform/tcm'))
|
||||
LINK_SEARCH_PATH.append(os.path.join(MSP_PATH,'third-party/libsamplerate/lib'))
|
||||
INCLUDE.append(os.path.join(MSP_PATH,'third-party/openssl/include'))
|
||||
LINK_SEARCH_PATH.append(os.path.join(MSP_PATH,'third-party/openssl/arm64/glibc/lib'))
|
||||
LINK_SEARCH_PATH.append(os.path.join(MSP_PATH,'third-party/tinyalsa/lib'))
|
||||
LINK_SEARCH_PATH.append(os.path.join(MSP_PATH,'third-party/fdk-aac/lib'))
|
||||
LINK_SEARCH_PATH.append(os.path.join(MSP_PATH,'third-party/opus/lib'))
|
||||
LINK_SEARCH_PATH.append(os.path.join(MSP_PATH,'app/lib'))
|
||||
|
||||
if 'CONFIG_SAMPLE_COMMON_ENABLED' in os.environ:
|
||||
if 'CONFIG_MSP_V3_0_0_ENABLED' in os.environ:
|
||||
INCLUDE.append(os.path.join(MSP_PATH,'sample/common'))
|
||||
SRCS += AGlob(os.path.join(MSP_PATH,'sample/common/*.c*'))
|
||||
|
||||
env['COMPONENTS'].append({'target':os.path.basename(env['component_dir']),
|
||||
'SRCS':SRCS,
|
||||
'INCLUDE':INCLUDE,
|
||||
'PRIVATE_INCLUDE':PRIVATE_INCLUDE,
|
||||
'REQUIREMENTS':REQUIREMENTS,
|
||||
'STATIC_LIB':STATIC_LIB,
|
||||
'DYNAMIC_LIB':DYNAMIC_LIB,
|
||||
'DEFINITIONS':DEFINITIONS,
|
||||
'DEFINITIONS_PRIVATE':DEFINITIONS_PRIVATE,
|
||||
'LDFLAGS':LDFLAGS,
|
||||
'LINK_SEARCH_PATH':LINK_SEARCH_PATH,
|
||||
'REGISTER':'static'
|
||||
})
|
||||
+1
-1
@@ -1134,7 +1134,7 @@ AX_S32 main(AX_S32 argc, AX_CHAR *argv[])
|
||||
// init engine
|
||||
AX_ENGINE_NPU_ATTR_T npu_attr;
|
||||
memset(&npu_attr, 0, sizeof(npu_attr));
|
||||
#if defined(CHIP_AX650)
|
||||
#if defined(CONFIG_AX_650N_MSP_ENABLED)
|
||||
if (nNpuType == 0) {
|
||||
npu_attr.eHardMode = AX_ENGINE_VIRTUAL_NPU_DISABLE;
|
||||
} else if (nNpuType >= 1 && nNpuType <= 4) {
|
||||
+1
-1
@@ -1073,7 +1073,7 @@ AX_S32 main(AX_S32 argc, AX_CHAR *argv[]) {
|
||||
// init engine
|
||||
AX_ENGINE_NPU_ATTR_T npu_attr;
|
||||
memset(&npu_attr, 0, sizeof(npu_attr));
|
||||
#if defined(CHIP_AX650)
|
||||
#if defined(CONFIG_AX_650N_MSP_ENABLED)
|
||||
if (nNpuType == 0) {
|
||||
npu_attr.eHardMode = AX_ENGINE_VIRTUAL_NPU_DISABLE;
|
||||
}
|
||||
@@ -22,9 +22,10 @@ DEFINITIONS += ['-std=c++17']
|
||||
LDFLAGS+=['-Wl,-rpath=/opt/m5stack/lib', '-Wl,-rpath=/usr/local/m5stack/lib', '-Wl,-rpath=/usr/local/m5stack/lib/gcc-10.3', '-Wl,-rpath=/opt/lib', '-Wl,-rpath=/opt/usr/lib', '-Wl,-rpath=./']
|
||||
# INCLUDE += [ADir('../include'), ADir('../include/cppzmq'), ADir('../include/libzmq')]
|
||||
LINK_SEARCH_PATH += [ADir('../static_lib')]
|
||||
|
||||
REQUIREMENTS += ['ax_sys', 'ax_interpreter', 'ax_audio', 'ax_audio_3a', 'ax_fdk', 'ax_opus']
|
||||
REQUIREMENTS += ['tinyalsa', 'opus', 'samplerate', 'fdk-aac']
|
||||
REQUIREMENTS += ['ax_sys', 'ax_audio', 'ax_audio_3a', 'tinyalsa', 'fdk-aac', 'opus', 'samplerate']
|
||||
if 'CONFIG_AX_620E_MSP_ENABLED' in os.environ:
|
||||
REQUIREMENTS += ['ax_sys', 'ax_interpreter', 'ax_audio', 'ax_audio_3a', 'ax_fdk', 'ax_opus']
|
||||
REQUIREMENTS += ['tinyalsa', 'opus', 'samplerate', 'fdk-aac']
|
||||
|
||||
STATIC_FILES += [AFile('audio.json')]
|
||||
STATIC_FILES += Glob('mode_*.json')
|
||||
|
||||
File diff suppressed because it is too large
Load Diff
Some files were not shown because too many files have changed in this diff Show More
Reference in New Issue
Block a user