[CI][MiniCPM-o] Add MiniCPM-o 4.5 accuracy and performance coverage (#5524)

Signed-off-by: R2-Y <ruiruyang2@gmail.com>
Signed-off-by: psv666 <2693925048@qq.com>
Co-authored-by: psv666 <2693925048@qq.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: amy-why-3459 <wuhaiyan17@huawei.com>
This commit is contained in:
authored and GitHub committed 2026-08-11 14:53:49 +08:00
1 parent c4a14091c9
commit 322600bfb2
59 files changed
+1908 -155

No files matched your search

+11 -1
View File
@@ -23,11 +23,19 @@ connectors:
stages:
- stage_id: 0
max_num_seqs: 4
# Leave headroom on the shared GPU for Talker KV + concurrent Code2Wav.
gpu_memory_utilization: 0.55
enforce_eager: false
tensor_parallel_size: 1
max_num_batched_tokens: 16384
devices: "0"
# Required for MiniCPM interleaved image/audio Daily-Omni packs (~78% recipe).
# Keep these in YAML so they survive both --deploy-config and --stage-configs-path.
interleave_mm_strings: true
media_io_kwargs:
video:
fps: 1
num_frames: 128
limit_mm_per_prompt:
# Daily-Omni MiniCPM pack: up to 64 interleaved image/audio pairs (1fps).
image: 64
@@ -49,6 +57,8 @@ stages:
gpu_memory_utilization: 0.15
enforce_eager: false
max_num_batched_tokens: 8192
# Talker context is 4096 (tts_config.max_position_embeddings).
max_model_len: 4096
devices: "0"
output_connectors:
to_stage_2: connector_of_shared_memory
@@ -66,7 +76,7 @@ stages:
- stage_id: 2
max_num_seqs: 4
gpu_memory_utilization: 0.15
gpu_memory_utilization: 0.18
enforce_eager: true
enable_chunked_prefill: false
max_num_batched_tokens: 65536
+5
View File
@@ -31,6 +31,11 @@ stages:
tensor_parallel_size: 1
max_num_batched_tokens: 16384
devices: "0"
interleave_mm_strings: true
media_io_kwargs:
video:
fps: 1
num_frames: 128
limit_mm_per_prompt:
video: 1
default_sampling_params:
@@ -43,6 +43,11 @@ stages:
max_model_len: 4096
max_num_batched_tokens: 4096
devices: "0,1,2,3"
interleave_mm_strings: true
media_io_kwargs:
video:
fps: 1
num_frames: 128
limit_mm_per_prompt:
video: 1
default_sampling_params:
@@ -27,6 +27,8 @@ stages:
- stage_id: 1
max_num_seqs: 2
async_scheduling: false
# Talker context is 4096 (tts_config.max_position_embeddings).
max_model_len: 4096
# The split Talker exposes a binary continue/stop vocabulary; native
# segment and turn boundaries travel as explicit output metadata.
default_sampling_params: