Update VLM Reamde for v6.1 Specifications#2591
Conversation
|
MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅ |
|
|
||
| ### Configure the benchmark | ||
|
|
||
| Example configs live under [endpoints/examples/08_Qwen3-VL-235B-A22B_Example](https://github.com/mlcommons/endpoints/tree/381d13bbd27d6d52306813a51dc4e44295222d7e/examples/08_Qwen3-VL-235B-A22B_Example). Set the endpoint URL in the YAML file to match your server address and port: |
There was a problem hiding this comment.
To be explicit, we should probably specify which fields in that yaml file might be adjusted by the submitter, and which fields are frozen (in which case if a submission changed it, the submission is invalid) -> which brings up a point that maybe we need a compliance test to validate the yaml file?
There was a problem hiding this comment.
Updated this part. I believe the yaml config should be verified by submission checker?
Co-authored-by: Shang Wang <shangw@nvidia.com>
wangshangsam
left a comment
There was a problem hiding this comment.
Some nits but LGTM overall!
|
|
||
| #### Fields that the submitter **should** update to match their server status: | ||
|
|
||
| - Served model name: |
There was a problem hiding this comment.
| - Served model name: | |
| - Served model name (to match the actual, probably quantized, model checkpoint): |
| name: "Qwen/Qwen3-VL-235B-A22B-Instruct" | ||
| ``` | ||
|
|
||
| - Endpoints url and port: |
There was a problem hiding this comment.
| - Endpoints url and port: | |
| - The URL and port number of the endpoint: |
| ```yaml | ||
| load_pattern: | ||
| type: "poisson" | ||
| target_qps: 6.5 | ||
| ``` |
There was a problem hiding this comment.
| ```yaml | |
| load_pattern: | |
| type: "poisson" | |
| target_qps: 6.5 | |
| ``` | |
| ```yaml | |
| load_pattern: | |
| target_qps: 6.5 | |
| ``` |
Removing type: "poisson" to avoid the misunderstanding that type could be changed.
|
|
||
| #### Fileds that the submitter **MUST NOT** change for valid results: | ||
|
|
||
| - Sampling parameters that specified in scection [Reference Implementation Specification](#reference-implementation-specification) |
There was a problem hiding this comment.
| - Sampling parameters that specified in scection [Reference Implementation Specification](#reference-implementation-specification) | |
| - Sampling parameters specified in the section [Reference Implementation Specification](#reference-implementation-specification) |
|
|
||
| - Sampling parameters that specified in scection [Reference Implementation Specification](#reference-implementation-specification) | ||
|
|
||
| - Datasets (Neither performance or accuracy dataset) |
There was a problem hiding this comment.
| - Datasets (Neither performance or accuracy dataset) | |
| - Datasets (neither for performance evaluation nor for accuracy evaluation) |
|
|
||
| #### Fields that the submitter **may** customize for performance tuning: | ||
|
|
||
| - Target_qps (for server and interactive mode): |
There was a problem hiding this comment.
| - Target_qps (for server and interactive mode): | |
| - Target QPS (for the server and interactive scenarios): |
No description provided.