Repository navigation
Add a benchmark for the serving cost of one Literature AI answer - #8
Merged
Merged
Conversation
literature_cost_benchmark.py sends N questions (built-in set or a file)
to POST /v1/literature/answers with an omni_sk_ key, optionally in
parallel, while sampling nvidia-smi on a background thread. It reports:
- answers, statuses, throughput, latency mean/p50/p95/max
- GPU busy-seconds and Wh per answer, peak GPU memory
- cost per answer at the measured throughput ("dedicated") and for GPU
busy time only, from --gpu-hourly-cost
- a suggested price per 1,000 answers (dedicated cost x --markup)
Every answer is billable, so it asks for confirmation unless --yes, and
the README says to use a test organization's key. Each request carries
its own Idempotency-Key. Tests cover the module at 100%.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016nxi4bp7DP1ancERNUUfzj
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
literature_cost_benchmark.py sends N questions (built-in set or a file) to POST /v1/literature/answers with an omni_sk_ key, optionally in parallel, while sampling nvidia-smi on a background thread. It reports:
Every answer is billable, so it asks for confirmation unless --yes, and the README says to use a test organization's key. Each request carries its own Idempotency-Key. Tests cover the module at 100%.
Claude-Session: https://claude.ai/code/session_016nxi4bp7DP1ancERNUUfzj