[CI/Build][Hardware][NVIDIA] Add Rubin CUDA 13.4 nightly images (#55953)
Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com> Co-authored-by: OpenAI Codex <noreply@openai.com> Co-authored-by: OpenAI Codex <codex@openai.com> Co-authored-by: Misha Goin <mgoin64@gmail.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
co-authored by
OpenAI Codex
OpenAI Codex
Misha Goin
mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent
3affd35f31
commit
4f9a41ac94
@@ -177,7 +177,8 @@ def test_incremental_detokenization(
|
||||
def test_remote_prefill_cached_tokens_override(do_remote_prefill: bool):
|
||||
"""P/D disaggregation: num_cached_tokens should report the P worker's
|
||||
cache hits (passed via kv_transfer_params) instead of the local count,
|
||||
which sees the KVs pulled from the remote prefill as a ~100% hit."""
|
||||
which sees the KVs pulled from the remote prefill as a ~100% hit.
|
||||
"""
|
||||
output_processor = OutputProcessor(tokenizer=None, log_stats=False)
|
||||
|
||||
prompt_tokens = [1, 2, 3, 4, 5, 6, 7, 8]
|
||||
|
||||
@@ -350,7 +350,8 @@ def test_prefill_exports_cached_tokens_in_kv_transfer_params():
|
||||
"""The P worker reports its own prefix-cache hits in the returned
|
||||
kv_transfer_params so the D worker can surface them in
|
||||
prompt_tokens_details instead of the ~100% local hit it measures
|
||||
when pulling the KVs from the remote."""
|
||||
when pulling the KVs from the remote.
|
||||
"""
|
||||
vllm_config = create_vllm_config()
|
||||
scheduler = create_scheduler(vllm_config)
|
||||
|
||||
|
||||
Reference in New Issue
Block a user