Dataset Viewer
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
arm: string
cell_id: string
concurrency: int64
corpus_receipt_sha256: string
experiment_id: string
model_revision: string
optimistic_prefill: int64
protected_preflight_receipt: string
public_base_image: string
recorded_at: string
router_base_url: string
router_source_tree: string
run_id: string
runtime_bundle_sha256: string
runtime_commit: string
schema_version: string
trace_root: string
trie_source_tree: string
finalized_at: string
result: struct<cached_prompt_tokens: int64, completed_completion_tokens: int64, completed_model_requests: in (... 2332 chars omitted)
child 0, cached_prompt_tokens: int64
child 1, completed_completion_tokens: int64
child 2, completed_model_requests: int64
child 3, completed_prompt_tokens: int64
child 4, completed_requests: int64
child 5, distribution_summaries: struct<block_aligned_prefix_tokens: struct<max: double, mean: double, min: double, p50: double, p90: (... 1088 chars omitted)
child 0, block_aligned_prefix_tokens: struct<max: double, mean: double, min: double, p50: double, p90: double, p95: double, p99: double>
child 0, max: double
child 1, mean: double
child 2, min: double
child 3, p50: double
child 4, p90: double
child 5, p95: double
child 6, p99: double
child 1, cache_hit_rate: struct<max: double, mean: double, min: double, p50: double, p90: double, p95: double, p99: double>
child 0, max: double
child 1, mean: double
...
string
child 7, exactness_expected_value: int64
child 8, exactness_metrics: list<item: string>
child 0, item: string
child 9, forbidden_shared_denominator: string
child 10, glm_dsa_full_only_required_zero_metrics: list<item: string>
child 0, item: string
child 11, glm_dsa_full_only_swa_capacity_tokens: int64
child 12, interval_seconds: int64
child 13, invariant_rank_consistent_metrics: list<item: string>
child 0, item: string
child 14, maximum_torn_scrape_fraction: double
child 15, measurement_semantics: string
child 16, metrics: list<item: string>
child 0, item: string
child 17, physical_tensor_hbm_pages_or_bytes_claimed: bool
child 18, prometheus_page_size_metric: string
child 19, reference_only_metrics: list<item: string>
child 0, item: string
child 20, require_exactness_every_sample: bool
child 21, require_rank_presence_every_sample: bool
child 22, require_torn_scrapes_isolated_and_bracketed: bool
child 23, required_analysis: list<item: string>
child 0, item: string
child 24, required_tp_ranks: list<item: int64>
child 0, item: int64
child 25, token_seconds_metrics: list<item: string>
child 0, item: string
child 26, tp8_rank_present_metric_count: int64
files: list<item: struct<path: string, sha256: string, size_bytes: int64>>
child 0, item: struct<path: string, sha256: string, size_bytes: int64>
child 0, path: string
child 1, sha256: string
child 2, size_bytes: int64
to
{'cell_id': Value('string'), 'concurrency': Value('int64'), 'decode_preallocation_measurement': {'allocator_page_id_metrics': List(Value('string')), 'allocator_request_count_metrics': List(Value('string')), 'allocator_token_slot_metrics': List(Value('string')), 'bytes_exact': Value('bool'), 'capacity_and_page_metrics': List(Value('string')), 'capacity_fractions': {'full': {'denominator_components': List(Value('string')), 'numerator': Value('string')}, 'swa': {'denominator_components': List(Value('string')), 'numerator': Value('string')}}, 'dynamic_non_atomic_metrics': List(Value('string')), 'exactness_expected_value': Value('int64'), 'exactness_metrics': List(Value('string')), 'forbidden_shared_denominator': Value('string'), 'glm_dsa_full_only_required_zero_metrics': List(Value('string')), 'glm_dsa_full_only_swa_capacity_tokens': Value('int64'), 'interval_seconds': Value('int64'), 'invariant_rank_consistent_metrics': List(Value('string')), 'maximum_torn_scrape_fraction': Value('float64'), 'measurement_semantics': Value('string'), 'metrics': List(Value('string')), 'physical_tensor_hbm_pages_or_bytes_claimed': Value('bool'), 'prometheus_page_size_metric': Value('string'), 'reference_only_metrics': List(Value('string')), 'require_exactness_every_sample': Value('bool'), 'require_rank_presence_every_sample': Value('bool'), 'require_torn_scrapes_isolated_and_bracketed': Value('bool'), 'required_analysis': List(Value('string')), 'required_tp_ranks': List(Value('int64')), 'token_seco
...
')}}, 'expected_requests': Value('null'), 'failed_requests': Value('int64'), 'new_prompt_tokens': Value('int64'), 'samples': {'block_aligned_prefix_tokens': List(Value('int64')), 'cache_hit_rate': List(Value('float64')), 'client_prefix_tokens': List(Value('int64')), 'decode_tpot_ms': List(Value('float64')), 'eligible_cache_hit_rate': List(Value('float64')), 'inter_token_latency_ms': List(Value('float64')), 'latency_s': List(Value('float64')), 'server_cached_tokens': List(Value('int64')), 'ttfat_s': List(Value('float64')), 'ttft_s': List(Value('float64'))}, 'schema_version': Value('int64'), 'throughput': {'last_30s': {'cached_prompt_tok_s': Value('float64'), 'completion_tok_s': Value('float64'), 'new_prompt_tok_s': Value('float64'), 'prompt_tok_s': Value('float64')}, 'overall': {'cached_prompt_tok_s': Value('float64'), 'completion_tok_s': Value('float64'), 'new_prompt_tok_s': Value('float64'), 'prompt_tok_s': Value('float64')}, 'steady_state': {'cached_prompt_tok_s': Value('float64'), 'completion_tok_s': Value('float64'), 'new_prompt_tok_s': Value('float64'), 'prompt_tok_s': Value('float64')}, 'steady_state_per_gpu': {'cached_prompt_tok_s': Value('float64'), 'completion_tok_s': Value('float64'), 'new_prompt_tok_s': Value('float64'), 'prompt_tok_s': Value('float64')}}, 'trace_per_s': Value('float64'), 'wall_time_s': Value('float64')}, 'runtime_bundle_sha256': Value('string'), 'runtime_commit': Value('string'), 'schema_version': Value('string'), 'workers': List(Value('string'))}
because column names don't match
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 149, in get_rows_or_raise
return get_rows(
dataset=dataset,
...<4 lines>...
column_names=column_names,
)
File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
return func(*args, **kwargs)
File "/src/services/worker/src/worker/utils.py", line 129, in get_rows
rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
File "/src/services/worker/src/worker/utils.py", line 489, in safe_iter
yield from ds.decode(False) if ds.features else ds
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2818, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2355, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2380, in _iter_arrow
for key, pa_table in self.ex_iterable._iter_arrow():
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, in _iter_arrow
for key, pa_table in self.generate_tables_fn(**gen_kwags):
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2369, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2297, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
arm: string
cell_id: string
concurrency: int64
corpus_receipt_sha256: string
experiment_id: string
model_revision: string
optimistic_prefill: int64
protected_preflight_receipt: string
public_base_image: string
recorded_at: string
router_base_url: string
router_source_tree: string
run_id: string
runtime_bundle_sha256: string
runtime_commit: string
schema_version: string
trace_root: string
trie_source_tree: string
finalized_at: string
result: struct<cached_prompt_tokens: int64, completed_completion_tokens: int64, completed_model_requests: in (... 2332 chars omitted)
child 0, cached_prompt_tokens: int64
child 1, completed_completion_tokens: int64
child 2, completed_model_requests: int64
child 3, completed_prompt_tokens: int64
child 4, completed_requests: int64
child 5, distribution_summaries: struct<block_aligned_prefix_tokens: struct<max: double, mean: double, min: double, p50: double, p90: (... 1088 chars omitted)
child 0, block_aligned_prefix_tokens: struct<max: double, mean: double, min: double, p50: double, p90: double, p95: double, p99: double>
child 0, max: double
child 1, mean: double
child 2, min: double
child 3, p50: double
child 4, p90: double
child 5, p95: double
child 6, p99: double
child 1, cache_hit_rate: struct<max: double, mean: double, min: double, p50: double, p90: double, p95: double, p99: double>
child 0, max: double
child 1, mean: double
...
string
child 7, exactness_expected_value: int64
child 8, exactness_metrics: list<item: string>
child 0, item: string
child 9, forbidden_shared_denominator: string
child 10, glm_dsa_full_only_required_zero_metrics: list<item: string>
child 0, item: string
child 11, glm_dsa_full_only_swa_capacity_tokens: int64
child 12, interval_seconds: int64
child 13, invariant_rank_consistent_metrics: list<item: string>
child 0, item: string
child 14, maximum_torn_scrape_fraction: double
child 15, measurement_semantics: string
child 16, metrics: list<item: string>
child 0, item: string
child 17, physical_tensor_hbm_pages_or_bytes_claimed: bool
child 18, prometheus_page_size_metric: string
child 19, reference_only_metrics: list<item: string>
child 0, item: string
child 20, require_exactness_every_sample: bool
child 21, require_rank_presence_every_sample: bool
child 22, require_torn_scrapes_isolated_and_bracketed: bool
child 23, required_analysis: list<item: string>
child 0, item: string
child 24, required_tp_ranks: list<item: int64>
child 0, item: int64
child 25, token_seconds_metrics: list<item: string>
child 0, item: string
child 26, tp8_rank_present_metric_count: int64
files: list<item: struct<path: string, sha256: string, size_bytes: int64>>
child 0, item: struct<path: string, sha256: string, size_bytes: int64>
child 0, path: string
child 1, sha256: string
child 2, size_bytes: int64
to
{'cell_id': Value('string'), 'concurrency': Value('int64'), 'decode_preallocation_measurement': {'allocator_page_id_metrics': List(Value('string')), 'allocator_request_count_metrics': List(Value('string')), 'allocator_token_slot_metrics': List(Value('string')), 'bytes_exact': Value('bool'), 'capacity_and_page_metrics': List(Value('string')), 'capacity_fractions': {'full': {'denominator_components': List(Value('string')), 'numerator': Value('string')}, 'swa': {'denominator_components': List(Value('string')), 'numerator': Value('string')}}, 'dynamic_non_atomic_metrics': List(Value('string')), 'exactness_expected_value': Value('int64'), 'exactness_metrics': List(Value('string')), 'forbidden_shared_denominator': Value('string'), 'glm_dsa_full_only_required_zero_metrics': List(Value('string')), 'glm_dsa_full_only_swa_capacity_tokens': Value('int64'), 'interval_seconds': Value('int64'), 'invariant_rank_consistent_metrics': List(Value('string')), 'maximum_torn_scrape_fraction': Value('float64'), 'measurement_semantics': Value('string'), 'metrics': List(Value('string')), 'physical_tensor_hbm_pages_or_bytes_claimed': Value('bool'), 'prometheus_page_size_metric': Value('string'), 'reference_only_metrics': List(Value('string')), 'require_exactness_every_sample': Value('bool'), 'require_rank_presence_every_sample': Value('bool'), 'require_torn_scrapes_isolated_and_bracketed': Value('bool'), 'required_analysis': List(Value('string')), 'required_tp_ranks': List(Value('int64')), 'token_seco
...
')}}, 'expected_requests': Value('null'), 'failed_requests': Value('int64'), 'new_prompt_tokens': Value('int64'), 'samples': {'block_aligned_prefix_tokens': List(Value('int64')), 'cache_hit_rate': List(Value('float64')), 'client_prefix_tokens': List(Value('int64')), 'decode_tpot_ms': List(Value('float64')), 'eligible_cache_hit_rate': List(Value('float64')), 'inter_token_latency_ms': List(Value('float64')), 'latency_s': List(Value('float64')), 'server_cached_tokens': List(Value('int64')), 'ttfat_s': List(Value('float64')), 'ttft_s': List(Value('float64'))}, 'schema_version': Value('int64'), 'throughput': {'last_30s': {'cached_prompt_tok_s': Value('float64'), 'completion_tok_s': Value('float64'), 'new_prompt_tok_s': Value('float64'), 'prompt_tok_s': Value('float64')}, 'overall': {'cached_prompt_tok_s': Value('float64'), 'completion_tok_s': Value('float64'), 'new_prompt_tok_s': Value('float64'), 'prompt_tok_s': Value('float64')}, 'steady_state': {'cached_prompt_tok_s': Value('float64'), 'completion_tok_s': Value('float64'), 'new_prompt_tok_s': Value('float64'), 'prompt_tok_s': Value('float64')}, 'steady_state_per_gpu': {'cached_prompt_tok_s': Value('float64'), 'completion_tok_s': Value('float64'), 'new_prompt_tok_s': Value('float64'), 'prompt_tok_s': Value('float64')}}, 'trace_per_s': Value('float64'), 'wall_time_s': Value('float64')}, 'runtime_bundle_sha256': Value('string'), 'runtime_commit': Value('string'), 'schema_version': Value('string'), 'workers': List(Value('string'))}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Optimistic Prefill × HiCache 2P1D Ablation — full experiment artifacts
This dataset contains the complete artifacts (client traces, router receipts, engine
telemetry/logs, per-request exports, analysis scripts, and results) of a matched-pair
ablation of SGLang optimistic prefill in a PD-disaggregated deployment, run on
2026-08-01 on Together's research-b200-ic1 cluster. It is published gated so results
can be re-analyzed later without cluster access.
What is NOT here (by confidentiality policy): the private SGLang runtime source bundle (referenced only by commit/SHA pins), and the protected EdgeBench bit-exact corpus (recorded prompt/output token IDs). All lifecycle records use hashed request / trace IDs and contain no prompt or response content.
1. The experiment I ran
Question: does --optimistic-prefill-attempts 1 (prefill starts computing before
the decode worker has allocated destination KV; computed KV is withheld until decode
signals ready) improve TTFT/E2E for a real agentic workload, at what decode-pressure
regimes, and at what cost?
Fixed apparatus (identical in every cell):
- Model
GLM-5.2-FP8(revisionba978f7d347eaf65d22f1a86833408afdb953541), 2P1D fleet on 3× 8×B200 nodes: P0/P1 = TP8 + PCP8 (--enable-prefill-cp, DSA CP-shared KV), D0 = TP8/DCP1. FP8 E4M3 KV, page 64,mem-fraction-static 0.8(device pool 949,248 tokens/rank),--hicache-size 100(≈2.23M host tokens/rank), decode radix cache ON, EAGLE 5-step (top-k 1, 6 draft tokens). - Router:
sgl-model-gatewaycompiled from the pinned bundle, retries disabled (one attempt total; 120 s prefill response-header timeout). - Private runtime pin: commit
09f74119bda699e96f2e0f16aa975082266eb0b4, bundle SHA-256a7cd8b08d65bf58a25102c482120fe19c26377ad614a3ea1c9c31c5d38166e76, plus two overlay patches (prefill CP-shared HiCache barrier; decode metrics-reporter fix). Public base imagedocker.io/weili0234/sglang@sha256:02fc722b.... - Workload: EdgeBench bit-exact strict replay — 137 recorded Claude-Code agent
sessions (17,968 explicit-token requests) replayed with recorded think-time gaps
(
delay_scale=1), forced recorded output tokens, session cache salt, seed 20260730, corpus cycled to hold the cell's session count constant.
Matrix (counterbalanced A/B, B/A, A/B): 6 cells = {24, 32, 40} sessions × optimistic {0, 1}; each cell = ~6 min tokenization prep + exactly 3600 s admission + ≤900 s drain, with a verified cache-clean transition (flush + per-rank flag check) between cells on the same live fleet — no engine restarts.
How it was executed: a manual controller pod (spec: scripts/manifests/cc1-pod.json)
ran each cell via scripts/controller/runcell.sh (per-cell: transition → runtime
preflight → metrics sampler (5 s, pod-local spool) → Trie strict-replay client →
client/lifecycle/metrics checkers → finalizer). scripts/controller/phase1.sh is the
one-time gate (corpus hash verify → baseline preflight → 39-request smoke → router
stability). Full flag-level configuration of every cell is inside each cell's
cell-config.json and experiment-config receipts in this dataset.
Validity policy (important for interpretation): science gates were enforced
unchanged — full 3600+900 s window, pinned config/process-epoch identity, strict-replay
exit 0, ≤5% router-confirmed prefill-header timeouts. Observability-side perfection
gates (torn multiprocess Prometheus scrapes, metric-fetch CurlExit28 timeouts under
load, router health-failure counter ticks, drain-boundary lifecycle correlation) were
downgraded from run-invalidating assertions to recorded annotations — see
scripts/tolerant-tooling/ for the exact patched checkers and each cell's
cc1-annotations.jsonl + telemetry receipts for what fired. All six cells passed the
science gates on their first attempt (confirmed timeouts: 1/1/1/1/0/0 out of
~3,500-3,800 admitted requests per cell).
2. The results I got
2.1 Latency (script: scripts/analysis/settlement.py, sign test = on-arm wins
across twelve 300 s admission windows; win_detail.py prints per-window detail)
| pair | TTFT p50 off→on | TTFT p90 | TTFT p99 | E2E p50 | sign test (TTFT p50) |
|---|---|---|---|---|---|
| c24 | 8,480 → 7,857 ms (−7.3%) | −2.9% | +13.5% | −3.3% | 5/12 (weak) |
| c32 | 17,959 → 16,605 ms (−7.5%) | −8.4% | +2.1% | −7.6% | 8/12 |
| c40 | 30,564 → 31,019 ms (+1.5%) | −1.4% | +9.4% | −1.0% | 4/12 (wash) |
2.2 Mechanism (scripts: pair_compare.py, kv_breakdown.py, putil2.py,
extract_kv_timeseries.py + plot_kv_composition.py → kv-composition-cell01-cell02.png)
- Decode KV allocation wait is unchanged by the treatment in every pair and scales brutally with concurrency: 6.6 s (c24) → ~15 s (c32) → ~26 s (c40) mean. The mechanism overlaps prefill compute under this wait; it never shortens it.
- Prefill retries (
sglang:num_prefill_retries_total) = 0 in all six cells — the optimistic yield/requeue cost path never fired. - Decode pool composition (see PNG): ~78-85% held by active decode, prealloc-new ≈ 0% (99% of preallocation is radix-prefix references), evictable cache ~6%, free ~15%.
- Prefill is ~97% cache-served (≈210k prompt TPM/GPU submitted vs ≈6.5k uncached TPM/GPU computed) and prefill GPUs idle at ~20% mean in both arms.
2.3 Throughput (scripts: total_avg.py, tail_throughput.py,
token_throughput.py, windowed_tables.py, cell_durations.py)
- Admission-window total decode TPS (avg): c24 816/817, c32 863/866, c40 910/898 (off/on) — treatment does not change decode capacity (closed-loop workload).
- Last-30-min pace: c32 on-arm +9.8% completions/min and +4.6% decode TPS (its faster TTFT compounds); c40 on-arm −2.7% / −3.8% (late-window degradation, cause not yet isolated — leading hypothesis is prefill-side KV residency from early-computed requests pressuring the prefill radix cache over the hour).
- Per-stream decode TPS p50 ≈ 96-108 tok/s in every cell and both arms (p90/p95 of per-stream TPS are short-output artifacts; trust p25-p70).
- Every cell: 60.0 min admission exactly, ~20.7 min drain+publish, ~3 min checkers.
2.4 Conclusion
Optimistic prefill is effectively free at this workload (no retries, no decode-side change) and buys meaningful latency in the mid-pressure regime — peak at c32 (−7.5% TTFT p50, −8.4% p90, consistent across windows) — but the benefit drowns at c40 where a ~1 s hidden prefill compute is noise against 26 s allocation waits, and the c24 gain is within single-run noise. p99 movements are inconsistent in sign across pairs (single-run tail noise); replicate a pair if tails matter.
3. File layout
cells/<cell-id>/ # the six valid cells (1.3-1.5 GB each)
timed-result.json # Trie client aggregate result
lifecycle/client.jsonl # per-request client events (ADMIT/FIRST_TOKEN/COMPLETE + token counts)
lifecycle-events.jsonl # correlated worker-side events (P_RECEIVE, bootstrap queue, D stages)
logs/timed-replay.log # full Trie client log
metrics/{p0,p1,d0,router}.openmetrics # 5 s Prometheus scrapes for the whole cell
metrics/sampling-control.jsonl # scrape-cadence control stream (+errors.jsonl)
preflight/*.json # transition/runtime/metrics/lifecycle/client gates
cell-config.json, corpus-receipt.json, transition-receipt.json,
process-epochs.json, artifact-manifest.json, phase-markers.jsonl
cc1-annotations.jsonl # annotate-only findings for this cell
engine/renamed-workers-r5/ # engine-side artifacts (whole fleet lifetime, incl. prior NR attempts)
workers/{p0,p1,d0}/requests/ # per-request engine exports (no bodies)
workers/{p0,p1,d0}/server.log # engine logs
workers/{p0,p1,d0}/gpu-util.csv # 1 Hz nvidia-smi utilization
forward/ # forward-pass telemetry shards
control/, recovery/, post-schedule/ # controller context + receipts
router/ablation-2p1d-noretry-r1/ # router logs/receipts (confirmed-timeout source)
replay-attempts/ # archived invalid attempts (NR-era provenance + cc1 receipts)
scripts/ # controller, analysis, tolerant tooling, pod manifests
kv-composition-cell01-cell02.png # decode KV pool composition figure (c24 pair)
4. Reproducing the analysis
Every table above regenerates from cells/ alone. Set the cells root, then e.g.:
python scripts/analysis/settlement.py # latency tables + sign tests
python scripts/analysis/windowed_tables.py # last-30/60-min throughput tables
python scripts/analysis/token_throughput.py # TPM/GPU + total TPS + per-stream TPS
(The scripts read RESULT_ROOT from env in their original form — point root at
your local cells/ download.)
5. Provenance & caveats
- Executed by a manual controller after an earlier automated controller spent 25 attempts failing its own observability gates on this matrix; the gate-vs-science separation above is the fix. Cell 01's data was collected 2026-08-01 ~06:00 UTC; cells 02-06 ran 19:45-03:07 UTC (same fleet, zero worker restarts throughout — process-epoch receipts prove identity).
- Closed-loop strict replay: throughput numbers are pacing, not capacity ceilings.
- Engine
requests/andserver.logspan the fleet's full 21 h lifetime (including pre-takeover attempts); filter by the cell time windows inphase-markers.jsonlwhen correlating. - The controller pod's wrapper logs (
/runtime/cell0*.log) were lost with the pod; the per-cell Trie logs incells/*/logs/carry the same client-side content.
- Downloads last month
- 135