ebca0c8921
Closes task #230. Measured PFlash NIAH compression at 16K-260K source contexts on 1× 24 GB / 3090 single-card. Result: PFlash works flawlessly up to 131K source. Compresses 131,068 tokens to 6,524 (5%) in 10.8s with NIAH key + answer both retained. Vanilla llama.cpp pp131072 takes ~257s per Luce's published numbers, so PFlash alone is ~24× faster at this context. End-to-end TTFT (PFlash + target prefill on 6.5K) would be ~12-13s vs ~257s = ~20×. Above 131K, drafter ephemeral forward-pass tensors (K_curr/V_curr/Q_last at full sequence length) exceed 24 GB. K-cache quantization (--pflash-k-type q8_0) doesn't help — the failing allocs are forward-pass not cache, confirmed by separate bench at 200K/260K with identical OOM at the same layer numbers. @weicj's PR #78 claim of 24K → 262K dual-GPU phase split is neither refuted nor reproduced. Their setup was 2× 22 GB Ti with target also loaded co-resident on one card; the "24K" was target+drafter combined. Our 131K is drafter-alone on 24 GB. Reproducing 262K specifically would require investigation of their drafter config (chunk_size, lookahead, BSA window) — drafter activation footprint at 200K+ is the binding constraint regardless of GPU count. Practical recommendation for 24 GB / 3090 single-card users: PFlash is shippable for source contexts ≤ 131K. The ~24× TTFT speedup is genuine and quality holds. Above 131K, fall back to vanilla llama.cpp prefill or wait for upstream drafter optimizations. Adds: - BENCHMARKS.md "PFlash long-context compression on 1× 3090" subsection with full per-context table + drafter ceiling explanation - results/lucebox-pflash-niah-20260504-150321/ (BF16 K cache run) - results/lucebox-pflash-niah-q8k-20260504-150600/ (q8_0 K cache run) This closes our active investigation of the Luce surface — three benches done (DFlash same-card 73.97 mean, K8V4 same-card 74.68 mean, PFlash compression ceiling 131K). Recommendation surface narrows to: PFlash at ≤131K is the one piece of Luce that beats vLLM dual.yml on TTFT for that workload class. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
18 lines
1.4 KiB
Plaintext
18 lines
1.4 KiB
Plaintext
[transformers] PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.
|
|
[transformers] Token indices sequence length is longer than the specified maximum sequence length for this model (251956 > 131072). Running this sequence through the model will result in indexing errors
|
|
Traceback (most recent call last):
|
|
File "/opt/ai/lucebox-hub/dflash/scripts/phase_split_dual_gpu.py", line 525, in <module>
|
|
main()
|
|
File "/opt/ai/lucebox-hub/dflash/scripts/phase_split_dual_gpu.py", line 521, in main
|
|
args.func(args)
|
|
File "/opt/ai/lucebox-hub/dflash/scripts/phase_split_dual_gpu.py", line 478, in run_bench_niah
|
|
run_cases(args, cases)
|
|
File "/opt/ai/lucebox-hub/dflash/scripts/phase_split_dual_gpu.py", line 332, in run_cases
|
|
ready_s = daemon.start()
|
|
^^^^^^^^^^^^^^
|
|
File "/opt/ai/lucebox-hub/dflash/scripts/phase_split_dual_gpu.py", line 124, in start
|
|
self.log.wait_for("[pflash-daemon] ready", 180)
|
|
File "/opt/ai/lucebox-hub/dflash/scripts/phase_split_dual_gpu.py", line 90, in wait_for
|
|
raise TimeoutError(f"timed out waiting for {needle!r}; tail={tail[-12:]}")
|
|
TimeoutError: timed out waiting for '[pflash-daemon] ready'; tail=['ggml_cuda_init: found 1 CUDA devices (Total VRAM: 24126 MiB):', ' Device 0: NVIDIA GeForce RTX 3090, compute capability 8.6, VMM: yes, VRAM: 24126 MiB']
|