{ "artifacts": [ { "file": "DeepSeek-R1-Distill-Qwen-7B_q4_block32_ekv4096.litertlm", "sha256": "511d59c11704f7ab39b9cb3a0eef1a88f8c05673d0b74bcb4b19c15845ab4a7f", "size_mb": 4322.031 } ], "benchmarks": [], "conversion": { "command": "TODO (the HF card states the recipe, not the invocation)", "quantization": "int4 blockwise (block 32) + OCTAV on linears, INT8 embedding externalized into its own bundle section; KV cache 4096 (HF card Conversion)", "tool": "litert-torch export_hf — the official upstream converter run from a clean git worktree at upstream/main, dev-fork patches excluded; Qwen2ForCausalLM rides the stock converter with no custom code (HF card Conversion)", "tool_version": "TODO (the HF card names the converter but not its version; no export log in the sources)" }, "cross_runtime": [], "delegation": null, "device": { "records": [ { "device": "galaxy-s26", "run": { "accelerator": "gpu", "context_length": null, "date": "2026-08-24", "decode_tokens_per_s": 8.43, "delegated_ops": 1243, "env": { "device": "Galaxy S26 (SM-S942Q)", "machine_label": "galaxy-s26-cl-pinned", "os_build": "Android 16", "runtime": "litert-lm", "runtime_version": "0.16.0", "soc": "Qualcomm SM8850", "vendor_sdk": null }, "error": null, "evidence": [ "VERBOSE: Replacing 1243 out of 1243 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 0 (prefill_128).", "VERBOSE: Replacing 1132 out of 1132 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 1 (decode).", "results block: prefill=107.53 decode=8.43 tokens/s" ], "failure_class": null, "full_delegation": true, "latency_p50_ms": null, "loads": true, "max_abs_diff": null, "max_rel_diff": null, "metrics": { "decode_tokens": 1089.0, "init_s": 28.15477, "prefill_tokens": 197.0 }, "output_match": null, "peak_mem_mb": null, "prefill_tokens_per_s": 107.53, "provenance": "measured", "runs": true, "total_ops": 1243, "ttft_ms": 1950.0 }, "source": "data/device_runs/0.16.0/2026-08-24/deepseek-r1-distill-qwen-7b-q4-block32-ekv4096__galaxy-s26.json" }, { "device": "raspberry-pi-5", "run": { "accelerator": "cpu", "context_length": null, "date": "2026-09-01", "decode_tokens_per_s": 1.76, "delegated_ops": null, "env": { "device": "Raspberry Pi 5 Model B Rev 1.1", "machine_label": "raspberry-pi-5-cooled-52c", "os_build": "Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41", "runtime": "litert-lm", "runtime_version": "0.16.1", "soc": "Broadcom BCM2712", "vendor_sdk": null }, "error": null, "evidence": [ "pi5 LLM sweep row: `litert-lm benchmark --backend cpu --cpu-thread-count 4 -p 256 -d 256 --runs 1 --cache memory` (wave-2 driver pi5_llm_bench.py; --cache memory rather than the house --cache no, which OOM-kills every >=1.2B file on the 8 GB Pi — equivalence measured on granite-350m int8, +2-3%), 3 invocations per file with cool-down to <=52 C between them, vcgencmd measure_temp + get_throttled logged per invocation, peak RSS polled from /proc; throughput = median of the three invocations (spread in metrics); a row counts as measured only when the real-generation gate (`litert-lm run`, degenerate-output check) passed and every invocation exited 0 with get_throttled 0x0", "versions: cpu=Raspberry Pi 5 Model B Rev 1.1, litert-lm=0.16.1, litert-lm-api=0.16.1, platform=Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41, python=3.13.5", "cache mode 'memory'; -p 256 -d 256 --runs 1 --cpu-thread-count 4", "gate ('What is 17 plus 26? Answer with the number only.'): status pass, exit 0, wall 306.0 s, output head \"[thought] I need to add 17 and 26 together. \\n\\nFirst, I'll add the units digits: \"", "invocation 0: exit=0 wall_s=522.3 temp 48.3->53.8C throttled=0x0 prefill_tps=10.41 decode_tps=1.73 ttft_s=29.8071 init_s=111.3886 peak_rss_mb=5509", "invocation 1: exit=0 wall_s=506.3 temp 49.9->53.8C throttled=0x0 prefill_tps=10.99 decode_tps=1.76 ttft_s=28.2533 init_s=111.3452 peak_rss_mb=5511", "invocation 2: exit=0 wall_s=506.3 temp 49.9->54.3C throttled=0x0 prefill_tps=11.01 decode_tps=1.77 ttft_s=28.1467 init_s=111.5353 peak_rss_mb=5497" ], "failure_class": null, "full_delegation": null, "latency_p50_ms": null, "loads": true, "max_abs_diff": null, "max_rel_diff": null, "metrics": { "decode_tokens": 256.0, "decode_tps_max": 1.77, "decode_tps_min": 1.73, "init_s": 111.3886, "init_s_max": 111.5353, "init_s_min": 111.3452, "invocations": 3.0, "prefill_tokens": 256.0, "prefill_tps_max": 11.01, "prefill_tps_min": 10.41, "runs_per_invocation": 1.0, "threads": 4.0, "ttft_s_max": 29.8071, "ttft_s_min": 28.1467 }, "output_match": null, "peak_mem_mb": 5511.0, "prefill_tokens_per_s": 10.99, "provenance": "measured", "runs": true, "total_ops": null, "ttft_ms": 28253.3 }, "source": "data/device_runs/0.16.1/2026-09-01/deepseek-r1-distill-qwen-7b-q4-block32-ekv4096__raspberry-pi-5.json" } ] }, "model": { "family": "deepseek-r1-distill-qwen", "id": "deepseek-r1-distill-qwen-7b-q4-block32-ekv4096", "license": "mit", "source_url": "https://huggingface.co/litert-community/DeepSeek-R1-Distill-Qwen-7B", "task": "text-generation" }, "pitfalls": [ "License: MIT for the model weights, inherited from deepseek-ai/DeepSeek-R1-Distill-Qwen-7B; the Qwen2.5 base is Apache-2.0; commercial use and derivatives permitted (HF card License).", "A ~4.2 GB single bundle (4,531,978,224 bytes): desktop / high-memory devices; the Raspberry Pi 5 CPU row (litert-lm 0.16.1, --cache memory) is the only phone-class-or-below measurement in this repo (device runs).", "LiteRT-LM .litertlm bundle: LiteRT.js cannot run it, so delegation stays null and there is no browser block; device rows come from data/device_runs/ (Pi 5 / S26 / Pixel 8a / Mac / iPhone as measured)." ], "schema_version": "1.2" }