{ "artifacts": [ { "file": "Tashkeel-350M-v2_int8.litertlm", "sha256": "e962e6287a3b1e109498d1bca09d20fe55131bb1cd7431165bf255cfd2de1cd7", "size_mb": 458.921 } ], "benchmarks": [], "conversion": { "command": "granite_work/convert_granite4h.py (hf-to-litertlm, 2026-08-25; invocation not stated in the README)", "quantization": "dynamic int8 on linears + embedding (README file table)", "tool": "hf-to-litertlm, family recipe granite_work/convert_granite4h.py (2026-08-25) (README Conversion & verification)", "tool_version": "TODO (the README names the converter but not its version)" }, "cross_runtime": [], "delegation": null, "device": { "records": [ { "device": "galaxy-s26", "run": { "accelerator": "cpu", "context_length": null, "date": "2026-09-05", "decode_tokens_per_s": 68.96, "delegated_ops": 3705, "env": { "device": "Galaxy S26 (SM-S942Q)", "machine_label": "galaxy-s26-cold-cache-cooled", "os_build": "Android 16", "runtime": "litert-lm", "runtime_version": "0.16.0", "soc": "Qualcomm SM8850", "vendor_sdk": null }, "error": null, "evidence": [ "VERBOSE: Replacing 3154 out of 3402 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 304 partitions for subgraph 0 (prefill_1024).", "VERBOSE: Replacing 3154 out of 3402 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 304 partitions for subgraph 1 (prefill_512).", "VERBOSE: Replacing 3126 out of 3374 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 304 partitions for subgraph 2 (prefill_256).", "VERBOSE: Replacing 3264 out of 3512 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 304 partitions for subgraph 3 (prefill_128).", "results block: prefill=212.13 decode=68.96 tokens/s" ], "failure_class": null, "full_delegation": false, "latency_p50_ms": null, "loads": true, "max_abs_diff": null, "max_rel_diff": null, "metrics": { "decode_tokens": 294.0, "init_s": 1.67486, "prefill_tokens": 224.0 }, "output_match": null, "peak_mem_mb": null, "prefill_tokens_per_s": 212.13, "provenance": "measured", "runs": true, "total_ops": 3890, "ttft_ms": 1070.0 }, "source": "data/device_runs/0.16.0/2026-09-05/tashkeel-350m-v2-int8__galaxy-s26.json" }, { "device": "galaxy-s26", "run": { "accelerator": "gpu", "context_length": null, "date": "2026-09-05", "decode_tokens_per_s": 46.26, "delegated_ops": 3512, "env": { "device": "Galaxy S26 (SM-S942Q)", "machine_label": "galaxy-s26-cold-cache-cooled", "os_build": "Android 16", "runtime": "litert-lm", "runtime_version": "0.16.0", "soc": "Qualcomm SM8850", "vendor_sdk": null }, "error": null, "evidence": [ "VERBOSE: Replacing 3402 out of 3402 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 0 (prefill_1024).", "VERBOSE: Replacing 3402 out of 3402 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 1 (prefill_512).", "VERBOSE: Replacing 3374 out of 3374 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 2 (prefill_256).", "VERBOSE: Replacing 3512 out of 3512 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 3 (prefill_128).", "results block: prefill=754.69 decode=46.26 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": 195.0, "init_s": 22.77085, "prefill_tokens": 224.0 }, "output_match": null, "peak_mem_mb": null, "prefill_tokens_per_s": 754.69, "provenance": "measured", "runs": true, "total_ops": 3512, "ttft_ms": 320.0 }, "source": "data/device_runs/0.16.0/2026-09-05/tashkeel-350m-v2-int8__galaxy-s26.json" }, { "device": "mac-studio-m4-max", "run": { "accelerator": "cpu", "context_length": null, "date": "2026-08-25", "decode_tokens_per_s": 97.21, "delegated_ops": null, "env": { "device": "mac-studio-m4-max", "machine_label": "mac-studio-m4-max", "os_build": null, "runtime": "litert-lm", "runtime_version": "0.16.0", "soc": "Apple M4 Max", "vendor_sdk": null }, "error": null, "evidence": [ "results block: prefill=761.25 decode=97.21 tokens/s", "token counts supplied by the caller from recorded evidence (this log states none): prefill=256 decode=256" ], "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, "prefill_tokens": 256.0 }, "output_match": null, "peak_mem_mb": null, "prefill_tokens_per_s": 761.25, "provenance": "measured", "runs": true, "total_ops": null, "ttft_ms": 346.7 }, "source": "data/device_runs/0.16.0/2026-08-25/tashkeel-350m-v2-int8__mac-studio-m4-max.json" }, { "device": "pixel-8a", "run": { "accelerator": "cpu", "context_length": null, "date": "2026-09-05", "decode_tokens_per_s": 49.06, "delegated_ops": 3705, "env": { "device": "Pixel 8a", "machine_label": "pixel-8a-cold-cache-cooled", "os_build": "Android 16", "runtime": "litert-lm", "runtime_version": "0.16.0", "soc": "Tensor G3", "vendor_sdk": null }, "error": null, "evidence": [ "VERBOSE: Replacing 3154 out of 3402 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 304 partitions for subgraph 0 (prefill_1024).", "VERBOSE: Replacing 3154 out of 3402 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 304 partitions for subgraph 1 (prefill_512).", "VERBOSE: Replacing 3126 out of 3374 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 304 partitions for subgraph 2 (prefill_256).", "VERBOSE: Replacing 3264 out of 3512 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 304 partitions for subgraph 3 (prefill_128).", "results block: prefill=111.8 decode=49.06 tokens/s" ], "failure_class": null, "full_delegation": false, "latency_p50_ms": null, "loads": true, "max_abs_diff": null, "max_rel_diff": null, "metrics": { "decode_tokens": 294.0, "init_s": 2.15629, "prefill_tokens": 224.0 }, "output_match": null, "peak_mem_mb": null, "prefill_tokens_per_s": 111.8, "provenance": "measured", "runs": true, "total_ops": 3890, "ttft_ms": 2020.0 }, "source": "data/device_runs/0.16.0/2026-09-05/tashkeel-350m-v2-int8__pixel-8a.json" }, { "device": "pixel-8a", "run": { "accelerator": "gpu", "context_length": null, "date": "2026-09-05", "decode_tokens_per_s": 18.37, "delegated_ops": 3512, "env": { "device": "Pixel 8a", "machine_label": "pixel-8a-cold-cache-cooled", "os_build": "Android 16", "runtime": "litert-lm", "runtime_version": "0.16.0", "soc": "Tensor G3", "vendor_sdk": null }, "error": null, "evidence": [ "VERBOSE: Replacing 3402 out of 3402 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 0 (prefill_1024).", "VERBOSE: Replacing 3402 out of 3402 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 1 (prefill_512).", "VERBOSE: Replacing 3374 out of 3374 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 2 (prefill_256).", "VERBOSE: Replacing 3512 out of 3512 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 3 (prefill_128).", "results block: prefill=250.92 decode=18.37 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": 195.0, "init_s": 25.46553, "prefill_tokens": 224.0 }, "output_match": null, "peak_mem_mb": null, "prefill_tokens_per_s": 250.92, "provenance": "measured", "runs": true, "total_ops": 3512, "ttft_ms": 950.0 }, "source": "data/device_runs/0.16.0/2026-09-05/tashkeel-350m-v2-int8__pixel-8a.json" } ] }, "model": { "family": "granite", "id": "tashkeel-350m-v2-int8", "license": "apache-2.0", "source_url": "https://huggingface.co/mlboydaisuke/Tashkeel-350M-v2-LiteRT", "task": "text-generation" }, "pitfalls": [ "Arabic diacritization (tashkeel) fine-tune of ibm-granite/granite-4.0-h-350m trained on Misraj/Sadeed_Tashkeela; the checkpoint's chat template (byte-equal to the granite base's, 6418/6418) is embedded and applied at runtime (README intro + Conversion).", "The metadata start token is dropped — granite's template has no leading BOS, and at 350M scale a prepended <|end_of_text|> flips correct diacritization into garbage (measured on this checkpoint) (README Conversion).", "Task gate: the model card's worked example plus nine undiacritized MSA probes, bundle vs HF fp32 greedy on identical rendered strings (granite_work/gate_tashkeel.py): fp16 10/10 byte-identical, int8 8/10 (the README carries a note on the two int8 misses) (README file table).", "License Apache-2.0, inherited from the source model and its granite base (README License).", "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/." ], "schema_version": "1.2" }