{
  "backend": "vulkan",
  "target": null,
  "tokenizer": "tokenizer.json",
  "output_repo": "experimentalmachines/Qwen2.5-Math-1.5B-Instruct-ExecuTorch",
  "source": {
    "id": "Qwen/Qwen2.5-Math-1.5B-Instruct",
    "sha": "aafeb0fc6f22cbf0eaeed126eff8be45b0360a35",
    "created_at": "2024-09-16T11:57:49+00:00",
    "total_params": 1543714304,
    "family": "qwen2_5",
    "variant": "instruct",
    "license": {
      "license": "apache-2.0",
      "license_link": "https://huggingface.co/Qwen/Qwen2.5-Math-1.5B-Instruct/blob/main/LICENSE"
    },
    "license_files": [
      "LICENSE"
    ]
  },
  "toolchain": {
    "executorch": "1.5.1",
    "torch": "2.14.0+cpu",
    "torchao": "0.18.0"
  },
  "recipe": {
    "model_class": "qwen2_5_1_5b",
    "converter": "qwen2",
    "params": {
      "dim": 1536,
      "ffn_dim_multiplier": 1,
      "hidden_dim": 8960,
      "n_heads": 12,
      "head_dim": 128,
      "n_kv_heads": 2,
      "n_layers": 28,
      "norm_eps": 1e-06,
      "rope_theta": 10000.0,
      "vocab_size": 151936,
      "use_scaled_rope": false,
      "use_hf_rope": true,
      "attention_qkv_bias": true
    },
    "export_llm_qmode": "8da4w",
    "embedding_hqq": false,
    "qmode": "8da4w-gptq",
    "embedding_quantize": "8,0",
    "prefill_chunk": 2048,
    "kv_cache_dtype": "fp32",
    "group_size": 32,
    "int4_codes": "gptq",
    "label": "8da4w-gptq-g32, int8 embeddings, int4 codes solved with GPTQ",
    "description": "ExecuTorch 1.5.1 `export_llm`: 8-bit dynamic activations and 4-bit weights in groups of 32, int8 per-channel embeddings, the Vulkan delegate, prefill chunk 2048, fp32 KV cache.",
    "codes_sha256": "387ca0fb1d70f9200f0dc1b6be27c392f113489f2eaeefde1e328760f8fc81e8"
  },
  "window": {
    "context": 16384,
    "reason": "requested window 16384",
    "kv_cache_bytes_per_token": 57344,
    "kv_cache_ring_bytes": 0,
    "device_resident_bytes": 2586061875,
    "phone_budget_bytes": 5000000000,
    "fits_phone_budget": true,
    "table": [
      {
        "context": 32768,
        "kv_cache_bytes": 1879048192,
        "device_resident_bytes": 3559475947,
        "export_peak_bytes": 41551591680,
        "fits_device": true,
        "fits_host": true
      },
      {
        "context": 16384,
        "kv_cache_bytes": 939524096,
        "device_resident_bytes": 2586061875,
        "export_peak_bytes": 18063489280,
        "fits_device": true,
        "fits_host": true
      },
      {
        "context": 8192,
        "kv_cache_bytes": 469762048,
        "device_resident_bytes": 2099354839,
        "export_peak_bytes": 11956582656,
        "fits_device": true,
        "fits_host": true
      },
      {
        "context": 4096,
        "kv_cache_bytes": 234881024,
        "device_resident_bytes": 1856001320,
        "export_peak_bytes": 10312415488,
        "fits_device": true,
        "fits_host": true
      },
      {
        "context": 2048,
        "kv_cache_bytes": 117440512,
        "device_resident_bytes": 1734324561,
        "export_peak_bytes": 9842653440,
        "fits_device": true,
        "fits_host": true
      }
    ]
  },
  "files": [
    {
      "path": "vulkan/Qwen2.5-Math-1.5B-Instruct-vulkan-8da4w-gptq-16k.pte",
      "bytes": 1442948096,
      "sha256": "9bc0a2f0437658bbddcf000b5100c786c156154411c7b935ee0f89d9d7b9d981"
    }
  ],
  "estimates": {
    "pte_bytes_estimate": 1146537779,
    "pte_bytes_actual": 1442948096,
    "export_peak_bytes_estimate": 18063489280
  },
  "host": {
    "nproc": 32,
    "mem_total_bytes": 265985806336,
    "swap_total_bytes": 0,
    "cpu_model": "Intel(R) Xeon(R) Platinum 8488C",
    "peak_rss_convert_bytes": 4022853632,
    "peak_rss_export_bytes": 24027811840,
    "peak_swap_used_bytes": 0,
    "min_mem_available_bytes": 221771304960,
    "peak_in_use_bytes": 44214501376,
    "peak_in_use_at": "2026-11-09T07:09:58Z",
    "peak_cgroup_bytes": null,
    "export_seconds": 195.2,
    "total_seconds": 252.4
  },
  "conversion": {
    "tensors": 339,
    "tied_embeddings": true,
    "dtypes": [
      "torch.bfloat16"
    ]
  },
  "metadata": {
    "enable_dynamic_shape": true,
    "get_bos_id": 151643,
    "get_eos_ids": [
      151645,
      151643
    ],
    "get_max_context_len": 16384,
    "get_max_seq_len": 2048,
    "get_n_layers": 28,
    "get_vocab_size": 151936,
    "use_kv_cache": true,
    "use_sdpa_with_kv_cache": true,
    "methods": [
      "enable_dynamic_shape",
      "forward",
      "get_bos_id",
      "get_eos_ids",
      "get_max_context_len",
      "get_max_seq_len",
      "get_n_layers",
      "get_vocab_size",
      "use_kv_cache",
      "use_sdpa_with_kv_cache"
    ]
  },
  "smoke": {
    "kind": "structural",
    "passed": true,
    "problems": [],
    "methods": [
      "enable_dynamic_shape",
      "forward",
      "get_bos_id",
      "get_eos_ids",
      "get_max_context_len",
      "get_max_seq_len",
      "get_n_layers",
      "get_vocab_size",
      "use_kv_cache",
      "use_sdpa_with_kv_cache"
    ],
    "delegates": {
      "forward": {
        "backends": [
          "VulkanBackend"
        ],
        "delegate_calls": 1
      }
    },
    "metadata": {
      "enable_dynamic_shape": true,
      "get_bos_id": 151643,
      "get_eos_ids": [
        151645,
        151643
      ],
      "get_max_context_len": 16384,
      "get_max_seq_len": 2048,
      "get_n_layers": 28,
      "get_vocab_size": 151936,
      "use_kv_cache": true,
      "use_sdpa_with_kv_cache": true,
      "methods": [
        "enable_dynamic_shape",
        "forward",
        "get_bos_id",
        "get_eos_ids",
        "get_max_context_len",
        "get_max_seq_len",
        "get_n_layers",
        "get_vocab_size",
        "use_kv_cache",
        "use_sdpa_with_kv_cache"
      ]
    }
  },
  "gate_proxy": {
    "report": "export-report-16k.json",
    "file": "xnnpack/Qwen2.5-Math-1.5B-Instruct-8da4w-gptq-16k.pte",
    "sha256": "38a3d150f0d4215d223057ab997c3d82b929ff067125a5a444f56bd9344277ad",
    "gate": {
      "passed": true,
      "graded": 3,
      "agreed": 3,
      "fp32_mean": 0.7674,
      "export_mean": 0.7408,
      "override": null
    }
  },
  "run": {}
}