| { |
| "status": "VALIDATED", |
| "created_at": "2026-09-21T18:16:33.223816+00:00", |
| "source_repository": "ukisai/Swift-1.5-Qwen3.8-27b", |
| "source_revision": "00ccd14e006897d28cb0ed5bf26390e60d274251", |
| "source_manifest_sha256": "0a00065b88ab003281853a7fb9bd5ce0086bc3781b36136d8c39da19933923ae", |
| "source_manifest_note": "SHA-256 of the original pre-publication-redaction export manifest.", |
| "source_weight_shards": 18, |
| "source_weight_bytes": 55563006776, |
| "source_file_bytes": 55586102681, |
| "source_recovery": "The pinned Hub revision was incomplete at conversion time. All 18 original shards and runtime assets were supplied by the verified project BF16 export and checked against the original export manifest; no weights were reconstructed or borrowed.", |
| "quantization": { |
| "mode": "affine", |
| "bits": 4, |
| "group_size": 64 |
| }, |
| "official_mlx_lm_repository": "https://github.com/ml-explore/mlx-lm", |
| "official_mlx_lm_base_commit": "c69d1288440a0dc4e6401fc417098b07598dccd5", |
| "architecture_patch_sha256": "f6f1d0bdafa45863bfbf93dac0398c481c993ea04fdf38b9bae98c643f89eaec", |
| "packages": { |
| "mlx": "0.32.2", |
| "mlx-cpu": "0.32.2", |
| "mlx-lm": "0.32.0", |
| "transformers": "5.14.1", |
| "huggingface_hub": "1.31.0", |
| "torch": "2.11.0+cpu", |
| "torchvision": "0.26.0+cpu", |
| "safetensors": "0.8.0" |
| }, |
| "python": "3.12.3", |
| "platform": "Linux x86_64", |
| "validation_runtime": { |
| "device": "CPU" |
| }, |
| "conversion_command": [ |
| "mlx_lm.convert", |
| "--hf-path", |
| "<SOURCE_MODEL_DIR>", |
| "--mlx-path", |
| "<OUTPUT_MODEL_DIR>", |
| "--quantize", |
| "--q-mode", |
| "affine", |
| "--q-bits", |
| "4", |
| "--q-group-size", |
| "64" |
| ], |
| "conversion_elapsed_seconds": 120.018799242, |
| "output_weight_shards": 3, |
| "output_weight_bytes": 15826764635, |
| "source_tensors": 1199, |
| "mapped_source_tensors": 1199, |
| "saved_tensors": 2379, |
| "categories": { |
| "text": 851, |
| "MTP": 15, |
| "vision": 333 |
| }, |
| "quantized_weight_tensors": 590, |
| "original_bf16_tensors": 609, |
| "ignored_tensors": 0, |
| "unexplained_tensors": 0, |
| "validation": { |
| "status": "PASS", |
| "recorded_at": "2026-09-21T18:14:21.887850+00:00", |
| "source_repo": "ukisai/Swift-1.5-Qwen3.8-27b", |
| "source_revision": "00ccd14e006897d28cb0ed5bf26390e60d274251", |
| "source_shards": 18, |
| "source_shard_bytes": 55563006776, |
| "source_tensors": 1199, |
| "mapped_source_tensors": 1199, |
| "saved_tensors": 2379, |
| "categories": { |
| "text": 851, |
| "MTP": 15, |
| "vision": 333 |
| }, |
| "ignored_tensors": 0, |
| "unexplained_tensors": 0, |
| "exact_unquantized_tensors": 609, |
| "quantization": { |
| "group_size": 64, |
| "bits": 4, |
| "mode": "affine" |
| }, |
| "all_floating_tensors_finite": true, |
| "tokenizer": "Qwen2Tokenizer", |
| "processor": "Qwen3VLProcessor", |
| "assets_sha256": { |
| "generation_config.json": "e70c136c1b78ddc1fb0905bac8e733a4dc448d4f852a5dd75143fffc70be550e", |
| "preprocessor_config.json": "27225450ac9c6529872ee1924fcb0962ff5634834f817040f444118116f4e516", |
| "video_preprocessor_config.json": "7768af27c1fafa9cc9011c1dc20067e03f8915e03b63504550e11d5066986d13", |
| "tokenizer.json": "0997f410c57a1f4e53b09e4be8f4a172d90edd9564368fb0847030937229b9f3", |
| "tokenizer_config.json": "b11349aafa7cdc6a320767cf7ceb29ed82f7eda5d65e8e0819e76f0ce947bf27", |
| "vocab.json": "ce99b4cb2983d118806ce0a8b777a35b093e2000a503ebde25853284c9dfa003", |
| "merges.txt": "a9d356d7bdf1ef4949e3e748e95b8e10ad9d4e2e838eddc38a0a7b6b94d1db8d", |
| "chat_template.jinja": "c3cf9e34abf4f9e36c2d72165aa9c132d3e2a725b6c2586aaa3a8af9d7a81041" |
| }, |
| "chat_templates": [ |
| { |
| "options": { |
| "enable_thinking": false |
| }, |
| "rendered": "<|im_start|>user\nSay hello.<|im_end|>\n<|im_start|>assistant\n<think>\n\n</think>\n\n" |
| }, |
| { |
| "options": { |
| "reasoning_effort": "low" |
| }, |
| "rendered": "<|im_start|>system\nReasoning effort is set to low. Keep your thinking brief and focused, moving directly to the conclusion without unnecessary elaboration.<|im_end|>\n<|im_start|>user\nSay hello.<|im_end|>\n<|im_start|>assistant\n<think>\n" |
| }, |
| { |
| "options": { |
| "reasoning_effort": "xhigh" |
| }, |
| "rendered": "<|im_start|>system\nReasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.<|im_end|>\n<|im_start|>user\nSay hello.<|im_end|>\n<|im_start|>assistant\n<think>\n" |
| } |
| ], |
| "load_seconds": 3.351265648000208, |
| "load_memory_bytes": 15826466152, |
| "process_peak_rss_bytes": 21142360064, |
| "inference_floating_dtype": "float32, CPU runtime only; stored floating tensors remain BF16", |
| "native_bf16_cpu_inference": "Aborted after reproducing incorrect accumulation in the official Linux BF16 quantized matmul. See cpu-quantized-matmul-diagnostic.json.", |
| "generation": { |
| "prompt": "<|im_start|>user\nReply with exactly: Hello from Swift.<|im_end|>\n<|im_start|>assistant\n<think>\n\n</think>\n\n", |
| "text": "Hello from Swift.", |
| "token_ids": [ |
| 9419, |
| 494, |
| 22929, |
| 13, |
| 248046 |
| ], |
| "tokens": 5, |
| "tokens_per_second": 0.0797672292904007, |
| "prompt_tokens_per_second": 0.06443886297275572, |
| "elapsed_seconds": 374.00736365800003, |
| "finish_reason": "stop" |
| }, |
| "mtp": { |
| "status": "PASS", |
| "shape": [ |
| 1, |
| 1, |
| 248320 |
| ], |
| "path": "Explicit MTP step with real text hidden states and shared LM head; speculative generation is not integrated" |
| }, |
| "vision": { |
| "status": "PASS", |
| "shape": [ |
| 64, |
| 5120 |
| ], |
| "grid": [ |
| [ |
| 1, |
| 16, |
| 16 |
| ] |
| ], |
| "path": "Vision encoder only; image/video insertion and multimodal text generation are not implemented" |
| }, |
| "total_validation_seconds": 467.15810263900016 |
| }, |
| "private_repository": "ukisai/Swift-1.5-4bit-MLX", |
| "apple_silicon_verification": { |
| "status": "PASS_MAC_NATIVE_MLX_FORMAT_AND_REAL_METAL_SAMPLES", |
| "platform": "macOS-26.6-arm64-arm-64bit", |
| "machine": "arm64", |
| "mlx": "0.32.2", |
| "mlx_lm": "0.32.0", |
| "device": "Device(gpu, 0)", |
| "quantization": { |
| "group_size": 64, |
| "bits": 4, |
| "mode": "affine" |
| }, |
| "all_saved_tensor_headers_validated": 2379, |
| "all_source_parameters_accounted_for": 1199, |
| "strict_complete_parameter_tree": "PASS using unevaluated header fixtures; no fabricated weights saved", |
| "actual_checkpoint_samples": [ |
| { |
| "sample": 0, |
| "category": "text", |
| "checkpoint_weight": "language_model.model.layers.0.linear_attn.in_proj_a.weight", |
| "original_shape": [ |
| 48, |
| 5120 |
| ], |
| "packed_shape": [ |
| 48, |
| 640 |
| ], |
| "native_metal_bf16": "PASS", |
| "metal_fp32": "PASS", |
| "bf16_max_absolute_error": 0.0004401206970214844, |
| "fp32_max_absolute_error": 0.0 |
| }, |
| { |
| "sample": 1, |
| "category": "vision", |
| "checkpoint_weight": "visual.blocks.0.attn.proj.weight", |
| "original_shape": [ |
| 1152, |
| 1152 |
| ], |
| "packed_shape": [ |
| 1152, |
| 144 |
| ], |
| "native_metal_bf16": "PASS", |
| "metal_fp32": "PASS", |
| "bf16_max_absolute_error": 0.0002315044403076172, |
| "fp32_max_absolute_error": 0.0 |
| }, |
| { |
| "sample": 2, |
| "category": "MTP", |
| "checkpoint_weight": "mtp.layers.0.self_attn.k_proj.weight", |
| "original_shape": [ |
| 1024, |
| 5120 |
| ], |
| "packed_shape": [ |
| 1024, |
| 640 |
| ], |
| "native_metal_bf16": "PASS", |
| "metal_fp32": "PASS", |
| "bf16_max_absolute_error": 0.0009589195251464844, |
| "fp32_max_absolute_error": 0.0 |
| } |
| ], |
| "full_model_mac_generation": "NOT_RUN: targeted native Metal format and real-weight component checks only", |
| "peak_mlx_bytes": 4597960, |
| "peak_process_rss_bytes": 360693760 |
| } |
| } |
|
|
