random_models / deepseek-coder /gguf /metadata.json
shibatch's picture
Upload folder using huggingface_hub
ce7ccba verified
Raw
History Blame Contribute Delete
2.74 kB
{
"commands": {
"convert": [
"PYTHON",
"LLAMA_CPP/convert_hf_to_gguf.py",
"STAGING",
"--outfile",
"INTERMEDIATE",
"--outtype",
"f16"
],
"quantize": [
"LLAMA_QUANTIZE",
"--output-tensor-type",
"f16",
"--token-embedding-type",
"f16",
"INTERMEDIATE",
"OUTPUT",
"Q4_K_M"
]
},
"comparison_to_f32_reference": {
"decode": {
"max_abs_error": 6.486475467681885e-05,
"mean_abs_error": 2.3104150386643596e-05,
"top1_agreement": true,
"top5_overlap": 5
},
"prefill": {
"max_abs_error": 7.614446803927422e-05,
"mean_abs_error": 2.0526116713881493e-05,
"top1_agreement": true,
"top5_overlap": 5
}
},
"converter": "convert_hf_to_gguf.py",
"converter_log_sha256": "5029f6e7c0af519f829e53b9db930b09f62ea2444d3d267ca839d7202ed6c3eb",
"direct_token_validation": {
"all_logits_finite": true,
"decode_token_id": 13,
"exactly_reproducible": true,
"prefill_token_ids": [
3,
5,
7,
11
]
},
"effective_eog_token_ids": [
1
],
"eog_validation_range": [
0,
127
],
"format": "GGUF",
"intermediate_f16_sha256": "58fe175f8292f78728181fe74107e7bdc107748900cc5e3e7cc9e628daaad4ea",
"intermediate_retained": false,
"llama_cpp_revision": "40b740ad05c531b9d57aca6698c3ed553a9e784c",
"model_sha256": "8557d99db9de6b588628f8fead576ff85b26edbdc8df176050b59a28acf720f3",
"model_size_bytes": 22784,
"notes": "Q4_K_M is lossy; numerical differences are recorded as observations, not pass thresholds.",
"quantization_requested": "Q4_K_M",
"quantizer": "llama-quantize",
"quantizer_fallback_observed": true,
"quantizer_log_sha256": "91680988560477d86ec6ecc8b7e3b7b6fd21236183dd3c918afb9f4c69e07f85",
"schema_version": 1,
"source_auxiliary_tokenizer_sha256": "0f88117a2f8c52a9eabe2d61f7301b6f9b58b7a56644888da4ec9970b1efbcee",
"source_package_model_sha256": "cba3bb2298077f3bdeee9da78a4d0ae7586c3a3964e7a40eb2f9e5dc2631c4cc",
"staging_special_token_normalization": {
"applied": false,
"gguf_primary_eos_token_id": 1,
"reason": "GGUF exposes one primary EOS metadata field; all source EOS IDs remain documented by tokenizer metadata.",
"source_eos_token_id": 1
},
"tensor_type_histogram": {
"F16": 14,
"F32": 5,
"Q5_0": 1,
"Q8_0": 1
},
"tiny_dimension_tensor_overrides": {
"output.weight": "F16",
"reason": "These tensors can be too narrow for the Q8_0 types selected by the Q4_K_M mixture.",
"token_embd.weight": "F16"
},
"tokenizer_smoke": {
"hf_native_ids_equal": true,
"text": "a",
"token_ids": [
126,
124
]
}
}