@simulatte/doppler 0.1.9 → 0.2.0

package/src/inference/pipelines/text/config.js

@@ -1,731 +0,0 @@
-import { log } from '../../../debug/index.js';
-import { mergeConfig } from '../../../config/merge.js';
-import { selectRuleValue } from '../../../rules/rule-registry.js';
-
-const UNSUPPORTED_RUNTIME_MODEL_TYPES = new Set(['mamba', 'rwkv']);
-
-// =============================================================================
-// Model Detection Functions
-// =============================================================================
-
-function assertSupportedRuntimeModelType(manifest) {
-  const modelType = typeof manifest?.modelType === 'string'
-    ? manifest.modelType.trim().toLowerCase()
-    : '';
-  if (!modelType) return;
-  if (!UNSUPPORTED_RUNTIME_MODEL_TYPES.has(modelType)) return;
-
-  const modelId = manifest?.modelId ?? 'unknown';
-  throw new Error(
-    `Manifest "${modelId}" declares modelType "${modelType}", but that runtime family is not implemented yet.`
-  );
-}
-
-function resolveRotaryDim(headDim, partialRotaryFactor, modelId) {
-  if (partialRotaryFactor == null) {
-    return headDim;
-  }
-  if (typeof partialRotaryFactor !== 'number' || Number.isNaN(partialRotaryFactor)) {
-    throw new Error(`Manifest "${modelId}" has invalid rope.partialRotaryFactor.`);
-  }
-  if (partialRotaryFactor <= 0 || partialRotaryFactor > 1) {
-    throw new Error(
-      `Manifest "${modelId}" requires 0 < rope.partialRotaryFactor <= 1; got ${partialRotaryFactor}.`
-    );
-  }
-  const rotaryDim = Math.trunc(headDim * partialRotaryFactor);
-  if (rotaryDim <= 0 || (rotaryDim % 2) !== 0) {
-    throw new Error(
-      `Manifest "${modelId}" resolves rope rotary dim ${rotaryDim} from headDim=${headDim} ` +
-      `and partialRotaryFactor=${partialRotaryFactor}, but rotary dim must be a positive even integer.`
-    );
-  }
-  return rotaryDim;
-}
-
-export function getStopTokenIds(manifest) {
-  const eosTokenId = manifest?.eos_token_id;
-  if (Array.isArray(eosTokenId)) return eosTokenId;
-  if (typeof eosTokenId === 'number') return [eosTokenId];
-  const modelId = manifest?.modelId ?? 'unknown';
-  throw new Error(
-    `Manifest "${modelId}" is missing eos_token_id. Re-convert the model with tokenizer metadata.`
-  );
-}
-
-function normalizeFfnTensorShape(value) {
-  if (!Array.isArray(value) || value.length !== 2) return null;
-  const rows = Number(value[0]);
-  const cols = Number(value[1]);
-  if (!Number.isFinite(rows) || !Number.isFinite(cols)) return null;
-  if (rows <= 0 || cols <= 0) return null;
-  return [Math.trunc(rows), Math.trunc(cols)];
-}
-
-function isExpertTensorName(name) {
-  const lower = String(name || '').toLowerCase();
-  return lower.includes('.experts.') || lower.includes('.expert.') || lower.includes('block_sparse_moe');
-}
-
-function inferLfm2IntermediateSizeFromManifest(manifest) {
-  const tensors = manifest?.tensors;
-  if (!tensors || typeof tensors !== 'object') return null;
-  const candidates = [];
-  for (const [name, entry] of Object.entries(tensors)) {
-    if (!name || isExpertTensorName(name)) continue;
-    const shape = normalizeFfnTensorShape(entry?.shape);
-    if (!shape) continue;
-    const lower = name.toLowerCase();
-    if (
-      lower.endsWith('.feed_forward.w1.weight')
-      || lower.endsWith('.feed_forward.w3.weight')
-      || lower.endsWith('.ffn_gate.weight')
-      || lower.endsWith('.ffn_up.weight')
-      || lower.endsWith('.ffn.gate_proj.weight')
-      || lower.endsWith('.ffn.up_proj.weight')
-      || lower.endsWith('.mlp.gate_proj.weight')
-      || lower.endsWith('.mlp.up_proj.weight')
-    ) {
-      candidates.push(shape[0]);
-      continue;
-    }
-    if (
-      lower.endsWith('.feed_forward.w2.weight')
-      || lower.endsWith('.ffn_down.weight')
-      || lower.endsWith('.ffn.down_proj.weight')
-      || lower.endsWith('.mlp.down_proj.weight')
-    ) {
-      candidates.push(shape[1]);
-      continue;
-    }
-    if (
-      lower.endsWith('.feed_forward.w1_w3.weight')
-      || lower.endsWith('.ffn_gate_up.weight')
-      || lower.endsWith('.ffn.gate_up_proj.weight')
-      || lower.endsWith('.mlp.gate_up_proj.weight')
-    ) {
-      if (shape[0] % 2 === 0) {
-        candidates.push(Math.trunc(shape[0] / 2));
-      }
-    }
-  }
-  if (candidates.length === 0) return null;
-  const counts = new Map();
-  for (const value of candidates) {
-    counts.set(value, (counts.get(value) ?? 0) + 1);
-  }
-  return [...counts.entries()]
-    .sort((a, b) => {
-      if (b[1] !== a[1]) return b[1] - a[1];
-      return a[0] - b[0];
-    })[0]?.[0] ?? null;
-}
-
-function resolveIntermediateSizeForRuntime(manifest, inf, arch, modelId) {
-  const fromArch = arch?.intermediateSize;
-  if (typeof fromArch !== 'number' || !Number.isFinite(fromArch) || fromArch <= 0) {
-    return fromArch;
-  }
-  const presetId = String(manifest?.inference?.presetId ?? inf?.presetId ?? '').toLowerCase();
-  if (presetId !== 'lfm2') {
-    return fromArch;
-  }
-  const inferred = inferLfm2IntermediateSizeFromManifest(manifest);
-  if (inferred == null || inferred === fromArch) {
-    return fromArch;
-  }
-  throw new Error(
-    `Manifest "${modelId}" has intermediateSize=${fromArch}, but FFN tensors imply ${inferred}. ` +
-    'Re-convert the model so manifest architecture matches the weights.'
-  );
-}
-
-// =============================================================================
-// Manifest-First Config Resolution (NEW)
-// =============================================================================
-
-
-export function hasManifestInference(manifest) {
-  return 'inference' in manifest && manifest.inference != null;
-}
-
-
-export function validateRequiredInferenceFields(inf, modelId) {
-  inf = inf ?? {};
-  inf.attention = inf.attention ?? {};
-  inf.normalization = inf.normalization ?? {};
-  inf.ffn = inf.ffn ?? {};
-  inf.rope = inf.rope ?? {};
-  inf.output = inf.output ?? {};
-  inf.layerPattern = inf.layerPattern ?? {};
-  inf.chatTemplate = inf.chatTemplate ?? {};
-  const errors = [];
-
-  // Attention fields - non-nullable required
-  if (inf.attention.queryPreAttnScalar == null) {
-    errors.push('attention.queryPreAttnScalar is required');
-  }
-  if (inf.attention.queryKeyNorm == null) {
-    errors.push('attention.queryKeyNorm is required');
-  }
-  if (inf.attention.attentionBias == null) {
-    errors.push('attention.attentionBias is required');
-  }
-  if (inf.attention.causal == null) {
-    errors.push('attention.causal is required');
-  }
-  // Attention fields - nullable required (undefined = missing, null = disabled)
-  if (inf.attention.slidingWindow === undefined) {
-    errors.push('attention.slidingWindow must be explicitly set (null for no sliding window, or number)');
-  }
-  if (inf.attention.attnLogitSoftcapping === undefined) {
-    errors.push('attention.attnLogitSoftcapping must be explicitly set (null for no softcapping, or number)');
-  }
-
-  // Normalization fields
-  if (inf.normalization.rmsNormWeightOffset == null) {
-    errors.push('normalization.rmsNormWeightOffset is required');
-  }
-  if (inf.normalization.rmsNormEps == null) {
-    errors.push('normalization.rmsNormEps is required');
-  }
-  if (inf.normalization.postAttentionNorm == null) {
-    errors.push('normalization.postAttentionNorm is required');
-  }
-  if (inf.normalization.preFeedforwardNorm == null) {
-    errors.push('normalization.preFeedforwardNorm is required');
-  }
-  if (inf.normalization.postFeedforwardNorm == null) {
-    errors.push('normalization.postFeedforwardNorm is required');
-  }
-
-  // FFN fields
-  if (inf.ffn.activation == null) {
-    errors.push('ffn.activation is required');
-  }
-  if (inf.ffn.gatedActivation == null) {
-    errors.push('ffn.gatedActivation is required');
-  }
-  if (inf.ffn.swigluLimit === undefined) {
-    errors.push('ffn.swigluLimit must be explicitly set (null for no limit, or number)');
-  } else {
-    const limit = inf.ffn.swigluLimit;
-    if (limit !== null && (typeof limit !== 'number' || Number.isNaN(limit) || limit <= 0)) {
-      errors.push('ffn.swigluLimit must be a positive number or null');
-    }
-  }
-
-  // RoPE fields - non-nullable required
-  if (inf.rope.ropeTheta == null) {
-    errors.push('rope.ropeTheta is required');
-  }
-  if (inf.rope.ropeScalingFactor == null) {
-    errors.push('rope.ropeScalingFactor is required (use 1.0 for no scaling)');
-  }
-  // RoPE fields - nullable required (undefined = missing, null = disabled)
-  if (inf.rope.ropeScalingType === undefined) {
-    errors.push('rope.ropeScalingType must be explicitly set (null for no scaling, or scaling type string)');
-  }
-  if (inf.rope.ropeLocalTheta === undefined) {
-    errors.push('rope.ropeLocalTheta must be explicitly set (null for no local theta, or number)');
-  }
-  if (inf.rope.mropeInterleaved == null) {
-    errors.push('rope.mropeInterleaved is required');
-  }
-  if (inf.rope.mropeSection === undefined) {
-    errors.push('rope.mropeSection must be explicitly set (null when unused, or an array of positive integers)');
-  }
-  if (inf.rope.partialRotaryFactor === undefined) {
-    errors.push('rope.partialRotaryFactor must be explicitly set (null when unused, or a number in (0, 1])');
-  } else {
-    const factor = inf.rope.partialRotaryFactor;
-    if (factor !== null && (typeof factor !== 'number' || Number.isNaN(factor) || factor <= 0 || factor > 1)) {
-      errors.push('rope.partialRotaryFactor must be a number in (0, 1] or null');
-    }
-  }
-
-  // Output fields - non-nullable required
-  if (inf.output.tieWordEmbeddings == null) {
-    errors.push('output.tieWordEmbeddings is required');
-  }
-  if (inf.output.scaleEmbeddings == null) {
-    errors.push('output.scaleEmbeddings is required');
-  }
-  if (inf.output.embeddingTranspose == null) {
-    errors.push('output.embeddingTranspose is required');
-  }
-  // Output fields - nullable required (undefined = missing, null = disabled)
-  if (inf.output.finalLogitSoftcapping === undefined) {
-    errors.push('output.finalLogitSoftcapping must be explicitly set (null for no softcapping, or number)');
-  }
-  if (inf.output.embeddingVocabSize === undefined) {
-    errors.push('output.embeddingVocabSize must be explicitly set (null to use architecture.vocabSize, or number)');
-  }
-
-  // Layer pattern fields
-  if (inf.layerPattern?.type == null) {
-    errors.push('layerPattern.type is required');
-  }
-  if (inf.layerPattern?.globalPattern === undefined) {
-    errors.push('layerPattern.globalPattern must be explicitly set (null if not applicable)');
-  }
-  if (inf.layerPattern?.period === undefined) {
-    errors.push('layerPattern.period must be explicitly set (null if not applicable)');
-  }
-  if (inf.layerPattern?.offset === undefined) {
-    errors.push('layerPattern.offset must be explicitly set (null if not applicable)');
-  }
-  if (inf.layerPattern?.type === 'custom' && inf.layerPattern?.layerTypes === undefined) {
-    errors.push('layerPattern.layerTypes must be explicitly set for custom patterns');
-  }
-
-  // Chat template fields
-  if (inf.chatTemplate?.type === undefined) {
-    errors.push('chatTemplate.type must be explicitly set (null for no template)');
-  }
-  if (inf.chatTemplate?.enabled == null) {
-    errors.push('chatTemplate.enabled is required');
-  }
-
-  // RoPE YARN fields
-  if (inf.rope.yarnBetaFast === undefined) {
-    errors.push('rope.yarnBetaFast must be explicitly set (null if not YARN)');
-  }
-  if (inf.rope.yarnBetaSlow === undefined) {
-    errors.push('rope.yarnBetaSlow must be explicitly set (null if not YARN)');
-  }
-  if (inf.rope.yarnOriginalMaxPos === undefined) {
-    errors.push('rope.yarnOriginalMaxPos must be explicitly set (null if not YARN)');
-  }
-
-  if (errors.length > 0) {
-    throw new Error(
-      `Manifest "${modelId}" has incomplete inference config. ` +
-      `Missing required fields:\n  - ${errors.join('\n  - ')}\n` +
-      `Re-convert the model using the latest converter.`
-    );
-  }
-}
-
-function normalizeLayerTypeTag(value) {
-  const normalized = typeof value === 'string' ? value.trim().toLowerCase() : '';
-  if (!normalized) return null;
-  if (
-    normalized === 'full_attention'
-    || normalized === 'global_attention'
-    || normalized === 'full'
-    || normalized === 'global'
-    || normalized === 'attention'
-  ) {
-    return 'full_attention';
-  }
-  if (
-    normalized === 'sliding_attention'
-    || normalized === 'local_attention'
-    || normalized === 'local'
-    || normalized === 'sliding'
-  ) {
-    return 'sliding_attention';
-  }
-  if (
-    normalized === 'linear_attention'
-    || normalized === 'linear'
-    || normalized === 'gated_delta'
-    || normalized === 'gated_delta_net'
-  ) {
-    return 'linear_attention';
-  }
-  if (
-    normalized === 'conv'
-    || normalized === 'convolution'
-    || normalized === 'liv_conv'
-    || normalized === 'liv_convolution'
-  ) {
-    return 'conv';
-  }
-  if (normalized === 'moe' || normalized === 'mamba' || normalized === 'rwkv') {
-    return normalized;
-  }
-  return null;
-}
-
-function resolveVisionConfig(rawConfig, manifest) {
-  const vc = rawConfig?.vision_config ?? manifest?.config?.vision_config;
-  if (!vc || typeof vc !== 'object') return null;
-  return {
-    depth: vc.depth ?? 24,
-    hiddenSize: vc.hidden_size ?? 1024,
-    intermediateSize: vc.intermediate_size ?? 4096,
-    numHeads: vc.num_heads ?? 16,
-    outHiddenSize: vc.out_hidden_size ?? vc.hidden_size ?? 1024,
-    patchSize: vc.patch_size ?? 16,
-    spatialMergeSize: vc.spatial_merge_size ?? 2,
-    temporalPatchSize: vc.temporal_patch_size ?? 2,
-    eps: vc.eps ?? 1e-6,
-    deepstackVisualIndexes: Array.isArray(vc.deepstack_visual_indexes) ? vc.deepstack_visual_indexes : [],
-    imageTokenId: rawConfig?.image_token_id ?? manifest?.image_token_id ?? null,
-  };
-}
-
-function parseCustomLayerTypes(layerTypes, numLayers, modelId) {
-  if (!Array.isArray(layerTypes) || layerTypes.length === 0) {
-    throw new Error(
-      `Manifest "${modelId}" has layerPattern.type='custom' but layerPattern.layerTypes is missing or empty. ` +
-      'Re-convert the model to include explicit layer types.'
-    );
-  }
-  if (layerTypes.length !== numLayers) {
-    throw new Error(
-      `Manifest "${modelId}" has layerPattern.type='custom' with ${layerTypes.length} layer types, ` +
-      `expected ${numLayers}. Re-convert the model to preserve full per-layer metadata.`
-    );
-  }
-  return layerTypes.map((layerType, index) => {
-    const normalized = normalizeLayerTypeTag(layerType);
-    if (!normalized) {
-      throw new Error(
-        `Manifest "${modelId}" has unknown layerPattern.layerTypes[${index}]="${layerType}". ` +
-        'Supported types: conv, full_attention, sliding_attention, linear_attention, moe, mamba, rwkv.'
-      );
-    }
-    return normalized;
-  });
-}
-
-function parseLinearNormMode(value, sharedFlag = null, modelId = 'unknown') {
-  if (typeof value === 'string') {
-    const normalized = value.trim().toLowerCase();
-    if (normalized === 'shared') return 'shared';
-    if (normalized === 'per_head' || normalized === 'per-head' || normalized === 'perhead') {
-      return 'per_head';
-    }
-    throw new Error(
-      `Manifest "${modelId}" has unsupported linear_norm_mode="${value}". ` +
-      'Supported values: "shared", "per_head".'
-    );
-  }
-  if (typeof sharedFlag === 'boolean') {
-    return sharedFlag ? 'shared' : 'per_head';
-  }
-  return null;
-}
-
-
-export function toParsedConfigFromMerged(merged, manifest) {
-  const rawConfig = manifest.config ?? {};
-  const config = rawConfig.text_config ?? rawConfig;
-  const inf = merged.inference;
-
-  // Validate required fields are present (fail fast on incomplete manifests)
-  validateRequiredInferenceFields(inf, merged.modelId);
-  if (manifest.quantization == null) {
-    throw new Error(`Manifest "${merged.modelId}" is missing quantization.`);
-  }
-
-  // Get architecture dimensions
-  const arch = (manifest.architecture && typeof manifest.architecture === 'object')
-    ? manifest.architecture
-    : null;
-  if (!arch) {
-    throw new Error(
-      `Manifest "${merged.modelId}" is missing architecture config. ` +
-      `Re-convert the model using the latest converter to add manifest.architecture.`
-    );
-  }
-  const resolvedIntermediateSize = resolveIntermediateSizeForRuntime(manifest, inf, arch, merged.modelId);
-
-  // Compute layer types from layerPattern
-  
-  let layerTypes = null;
-  if (inf.layerPattern) {
-    const numLayers = arch.numLayers;
-    const patternType = inf.layerPattern.type;
-
-    if (patternType === 'custom') {
-      layerTypes = parseCustomLayerTypes(inf.layerPattern.layerTypes, numLayers, merged.modelId);
-    } else {
-      // Fail fast if alternating pattern lacks required globalPattern
-      if (patternType === 'alternating' && inf.layerPattern.globalPattern == null) {
-        throw new Error(
-          `Manifest "${merged.modelId}" has layerPattern.type='alternating' but globalPattern is missing. ` +
-          `Re-convert the model to include layerPattern.globalPattern.`
-        );
-      }
-
-      // Fail fast if every_n pattern lacks required period
-      if (patternType === 'every_n' && inf.layerPattern.period == null) {
-        throw new Error(
-          `Manifest "${merged.modelId}" has layerPattern.type='every_n' but period is missing. ` +
-          `Re-convert the model to include layerPattern.period.`
-        );
-      }
-      const period = inf.layerPattern.period;
-      const rawOffset = inf.layerPattern.offset;
-      const offset = (
-        Number.isFinite(rawOffset) && period != null && period > 0
-      )
-        ? ((Math.trunc(rawOffset) % period) + period) % period
-        : 0;
-      const pattern = inf.layerPattern.globalPattern;
-      const patternKind = selectRuleValue(
-        'inference',
-        'layerPattern',
-        'patternKind',
-        { patternType, globalPattern: pattern }
-      );
-      if (patternKind) {
-        layerTypes = Array.from({ length: numLayers }, (_, i) => {
-          const isEven = i % 2 === 0;
-          // For every_n pattern: global at layer "offset" and every N thereafter.
-          // e.g. period=6, offset=5 => indices 5,11,17,...
-          const isStride = period == null ? false : (((i - offset) % period + period) % period) === 0;
-          return selectRuleValue(
-            'inference',
-            'layerPattern',
-            'layerType',
-            { patternKind, isEven, isStride }
-          );
-        });
-      }
-    }
-  }
-
-  if (!Array.isArray(layerTypes) && Array.isArray(config.layer_types) && config.layer_types.length > 0) {
-    layerTypes = parseCustomLayerTypes(config.layer_types, arch.numLayers, merged.modelId);
-  }
-
-  // Compute queryPreAttnScalar from manifest inference (NOT from preset detection)
-  // Manifest-first: queryPreAttnScalar is required in ManifestAttentionSchema
-  const headDim = arch.headDim;
-  const queryPreAttnScalar = inf.attention.queryPreAttnScalar;
-  const causalAttention = inf.attention.causal;
-
-  // Cross-field sanity: queryPreAttnScalar should typically equal headDim.
-  // A value of sqrt(headDim) indicates a known converter bug that produces
-  // attnScale = 1/sqrt(sqrt(headDim)) instead of the correct 1/sqrt(headDim).
-  if (queryPreAttnScalar != null && headDim != null
-      && queryPreAttnScalar !== headDim
-      && Math.abs(queryPreAttnScalar - Math.sqrt(headDim)) < 0.01) {
-    throw new Error(
-      `Model "${merged.modelId}": queryPreAttnScalar (${queryPreAttnScalar}) ` +
-      `equals sqrt(headDim) instead of headDim (${headDim}). ` +
-      `This is a known converter bug — the manifest must be regenerated ` +
-      `with the corrected converter.`
-    );
-  }
-
-  // Get stop token IDs (cast to Manifest for compatibility)
-  const stopTokenIds = getStopTokenIds(manifest);
-
-  // Get MoE config
-  const moeConfig = manifest.moeConfig ?? null;
-  const useMoE = (moeConfig?.numExperts ?? 0) > 1;
-  if (useMoE && (moeConfig?.numExperts == null || moeConfig?.numExpertsPerToken == null || !moeConfig?.expertFormat)) {
-    throw new Error(`Manifest "${manifest.modelId}" is missing moeConfig fields for MoE inference.`);
-  }
-  const numExperts = useMoE ? moeConfig.numExperts : 0;
-  const moeTopK = useMoE ? moeConfig.numExpertsPerToken : 0;
-  const expertFormat = useMoE ? moeConfig.expertFormat : null;
-
-  // RoPE scaling - use manifest inference as source of truth (not raw config)
-  const ropeScale = inf.rope.ropeScalingFactor;
-  const ropeScalingType = inf.rope.ropeScalingType;
-  const ropeLocalScale = inf.rope.ropeLocalScalingFactor;
-  const ropeLocalScalingType = inf.rope.ropeLocalScalingType;
-  const partialRotaryFactor = inf.rope.partialRotaryFactor;
-  const mropeInterleaved = inf.rope.mropeInterleaved === true;
-  const ropeInterleaved = false;
-
-  if (ropeLocalScale == null && (inf.rope.ropeLocalTheta != null || inf.rope.mropeSection != null)) {
-    throw new Error(
-      `Model "${merged.modelId}" uses hybrid/mRoPE but is missing rope.ropeLocalScalingFactor in manifest. ` +
-      `Re-convert the model using the latest converter or update the manifest to include an explicit scale.`
-    );
-  }
-  const mropeSection = Array.isArray(inf.rope.mropeSection)
-    ? inf.rope.mropeSection.map((entry) => Math.trunc(Number(entry)))
-    : null;
-  const ropeRotaryDim = resolveRotaryDim(arch.headDim, partialRotaryFactor, merged.modelId);
-  if (mropeSection && mropeSection.some((entry) => !Number.isFinite(entry) || entry <= 0)) {
-    throw new Error(
-      `Manifest "${merged.modelId}" has invalid rope.mropeSection; expected positive integers.`
-    );
-  }
-  if (mropeInterleaved && mropeSection) {
-    const doubledMropeDim = mropeSection.reduce((sum, entry) => sum + entry, 0) * 2;
-    if (doubledMropeDim !== ropeRotaryDim) {
-      throw new Error(
-        `Manifest "${merged.modelId}" declares rope.mropeSection=${JSON.stringify(mropeSection)}, ` +
-        `which expands to rotary dim ${doubledMropeDim}, but the resolved rotary dim is ${ropeRotaryDim}.`
-      );
-    }
-  }
-
-  // Build ropeScaling object from manifest values if scaling is enabled
-  // Include YARN params when present
-  
-  const ropeScaling = ropeScalingType ? {
-    type: ropeScalingType,
-    factor: ropeScale,
-    ...(ropeScalingType === 'yarn' && inf.rope.yarnBetaFast != null && { beta_fast: inf.rope.yarnBetaFast }),
-    ...(ropeScalingType === 'yarn' && inf.rope.yarnBetaSlow != null && { beta_slow: inf.rope.yarnBetaSlow }),
-    ...(ropeScalingType === 'yarn' && inf.rope.yarnOriginalMaxPos != null && {
-      original_max_position_embeddings: inf.rope.yarnOriginalMaxPos
-    }),
-  } : null;
-  const ropeLocalScaling = ropeLocalScalingType ? {
-    type: ropeLocalScalingType,
-    factor: ropeLocalScale,
-    ...(ropeLocalScalingType === 'yarn' && (inf.rope.ropeLocalYarnBetaFast ?? inf.rope.yarnBetaFast) != null && {
-      beta_fast: inf.rope.ropeLocalYarnBetaFast ?? inf.rope.yarnBetaFast
-    }),
-    ...(ropeLocalScalingType === 'yarn' && (inf.rope.ropeLocalYarnBetaSlow ?? inf.rope.yarnBetaSlow) != null && {
-      beta_slow: inf.rope.ropeLocalYarnBetaSlow ?? inf.rope.yarnBetaSlow
-    }),
-    ...(ropeLocalScalingType === 'yarn'
-      && (inf.rope.ropeLocalYarnOriginalMaxPos ?? inf.rope.yarnOriginalMaxPos) != null && {
-      original_max_position_embeddings:
-        inf.rope.ropeLocalYarnOriginalMaxPos ?? inf.rope.yarnOriginalMaxPos
-    }),
-  } : null;
-
-  // Activation type
-  const activation = inf.ffn.activation;
-  
-  const hiddenActivation = selectRuleValue(
-    'inference',
-    'config',
-    'hiddenActivation',
-    { activation }
-  );
-
-  const chatTemplateType = inf.chatTemplate.type;
-  const chatTemplateEnabled = inf.chatTemplate.enabled;
-  const parsePositiveInt = (value) => {
-    const num = Number(value);
-    if (!Number.isFinite(num) || num <= 0) return null;
-    return Math.trunc(num);
-  };
-
-  const linearNumKeyHeads = parsePositiveInt(arch.linearNumKeyHeads ?? config.linear_num_key_heads);
-  const linearNumValueHeads = parsePositiveInt(arch.linearNumValueHeads ?? config.linear_num_value_heads);
-  const linearKeyHeadDim = parsePositiveInt(arch.linearKeyHeadDim ?? config.linear_key_head_dim);
-  const linearValueHeadDim = parsePositiveInt(arch.linearValueHeadDim ?? config.linear_value_head_dim);
-  const linearConvKernelDim = parsePositiveInt(arch.linearConvKernelDim ?? config.linear_conv_kernel_dim);
-  const linearNormMode = parseLinearNormMode(
-    arch.linearNormMode ?? config.linear_norm_mode,
-    config.linear_norm_shared,
-    merged.modelId
-  );
-
-  return {
-    numLayers: arch.numLayers,
-    hiddenSize: arch.hiddenSize,
-    intermediateSize: resolvedIntermediateSize,
-    numHeads: arch.numAttentionHeads,
-    numKVHeads: arch.numKeyValueHeads,
-    headDim: arch.headDim,
-    vocabSize: arch.vocabSize,
-    maxSeqLen: arch.maxSeqLen,
-    useMoE,
-    numExperts,
-    moeTopK,
-    expertFormat,
-    slidingWindow: inf.attention.slidingWindow,
-    ropeTheta: inf.rope.ropeTheta,
-    ropeLocalTheta: inf.rope.ropeLocalTheta,
-    ropeRotaryDim,
-    ropeInterleaved,
-    mropeInterleaved,
-    mropeSection,
-    partialRotaryFactor,
-    ropeScale,
-    ropeLocalScale,
-    ropeScalingType,
-    ropeLocalScalingType,
-    ropeScaling,
-    ropeLocalScaling,
-    quantization: manifest.quantization,
-    quantMethod: config.quantization_config?.quant_method ?? null,
-    rmsNormEps: inf.normalization.rmsNormEps,
-    rmsNormWeightOffset: inf.normalization.rmsNormWeightOffset,
-    postAttentionNorm: inf.normalization.postAttentionNorm,
-    preFeedforwardNorm: inf.normalization.preFeedforwardNorm,
-    postFeedforwardNorm: inf.normalization.postFeedforwardNorm,
-    scaleEmbeddings: inf.output.scaleEmbeddings,
-    useTiedEmbeddings: inf.output.tieWordEmbeddings,
-    embeddingTranspose: inf.output.embeddingTranspose,
-    embeddingVocabSize: inf.output.embeddingVocabSize,
-    hiddenActivation,
-    swigluLimit: inf.ffn.swigluLimit,
-    stopTokenIds,
-    layerTypes,
-    linearNumKeyHeads,
-    linearNumValueHeads,
-    linearKeyHeadDim,
-    linearValueHeadDim,
-    linearConvKernelDim,
-    linearNormMode,
-    attentionBias: inf.attention.attentionBias,
-    causalAttention,
-    finalLogitSoftcapping: inf.output.finalLogitSoftcapping,
-    attnLogitSoftcapping: inf.attention.attnLogitSoftcapping,
-    queryKeyNorm: inf.attention.queryKeyNorm,
-    attentionOutputGate: inf.attention.attentionOutputGate === true,
-    queryPreAttnScalar,
-    layerPipeline: inf.pipeline ?? null,
-    chatTemplateType,
-    chatTemplateEnabled,
-    kernelPath: inf.defaultKernelPath,
-    visionConfig: resolveVisionConfig(config, manifest),
-  };
-}
-
-
-export function parseModelConfigFromManifest(manifest, runtimeOverrides) {
-  assertSupportedRuntimeModelType(manifest);
-
-  // Merge manifest inference with runtime overrides
-  const merged = mergeConfig(
-    {
-      modelId: manifest.modelId ?? 'unknown',
-      inference: manifest.inference,
-      architecture: manifest.architecture,
-    },
-    runtimeOverrides
-  );
-
-  // Log config source info
-  const runtimeSources = Array.from(merged._sources.entries())
-    .filter(([, src]) => src === 'runtime')
-    .length;
-  const totalSources = merged._sources.size;
-  if (runtimeSources > 0) {
-    log.info('Config', `Manifest-first config: ${totalSources - runtimeSources} from manifest, ${runtimeSources} from runtime`);
-  } else {
-    log.debug('Config', `Manifest-first config: ${totalSources} values from manifest`);
-  }
-
-  // Convert to ParsedModelConfig
-  return toParsedConfigFromMerged(merged, manifest);
-}
-
-// =============================================================================
-// Main Entry Point
-// =============================================================================
-
-
-export function parseModelConfig(manifest, runtimeOverrides) {
-  // Manifest-first architecture: inference config is required
-  if (!hasManifestInference(manifest)) {
-    const modelId = manifest.modelId ?? 'unknown';
-    throw new Error(
-      `Manifest "${modelId}" is missing inference config. ` +
-      `Re-convert the model using the latest converter to add manifest.inference. ` +
-      `Legacy preset-based resolution has been removed.`
-    );
-  }
-
-  log.info('Config', 'Using manifest-first config (source of truth)');
-  return parseModelConfigFromManifest(manifest, runtimeOverrides);
-}