MarkBase Admin
ac75faa0cc
CI / build-and-test (push) Has been cancelled
- E4B-MarkBase model (42 layers, 4.4GB) loaded successfully - All Phase 1-6 tests passed (model loading, forward pass, vision/audio towers, token generation, performance) - All stress tests passed (5/5 in 127.6s) - Concurrent inference - Memory stress (67.5 tok/s, 0 NaN) - Continuous generation - Batch processing - Long-running stability - Swift Metal inference engine with multimodal support
113 lines
3.4 KiB
Swift
113 lines
3.4 KiB
Swift
import Foundation
|
|
|
|
// ─────────────────────────────────────────────────────────────
|
|
// Embeddings API Models
|
|
// OpenAI Compatible Format
|
|
// ─────────────────────────────────────────────────────────────
|
|
|
|
/// Embeddings request
|
|
public struct EmbeddingsRequest: Codable {
|
|
public let model: String
|
|
public let input: InputType
|
|
public let encoding_format: String?
|
|
|
|
public enum InputType: Codable, Sendable {
|
|
case string(String)
|
|
case strings([String])
|
|
case tokens([Int])
|
|
case tokensList([[Int]])
|
|
|
|
public init(from decoder: Decoder) throws {
|
|
let container = try decoder.singleValueContainer()
|
|
|
|
if let string = try? container.decode(String.self) {
|
|
self = .string(string)
|
|
} else if let strings = try? container.decode([String].self) {
|
|
self = .strings(strings)
|
|
} else if let tokens = try? container.decode([Int].self) {
|
|
self = .tokens(tokens)
|
|
} else if let tokensList = try? container.decode([[Int]].self) {
|
|
self = .tokensList(tokensList)
|
|
} else {
|
|
throw DecodingError.dataCorruptedError(
|
|
in: container,
|
|
debugDescription: "Invalid input type"
|
|
)
|
|
}
|
|
}
|
|
|
|
public func encode(to encoder: Encoder) throws {
|
|
var container = encoder.singleValueContainer()
|
|
|
|
switch self {
|
|
case .string(let string):
|
|
try container.encode(string)
|
|
case .strings(let strings):
|
|
try container.encode(strings)
|
|
case .tokens(let tokens):
|
|
try container.encode(tokens)
|
|
case .tokensList(let tokensList):
|
|
try container.encode(tokensList)
|
|
}
|
|
}
|
|
}
|
|
|
|
public init(
|
|
model: String,
|
|
input: InputType,
|
|
encoding_format: String? = nil
|
|
) {
|
|
self.model = model
|
|
self.input = input
|
|
self.encoding_format = encoding_format
|
|
}
|
|
}
|
|
|
|
/// Embeddings response
|
|
public struct EmbeddingsResponse: Codable, Sendable {
|
|
public let object: String
|
|
public let data: [EmbeddingData]
|
|
public let model: String
|
|
public let usage: EmbeddingUsage
|
|
|
|
public init(
|
|
object: String = "list",
|
|
data: [EmbeddingData],
|
|
model: String,
|
|
usage: EmbeddingUsage
|
|
) {
|
|
self.object = object
|
|
self.data = data
|
|
self.model = model
|
|
self.usage = usage
|
|
}
|
|
}
|
|
|
|
/// Single embedding data
|
|
public struct EmbeddingData: Codable, Sendable {
|
|
public let object: String
|
|
public let index: Int
|
|
public let embedding: [Float]
|
|
|
|
public init(
|
|
object: String = "embedding",
|
|
index: Int,
|
|
embedding: [Float]
|
|
) {
|
|
self.object = object
|
|
self.index = index
|
|
self.embedding = embedding
|
|
}
|
|
}
|
|
|
|
/// Embedding usage
|
|
public struct EmbeddingUsage: Codable, Sendable {
|
|
public let prompt_tokens: Int
|
|
public let total_tokens: Int
|
|
|
|
public init(prompt_tokens: Int, total_tokens: Int) {
|
|
self.prompt_tokens = prompt_tokens
|
|
self.total_tokens = total_tokens
|
|
}
|
|
}
|