Skip to content

Configure llama temperature only for non-greedy samplers #18

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
mattt merged 1 commit into from
Nov 3, 2025
Merged

Configure llama temperature only for non-greedy samplers #18

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 12 additions & 8 deletions Sources/AnyLanguageModel/Models/LlamaLanguageModel.swift
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -308,10 +308,6 @@ import Foundation
}
defer { llama_sampler_free(sampler) }

// Use temperature from options if provided, otherwise use model's default
let effectiveTemperature = options.temperature.map { Float($0) } ?? temperature
llama_sampler_chain_add(sampler, llama_sampler_init_temp(effectiveTemperature))

// Use sampling parameters from options if provided
if let sampling = options.sampling {
switch sampling.mode {
Expand All @@ -321,12 +317,18 @@ import Foundation
case .topK(let k, let seed):
llama_sampler_chain_add(sampler, llama_sampler_init_top_k(Int32(k)))
llama_sampler_chain_add(sampler, llama_sampler_init_top_p(1.0, 1))
if let temperature = options.temperature {
llama_sampler_chain_add(sampler, llama_sampler_init_temp(Float(temperature)))
}
if let seed = seed {
llama_sampler_chain_add(sampler, llama_sampler_init_dist(UInt32(seed)))
}
case .nucleus(let threshold, let seed):
llama_sampler_chain_add(sampler, llama_sampler_init_top_k(0)) // Disable top-k
llama_sampler_chain_add(sampler, llama_sampler_init_top_p(Float(threshold), 1))
if let temperature = options.temperature {
llama_sampler_chain_add(sampler, llama_sampler_init_temp(Float(temperature)))
}
if let seed = seed {
llama_sampler_chain_add(sampler, llama_sampler_init_dist(UInt32(seed)))
}
Expand Down Expand Up @@ -450,10 +452,6 @@ import Foundation
}
defer { llama_sampler_free(sampler) }

// Use temperature from options if provided, otherwise use model's default
let effectiveTemperature = options.temperature.map { Float($0) } ?? self.temperature
llama_sampler_chain_add(sampler, llama_sampler_init_temp(effectiveTemperature))

// Use sampling parameters from options if provided
if let sampling = options.sampling {
switch sampling.mode {
Expand All @@ -463,12 +461,18 @@ import Foundation
case .topK(let k, let seed):
llama_sampler_chain_add(sampler, llama_sampler_init_top_k(Int32(k)))
llama_sampler_chain_add(sampler, llama_sampler_init_top_p(1.0, 1))
if let temperature = options.temperature {
llama_sampler_chain_add(sampler, llama_sampler_init_temp(Float(temperature)))
}
if let seed = seed {
llama_sampler_chain_add(sampler, llama_sampler_init_dist(UInt32(seed)))
}
case .nucleus(let threshold, let seed):
llama_sampler_chain_add(sampler, llama_sampler_init_top_k(0)) // Disable top-k
llama_sampler_chain_add(sampler, llama_sampler_init_top_p(Float(threshold), 1))
if let temperature = options.temperature {
llama_sampler_chain_add(sampler, llama_sampler_init_temp(Float(temperature)))
}
if let seed = seed {
llama_sampler_chain_add(sampler, llama_sampler_init_dist(UInt32(seed)))
}
Expand Down
14 changes: 14 additions & 0 deletions Tests/AnyLanguageModelTests/LlamaLanguageModelTests.swift
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -125,5 +125,19 @@ import Testing
// Response should be limited by max tokens
#expect(!response.content.isEmpty)
}

@Test func greedySamplingWithTemperature() async throws {
let session = LanguageModelSession(model: model)
let options = GenerationOptions(
sampling: .greedy,
temperature: 0.7,
maximumResponseTokens: 50
)
let response = try await session.respond(
to: "Tell me a fact",
options: options
)
#expect(!response.content.isEmpty)
}
}
#endif // Llama