fix: remove tfs_z

ref: ggml-org/llama.cpp#10071

Author: jhen0409

android/src/main/jni.cpp

@@ -176,7 +176,7 @@ Java_com_rnllama_LlamaContext_initContext(
     defaultParams.cpuparams.n_threads = n_threads > 0 ? n_threads : default_n_threads;
 
     defaultParams.n_gpu_layers = n_gpu_layers;
-    
+
     defaultParams.use_mlock = use_mlock;
     defaultParams.use_mmap = use_mmap;
 
@@ -380,7 +380,6 @@ Java_com_rnllama_LlamaContext_doCompletion(
     jfloat min_p,
     jfloat xtc_threshold,
     jfloat xtc_probability,
-    jfloat tfs_z,
     jfloat typical_p,
     jint seed,
     jobjectArray stop,
@@ -419,7 +418,6 @@ Java_com_rnllama_LlamaContext_doCompletion(
     sparams.top_k = top_k;
     sparams.top_p = top_p;
     sparams.min_p = min_p;
-    sparams.tfs_z = tfs_z;
     sparams.typ_p = typical_p;
     sparams.n_probs = n_probs;
     sparams.grammar = env->GetStringUTFChars(grammar, nullptr);
@@ -546,7 +544,7 @@ Java_com_rnllama_LlamaContext_doCompletion(
     putInt(env, result, "tokens_cached", llama->n_past);
 
     const auto timings_token = llama_perf_context(llama -> ctx);
-    
+
     auto timingsResult = createWriteableMap(env);
     putInt(env, timingsResult, "prompt_n", timings_token.n_p_eval);
     putInt(env, timingsResult, "prompt_ms", timings_token.t_p_eval_ms);
@@ -643,7 +641,7 @@ Java_com_rnllama_LlamaContext_embedding(
     llama->rewind();
 
     llama_perf_context_reset(llama->ctx);
-    
+
     llama->params.prompt = text_chars;
 
     llama->params.n_predict = 0;

example/src/App.tsx

@@ -353,7 +353,6 @@ export default function App() {
           temperature: 0.7,
           top_k: 40, // <= 0 to use vocab size
           top_p: 0.5, // 1.0 = disabled
-          tfs_z: 1.0, // 1.0 = disabled
           typical_p: 1.0, // 1.0 = disabled
           penalty_last_n: 256, // 0 = disable penalty, -1 = context size
           penalty_repeat: 1.18, // 1.0 = disabled

ios/RNLlamaContext.mm

@@ -209,7 +209,6 @@ - (NSDictionary *)completion:(NSDictionary *)params
     if (params[@"min_p"]) sparams.min_p = [params[@"min_p"] doubleValue];
     if (params[@"xtc_threshold"]) sparams.xtc_threshold = [params[@"xtc_threshold"] doubleValue];
     if (params[@"xtc_probability"]) sparams.xtc_probability = [params[@"xtc_probability"] doubleValue];
-    if (params[@"tfs_z"]) sparams.tfs_z = [params[@"tfs_z"] doubleValue];
     if (params[@"typical_p"]) sparams.typ_p = [params[@"typical_p"] doubleValue];
 
     if (params[@"grammar"]) {

src/NativeRNLlama.ts

@@ -37,7 +37,6 @@ export type NativeCompletionParams = {
   min_p?: number
   xtc_threshold?: number
   xtc_probability?: number
-  tfs_z?: number
   typical_p?: number
   temperature?: number // -> temp
   penalty_last_n?: number