cui-llama.rn 1.2.6 → 1.3.0

package/ios/RNLlama.mm

@@ -21,10 +21,25 @@ RCT_EXPORT_METHOD(setContextLimit:(double)limit
     resolve(nil);
 }
 
-RCT_EXPORT_METHOD(initContext:(NSDictionary *)contextParams
+RCT_EXPORT_METHOD(modelInfo:(NSString *)path
+                 withSkip:(NSArray *)skip
                  withResolver:(RCTPromiseResolveBlock)resolve
                  withRejecter:(RCTPromiseRejectBlock)reject)
 {
+    resolve([RNLlamaContext modelInfo:path skip:skip]);
+}
+
+RCT_EXPORT_METHOD(initContext:(double)contextId
+                 withContextParams:(NSDictionary *)contextParams
+                 withResolver:(RCTPromiseResolveBlock)resolve
+                 withRejecter:(RCTPromiseRejectBlock)reject)
+{
+    NSNumber *contextIdNumber = [NSNumber numberWithDouble:contextId];
+    if (llamaContexts[contextIdNumber] != nil) {
+        reject(@"llama_error", @"Context already exists", nil);
+        return;
+    }
+
     if (llamaDQueue == nil) {
       llamaDQueue = dispatch_queue_create("com.rnllama", DISPATCH_QUEUE_SERIAL);
     }
@@ -38,23 +53,27 @@ RCT_EXPORT_METHOD(initContext:(NSDictionary *)contextParams
         return;
     }
 
-    RNLlamaContext *context = [RNLlamaContext initWithParams:contextParams];
-    if (![context isModelLoaded]) {
-        reject(@"llama_cpp_error", @"Failed to load the model", nil);
-        return;
+    @try {
+      RNLlamaContext *context = [RNLlamaContext initWithParams:contextParams onProgress:^(unsigned int progress) {
+          dispatch_async(dispatch_get_main_queue(), ^{
+              [self sendEventWithName:@"@RNLlama_onInitContextProgress" body:@{ @"contextId": @(contextId), @"progress": @(progress) }];
+          });
+      }];
+      if (![context isModelLoaded]) {
+          reject(@"llama_cpp_error", @"Failed to load the model", nil);
+          return;
+      }
+
+      [llamaContexts setObject:context forKey:contextIdNumber];
+
+      resolve(@{
+          @"gpu": @([context isMetalEnabled]),
+          @"reasonNoGPU": [context reasonNoMetal],
+          @"model": [context modelInfo],
+      });
+    } @catch (NSException *exception) {
+      reject(@"llama_cpp_error", exception.reason, nil);
     }
-
-    double contextId = (double) arc4random_uniform(1000000);
-
-    NSNumber *contextIdNumber = [NSNumber numberWithDouble:contextId];
-    [llamaContexts setObject:context forKey:contextIdNumber];
-
-    resolve(@{
-        @"contextId": contextIdNumber,
-        @"gpu": @([context isMetalEnabled]),
-        @"reasonNoGPU": [context reasonNoMetal],
-        @"model": [context modelInfo],
-    });
 }
 
 RCT_EXPORT_METHOD(getFormattedChat:(double)contextId
@@ -125,6 +144,7 @@ RCT_EXPORT_METHOD(saveSession:(double)contextId
 
 - (NSArray *)supportedEvents {
   return@[
+    @"@RNLlama_onInitContextProgress",
     @"@RNLlama_onToken",
   ];
 }
@@ -213,6 +233,7 @@ RCT_EXPORT_METHOD(detokenize:(double)contextId
 
 RCT_EXPORT_METHOD(embedding:(double)contextId
                   text:(NSString *)text
+                  params:(NSDictionary *)params
                   withResolver:(RCTPromiseResolveBlock)resolve
                   withRejecter:(RCTPromiseRejectBlock)reject)
 {
@@ -222,9 +243,8 @@ RCT_EXPORT_METHOD(embedding:(double)contextId
         return;
     }
     @try {
-        NSMutableArray *embedding = [context embedding:text];
-        resolve(@{ @"embedding": embedding });
-        [embedding release];
+        NSDictionary *embedding = [context embedding:text params:params];
+        resolve(embedding);
     } @catch (NSException *exception) {
         reject(@"llama_cpp_error", exception.reason, nil);
     }
@@ -260,6 +280,9 @@ RCT_EXPORT_METHOD(releaseContext:(double)contextId
         reject(@"llama_error", @"Context not found", nil);
         return;
     }
+    if (![context isModelLoaded]) {
+      [context interruptLoad];
+    }
     [context stopCompletion];
     dispatch_barrier_sync(llamaDQueue, ^{});
     [context invalidate];

package/ios/RNLlamaContext.h

@@ -1,18 +1,24 @@
 #ifdef __cplusplus
 #import "llama.h"
+#import "llama-impl.h"
+#import "ggml.h"
 #import "rn-llama.hpp"
 #endif
 
 
 @interface RNLlamaContext : NSObject {
     bool is_metal_enabled;
-    NSString * reason_no_metal;
     bool is_model_loaded;
+    NSString * reason_no_metal;
+
+    void (^onProgress)(unsigned int progress);
 
     rnllama::llama_rn_context * llama;
 }
 
-+ (instancetype)initWithParams:(NSDictionary *)params;
++ (NSDictionary *)modelInfo:(NSString *)path skip:(NSArray *)skip;
++ (instancetype)initWithParams:(NSDictionary *)params onProgress:(void (^)(unsigned int progress))onProgress;
+- (void)interruptLoad;
 - (bool)isMetalEnabled;
 - (NSString *)reasonNoMetal;
 - (NSDictionary *)modelInfo;
@@ -22,7 +28,7 @@
 - (void)stopCompletion;
 - (NSArray *)tokenize:(NSString *)text;
 - (NSString *)detokenize:(NSArray *)tokens;
-- (NSArray *)embedding:(NSString *)text;
+- (NSDictionary *)embedding:(NSString *)text params:(NSDictionary *)params;
 - (NSString *)getFormattedChat:(NSArray *)messages withTemplate:(NSString *)chatTemplate;
 - (NSDictionary *)loadSession:(NSString *)path;
 - (int)saveSession:(NSString *)path size:(int)size;

package/ios/RNLlamaContext.mm

@@ -3,9 +3,53 @@
 
 @implementation RNLlamaContext
 
-+ (instancetype)initWithParams:(NSDictionary *)params {
++ (NSDictionary *)modelInfo:(NSString *)path skip:(NSArray *)skip {
+    struct lm_gguf_init_params params = {
+        /*.no_alloc = */ false,
+        /*.ctx      = */ NULL,
+    };
+
+    struct lm_gguf_context * ctx = lm_gguf_init_from_file([path UTF8String], params);
+
+    if (!ctx) {
+        NSLog(@"%s: failed to load '%s'\n", __func__, [path UTF8String]);
+        return @{};
+    }
+
+    NSMutableDictionary *info = [[NSMutableDictionary alloc] init];
+
+    info[@"version"] = @(lm_gguf_get_version(ctx));
+    info[@"alignment"] = @(lm_gguf_get_alignment(ctx));
+    info[@"data_offset"] = @(lm_gguf_get_data_offset(ctx));
+
+    // kv
+    {
+        const int n_kv = lm_gguf_get_n_kv(ctx);
+
+        for (int i = 0; i < n_kv; ++i) {
+            const char * key = lm_gguf_get_key(ctx, i);
+
+            if (skip && [skip containsObject:[NSString stringWithUTF8String:key]]) {
+                continue;
+            }
+            const std::string value = rnllama::lm_gguf_kv_to_str(ctx, i);
+            info[[NSString stringWithUTF8String:key]] = [NSString stringWithUTF8String:value.c_str()];
+        }
+    }
+
+    lm_gguf_free(ctx);
+
+    return info;
+}
+
++ (instancetype)initWithParams:(NSDictionary *)params onProgress:(void (^)(unsigned int progress))onProgress {
     // llama_backend_init(false);
-    gpt_params defaultParams;
+    common_params defaultParams;
+
+    if (params[@"vocab_only"]) {
+        defaultParams.vocab_only = [params[@"vocab_only"] boolValue];
+        defaultParams.warmup = false;
+    }
 
     NSString *modelPath = params[@"model"];
     BOOL isAsset = [params[@"is_model_asset"] boolValue];
@@ -13,10 +57,6 @@
     if (isAsset) path = [[NSBundle mainBundle] pathForResource:modelPath ofType:nil];
     defaultParams.model = [path UTF8String];
 
-    if (params[@"embedding"] && [params[@"embedding"] boolValue]) {
-        defaultParams.embedding = true;
-    }
-
     if (params[@"n_ctx"]) defaultParams.n_ctx = [params[@"n_ctx"] intValue];
     if (params[@"use_mlock"]) defaultParams.use_mlock = [params[@"use_mlock"]boolValue];
 
@@ -56,35 +96,79 @@
     if (params[@"n_batch"]) defaultParams.n_batch = [params[@"n_batch"] intValue];
     if (params[@"use_mmap"]) defaultParams.use_mmap = [params[@"use_mmap"] boolValue];
 
+    if (params[@"pooling_type"] && [params[@"pooling_type"] isKindOfClass:[NSNumber class]]) {
+      defaultParams.pooling_type = static_cast<enum llama_pooling_type>([params[@"pooling_type"] intValue]);
+    }
+
+    if (params[@"embedding"] && [params[@"embedding"] boolValue]) {
+        defaultParams.embedding = true;
+        // For non-causal models, batch size must be equal to ubatch size
+        defaultParams.n_ubatch = defaultParams.n_batch;
+
+        if (params[@"embd_normalize"] && [params[@"embd_normalize"] isKindOfClass:[NSNumber class]]) {
+            defaultParams.embd_normalize = [params[@"embd_normalize"] intValue];
+        }
+    }
+
     if (params[@"lora"]) {
         float lora_scaled = 1.0f;
         if (params[@"lora_scaled"]) lora_scaled = [params[@"lora_scaled"] floatValue];
-        defaultParams.lora_adapter.push_back({[params[@"lora"] UTF8String], lora_scaled});
-        defaultParams.use_mmap = false;
+        defaultParams.lora_adapters.push_back({[params[@"lora"] UTF8String], lora_scaled});
     }
 
     if (params[@"rope_freq_base"]) defaultParams.rope_freq_base = [params[@"rope_freq_base"] floatValue];
     if (params[@"rope_freq_scale"]) defaultParams.rope_freq_scale = [params[@"rope_freq_scale"] floatValue];
 
-    if (params[@"seed"]) defaultParams.seed = [params[@"seed"] intValue];
+    if (params[@"flash_attn"] && [params[@"flash_attn"] boolValue]) defaultParams.flash_attn = true;
+
+    if (params[@"cache_type_k"]) defaultParams.cache_type_k = [params[@"cache_type_k"] UTF8String];
+    if (params[@"cache_type_v"]) defaultParams.cache_type_v = [params[@"cache_type_v"] UTF8String];
 
     int nThreads = params[@"n_threads"] ? [params[@"n_threads"] intValue] : 0;
     const int maxThreads = (int) [[NSProcessInfo processInfo] processorCount];
     // Use 2 threads by default on 4-core devices, 4 threads on more cores
     const int defaultNThreads = nThreads == 4 ? 2 : MIN(4, maxThreads);
-    defaultParams.n_threads = nThreads > 0 ? nThreads : defaultNThreads;
+    defaultParams.cpuparams.n_threads = nThreads > 0 ? nThreads : defaultNThreads;
 
     RNLlamaContext *context = [[RNLlamaContext alloc] init];
-    if (context->llama == nullptr) {
-        context->llama = new rnllama::llama_rn_context();
+    context->llama = new rnllama::llama_rn_context();
+    context->llama->is_load_interrupted = false;
+    context->llama->loading_progress = 0;
+    context->onProgress = onProgress;
+
+    if (params[@"use_progress_callback"] && [params[@"use_progress_callback"] boolValue]) {
+        defaultParams.progress_callback = [](float progress, void * user_data) {
+            RNLlamaContext *context = (__bridge RNLlamaContext *)(user_data);
+            unsigned percentage = (unsigned) (100 * progress);
+            if (percentage > context->llama->loading_progress) {
+                context->llama->loading_progress = percentage;
+                context->onProgress(percentage);
+            }
+            return !context->llama->is_load_interrupted;
+        };
+        defaultParams.progress_callback_user_data = context;
     }
+
     context->is_model_loaded = context->llama->loadModel(defaultParams);
+
+    if (
+        params[@"embedding"] && [params[@"embedding"] boolValue] &&
+        llama_model_has_encoder(context->llama->model) && llama_model_has_decoder(context->llama->model)
+    ) {
+        delete context->llama;
+        @throw [NSException exceptionWithName:@"LlamaException" reason:@"Embedding is not supported in encoder-decoder models" userInfo:nil];
+    }
+
     context->is_metal_enabled = isMetalEnabled;
     context->reason_no_metal = reasonNoMetal;
 
     return context;
 }
 
+- (void)interruptLoad {
+    llama->is_load_interrupted = true;
+}
+
 - (bool)isMetalEnabled {
     return is_metal_enabled;
 }
@@ -128,7 +212,7 @@
 }
 
 - (NSString *)getFormattedChat:(NSArray *)messages withTemplate:(NSString *)chatTemplate {
-  std::vector<llama_chat_msg> chat;
+  std::vector<common_chat_msg> chat;
 
   for (NSDictionary *msg in messages) {
     std::string role = [[msg objectForKey:@"role"] UTF8String];
@@ -137,7 +221,7 @@
   }
 
   auto tmpl = chatTemplate == nil ? "" : [chatTemplate UTF8String];
-  auto formatted_chat = llama_chat_apply_template(llama->model, tmpl, chat, true);
+  auto formatted_chat = common_chat_apply_template(llama->model, tmpl, chat, true);
   return [NSString stringWithUTF8String:formatted_chat.c_str()];
 }
 
@@ -168,21 +252,22 @@
 {
     llama->rewind();
 
-    llama_reset_timings(llama->ctx);
+    //llama_reset_timings(llama->ctx);
 
     NSString *prompt = [params objectForKey:@"prompt"];
 
     llama->params.prompt = [prompt UTF8String];
-    llama->params.seed = params[@"seed"] ? [params[@"seed"] intValue] : -1;
+    llama->params.sparams.seed = params[@"seed"] ? [params[@"seed"] intValue] : -1;
 
     if (params[@"n_threads"]) {
-        int nThreads = params[@"n_threads"] ? [params[@"n_threads"] intValue] : llama->params.n_threads;
+        int nThreads = params[@"n_threads"] ? [params[@"n_threads"] intValue] : llama->params.cpuparams.n_threads;
         const int maxThreads = (int) [[NSProcessInfo processInfo] processorCount];
         // Use 2 threads by default on 4-core devices, 4 threads on more cores
         const int defaultNThreads = nThreads == 4 ? 2 : MIN(4, maxThreads);
-        llama->params.n_threads = nThreads > 0 ? nThreads : defaultNThreads;
+        llama->params.cpuparams.n_threads = nThreads > 0 ? nThreads : defaultNThreads;
     }
     if (params[@"n_predict"]) llama->params.n_predict = [params[@"n_predict"] intValue];
+    if (params[@"ignore_eos"]) llama->params.sparams.ignore_eos = [params[@"ignore_eos"] boolValue];
 
     auto & sparams = llama->params.sparams;
 
@@ -203,9 +288,9 @@
     if (params[@"top_k"]) sparams.top_k = [params[@"top_k"] intValue];
     if (params[@"top_p"]) sparams.top_p = [params[@"top_p"] doubleValue];
     if (params[@"min_p"]) sparams.min_p = [params[@"min_p"] doubleValue];
-    if (params[@"tfs_z"]) sparams.tfs_z = [params[@"tfs_z"] doubleValue];
-
-    if (params[@"typical_p"]) sparams.typical_p = [params[@"typical_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[@"typical_p"]) sparams.typ_p = [params[@"typical_p"] doubleValue];
 
     if (params[@"grammar"]) {
         sparams.grammar = [params[@"grammar"] UTF8String];
@@ -221,7 +306,7 @@
 
     sparams.logit_bias.clear();
     if (params[@"ignore_eos"] && [params[@"ignore_eos"] boolValue]) {
-        sparams.logit_bias[llama_token_eos(llama->model)] = -INFINITY;
+        sparams.logit_bias[llama_token_eos(llama->model)].bias = -INFINITY;
     }
 
     if (params[@"logit_bias"] && [params[@"logit_bias"] isKindOfClass:[NSArray class]]) {
@@ -232,9 +317,9 @@
                 llama_token tok = [el[0] intValue];
                 if (tok >= 0 && tok < n_vocab) {
                     if ([el[1] isKindOfClass:[NSNumber class]]) {
-                        sparams.logit_bias[tok] = [el[1] doubleValue];
+                        sparams.logit_bias[tok].bias = [el[1] doubleValue];
                     } else if ([el[1] isKindOfClass:[NSNumber class]] && ![el[1] boolValue]) {
-                        sparams.logit_bias[tok] = -INFINITY;
+                        sparams.logit_bias[tok].bias = -INFINITY;
                     }
                 }
             }
@@ -255,7 +340,7 @@
         if (token_with_probs.tok == -1 || llama->incomplete) {
             continue;
         }
-        const std::string token_text = llama_token_to_piece(llama->ctx, token_with_probs.tok);
+        const std::string token_text = common_token_to_piece(llama->ctx, token_with_probs.tok);
 
         size_t pos = std::min(sent_count, llama->generated_text.size());
 
@@ -290,7 +375,7 @@
             tokenResult[@"token"] = [NSString stringWithUTF8String:to_send.c_str()];
 
             if (llama->params.sparams.n_probs > 0) {
-                const std::vector<llama_token> to_send_toks = llama_tokenize(llama->ctx, to_send, false);
+                const std::vector<llama_token> to_send_toks = common_tokenize(llama->ctx, to_send, false);
                 size_t probs_pos = std::min(sent_token_probs_index, llama->generated_token_probs.size());
                 size_t probs_stop_pos = std::min(sent_token_probs_index + to_send_toks.size(), llama->generated_token_probs.size());
                 if (probs_pos < probs_stop_pos) {
@@ -305,10 +390,10 @@
         }
     }
 
-    llama_print_timings(llama->ctx);
+    llama_perf_context_print(llama->ctx);
     llama->is_predicting = false;
 
-    const auto timings = llama_get_timings(llama->ctx);
+    const auto timings = llama_perf_context(llama->ctx);
     return @{
         @"text": [NSString stringWithUTF8String:llama->generated_text.c_str()],
         @"completion_probabilities": [self tokenProbsToDict:llama->generated_token_probs],
@@ -339,7 +424,7 @@
 }
 
 - (NSArray *)tokenize:(NSString *)text {
-    const std::vector<llama_token> toks = llama_tokenize(llama->ctx, [text UTF8String], false);
+    const std::vector<llama_token> toks = common_tokenize(llama->ctx, [text UTF8String], false);
     NSMutableArray *result = [[NSMutableArray alloc] init];
     for (llama_token tok : toks) {
         [result addObject:@(tok)];
@@ -356,14 +441,22 @@
     return [NSString stringWithUTF8String:text.c_str()];
 }
 
-- (NSArray *)embedding:(NSString *)text {
+- (NSDictionary *)embedding:(NSString *)text params:(NSDictionary *)params {
     if (llama->params.embedding != true) {
         @throw [NSException exceptionWithName:@"LlamaException" reason:@"Embedding is not enabled" userInfo:nil];
     }
 
+    common_params embdParams;
+    embdParams.embedding = true;
+    embdParams.embd_normalize = llama->params.embd_normalize;
+
+    if (params[@"embd_normalize"] && [params[@"embd_normalize"] isKindOfClass:[NSNumber class]]) {
+        embdParams.embd_normalize = [params[@"embd_normalize"] intValue];
+    }
+
     llama->rewind();
 
-    llama_reset_timings(llama->ctx);
+    llama_perf_context_reset(llama->ctx);
 
     llama->params.prompt = [text UTF8String];
 
@@ -376,15 +469,22 @@
     llama->loadPrompt();
     llama->doCompletion();
 
-    std::vector<float> result = llama->getEmbedding();
+    std::vector<float> result = llama->getEmbedding(embdParams);
 
+    NSMutableDictionary *resultDict = [[NSMutableDictionary alloc] init];
     NSMutableArray *embeddingResult = [[NSMutableArray alloc] init];
     for (float f : result) {
         [embeddingResult addObject:@(f)];
     }
+    resultDict[@"embedding"] = embeddingResult;
+    NSMutableArray *promptTokens = [[NSMutableArray alloc] init];
+    for (llama_token tok : llama->embd) {
+        [promptTokens addObject:[NSString stringWithUTF8String:common_token_to_piece(llama->ctx, tok).c_str()]];
+    }
+    resultDict[@"prompt_tokens"] = promptTokens;
 
     llama->is_predicting = false;
-    return embeddingResult;
+    return resultDict;
 }
 
 - (NSDictionary *)loadSession:(NSString *)path {

package/jest/mock.js

@@ -4,7 +4,6 @@ if (!NativeModules.RNLlama) {
   NativeModules.RNLlama = {
     initContext: jest.fn(() =>
       Promise.resolve({
-        contextId: 1,
         gpu: false,
         reasonNoGPU: 'Test',
       }),

package/lib/commonjs/NativeRNLlama.js.map

@@ -1 +1 @@
-{"version":3,"names":["_reactNative","require","_default","TurboModuleRegistry","get","exports","default"],"sourceRoot":"..\\..\\src","sources":["NativeRNLlama.ts"],"mappings":";;;;;;AACA,IAAAA,YAAA,GAAAC,OAAA;AAAkD,IAAAC,QAAA,GAqKnCC,gCAAmB,CAACC,GAAG,CAAO,SAAS,CAAC;AAAAC,OAAA,CAAAC,OAAA,GAAAJ,QAAA"}
+{"version":3,"names":["_reactNative","require","_default","TurboModuleRegistry","get","exports","default"],"sourceRoot":"..\\..\\src","sources":["NativeRNLlama.ts"],"mappings":";;;;;;AACA,IAAAA,YAAA,GAAAC,OAAA;AAAkD,IAAAC,QAAA,GAwMnCC,gCAAmB,CAACC,GAAG,CAAO,SAAS,CAAC;AAAAC,OAAA,CAAAC,OAAA,GAAAJ,QAAA"}

package/lib/commonjs/index.js

@@ -18,6 +18,7 @@ Object.defineProperty(exports, "convertJsonSchemaToGrammar", {
 });
 exports.getCpuFeatures = getCpuFeatures;
 exports.initLlama = initLlama;
+exports.loadLlamaModelInfo = loadLlamaModelInfo;
 exports.releaseAllLlama = releaseAllLlama;
 exports.setContextLimit = setContextLimit;
 var _reactNative = require("react-native");
@@ -25,8 +26,8 @@ var _NativeRNLlama = _interopRequireDefault(require("./NativeRNLlama"));
 var _grammar = require("./grammar");
 var _chat = require("./chat");
 function _interopRequireDefault(obj) { return obj && obj.__esModule ? obj : { default: obj }; }
+const EVENT_ON_INIT_CONTEXT_PROGRESS = '@RNLlama_onInitContextProgress';
 const EVENT_ON_TOKEN = '@RNLlama_onToken';
-const EVENT_ON_MODEL_PROGRESS = '@RNLlama_onModelProgress';
 let EventEmitter;
 if (_reactNative.Platform.OS === 'ios') {
   // @ts-ignore
@@ -67,16 +68,18 @@ class LlamaContext {
   async saveSession(filepath, options) {
     return _NativeRNLlama.default.saveSession(this.id, filepath, (options === null || options === void 0 ? void 0 : options.tokenSize) || -1);
   }
-  async getFormattedChat(messages) {
+  async getFormattedChat(messages, template) {
     var _this$model;
     const chat = (0, _chat.formatChat)(messages);
-    return _NativeRNLlama.default.getFormattedChat(this.id, chat, (_this$model = this.model) !== null && _this$model !== void 0 && _this$model.isChatTemplateSupported ? undefined : 'chatml');
+    let tmpl = (_this$model = this.model) !== null && _this$model !== void 0 && _this$model.isChatTemplateSupported ? undefined : 'chatml';
+    if (template) tmpl = template; // Force replace if provided
+    return _NativeRNLlama.default.getFormattedChat(this.id, chat, tmpl);
   }
   async completion(params, callback) {
     let finalPrompt = params.prompt;
     if (params.messages) {
       // messages always win
-      finalPrompt = await this.getFormattedChat(params.messages);
+      finalPrompt = await this.getFormattedChat(params.messages, params.chatTemplate);
     }
     let tokenListener = callback && EventEmitter.addListener(EVENT_ON_TOKEN, evt => {
       const {
@@ -116,8 +119,8 @@ class LlamaContext {
   detokenize(tokens) {
     return _NativeRNLlama.default.detokenize(this.id, tokens);
   }
-  embedding(text) {
-    return _NativeRNLlama.default.embedding(this.id, text);
+  embedding(text, params) {
+    return _NativeRNLlama.default.embedding(this.id, text, params || {});
   }
   async bench(pp, tg, pl, nr) {
     const result = await _NativeRNLlama.default.bench(this.id, pp, tg, pl, nr);
@@ -143,30 +146,64 @@ async function getCpuFeatures() {
 async function setContextLimit(limit) {
   return _NativeRNLlama.default.setContextLimit(limit);
 }
-async function initLlama(_ref2, progressCallback) {
+let contextIdCounter = 0;
+const contextIdRandom = () => process.env.NODE_ENV === 'test' ? 0 : Math.floor(Math.random() * 100000);
+const modelInfoSkip = [
+// Large fields
+'tokenizer.ggml.tokens', 'tokenizer.ggml.token_type', 'tokenizer.ggml.merges'];
+async function loadLlamaModelInfo(model) {
+  let path = model;
+  if (path.startsWith('file://')) path = path.slice(7);
+  return _NativeRNLlama.default.modelInfo(path, modelInfoSkip);
+}
+const poolTypeMap = {
+  // -1 is unspecified as undefined
+  none: 0,
+  mean: 1,
+  cls: 2,
+  last: 3,
+  rank: 4
+};
+async function initLlama(_ref2, onProgress) {
+  var _loraPath, _removeProgressListen2;
   let {
     model,
     is_model_asset: isModelAsset,
+    pooling_type: poolingType,
+    lora,
     ...rest
   } = _ref2;
   let path = model;
   if (path.startsWith('file://')) path = path.slice(7);
-  const modelProgressListener = EventEmitter.addListener(EVENT_ON_MODEL_PROGRESS, event => {
-    if (event.progress && progressCallback) progressCallback(event.progress);
-    if (event.progress === 100) {
-      modelProgressListener.remove();
-    }
-  });
+  let loraPath = lora;
+  if ((_loraPath = loraPath) !== null && _loraPath !== void 0 && _loraPath.startsWith('file://')) loraPath = loraPath.slice(7);
+  const contextId = contextIdCounter + contextIdRandom();
+  contextIdCounter += 1;
+  let removeProgressListener = null;
+  if (onProgress) {
+    removeProgressListener = EventEmitter.addListener(EVENT_ON_INIT_CONTEXT_PROGRESS, evt => {
+      if (evt.contextId !== contextId) return;
+      onProgress(evt.progress);
+    });
+  }
+  const poolType = poolTypeMap[poolingType];
   const {
-    contextId,
     gpu,
     reasonNoGPU,
     model: modelDetails
-  } = await _NativeRNLlama.default.initContext({
+  } = await _NativeRNLlama.default.initContext(contextId, {
     model: path,
     is_model_asset: !!isModelAsset,
+    use_progress_callback: !!onProgress,
+    pooling_type: poolType,
+    lora: loraPath,
     ...rest
+  }).catch(err => {
+    var _removeProgressListen;
+    (_removeProgressListen = removeProgressListener) === null || _removeProgressListen === void 0 ? void 0 : _removeProgressListen.remove();
+    throw err;
   });
+  (_removeProgressListen2 = removeProgressListener) === null || _removeProgressListen2 === void 0 ? void 0 : _removeProgressListen2.remove();
   return new LlamaContext({
     contextId,
     gpu,

package/lib/commonjs/index.js.map

@@ -1 +1 @@
-{"version":3,"names":["_reactNative","require","_NativeRNLlama","_interopRequireDefault","_grammar","_chat","obj","__esModule","default","EVENT_ON_TOKEN","EVENT_ON_MODEL_PROGRESS","EventEmitter","Platform","OS","NativeEventEmitter","RNLlama","DeviceEventEmitter","LlamaContext","gpu","reasonNoGPU","model","constructor","_ref","contextId","id","loadSession","filepath","path","startsWith","slice","saveSession","options","tokenSize","getFormattedChat","messages","_this$model","chat","formatChat","isChatTemplateSupported","undefined","completion","params","callback","finalPrompt","prompt","tokenListener","addListener","evt","tokenResult","Error","promise","emit_partial_completion","then","completionResult","_tokenListener","remove","catch","err","_tokenListener2","stopCompletion","tokenizeAsync","text","tokenizeSync","detokenize","tokens","embedding","bench","pp","tg","pl","nr","result","modelDesc","modelSize","modelNParams","ppAvg","ppStd","tgAvg","tgStd","JSON","parse","release","releaseContext","exports","getCpuFeatures","setContextLimit","limit","initLlama","_ref2","progressCallback","is_model_asset","isModelAsset","rest","modelProgressListener","event","progress","modelDetails","initContext","releaseAllLlama","releaseAllContexts"],"sourceRoot":"..\\..\\src","sources":["index.ts"],"mappings":";;;;;;;;;;;;;;;;;;;;;;AAAA,IAAAA,YAAA,GAAAC,OAAA;AAEA,IAAAC,cAAA,GAAAC,sBAAA,CAAAF,OAAA;AAYA,IAAAG,QAAA,GAAAH,OAAA;AAEA,IAAAI,KAAA,GAAAJ,OAAA;AAAmC,SAAAE,uBAAAG,GAAA,WAAAA,GAAA,IAAAA,GAAA,CAAAC,UAAA,GAAAD,GAAA,KAAAE,OAAA,EAAAF,GAAA;AAInC,MAAMG,cAAc,GAAG,kBAAkB;AAEzC,MAAMC,uBAAuB,GAAG,0BAA0B;AAE1D,IAAIC,YAA2D;AAC/D,IAAIC,qBAAQ,CAACC,EAAE,KAAK,KAAK,EAAE;EACzB;EACAF,YAAY,GAAG,IAAIG,+BAAkB,CAACC,sBAAO,CAAC;AAChD;AACA,IAAIH,qBAAQ,CAACC,EAAE,KAAK,SAAS,EAAE;EAC7BF,YAAY,GAAGK,+BAAkB;AACnC;AAgCO,MAAMC,YAAY,CAAC;EAGxBC,GAAG,GAAY,KAAK;EAEpBC,WAAW,GAAW,EAAE;EAExBC,KAAK,GAED,CAAC,CAAC;EAENC,WAAWA,CAAAC,IAAA,EAA6D;IAAA,IAA5D;MAAEC,SAAS;MAAEL,GAAG;MAAEC,WAAW;MAAEC;IAA0B,CAAC,GAAAE,IAAA;IACpE,IAAI,CAACE,EAAE,GAAGD,SAAS;IACnB,IAAI,CAACL,GAAG,GAAGA,GAAG;IACd,IAAI,CAACC,WAAW,GAAGA,WAAW;IAC9B,IAAI,CAACC,KAAK,GAAGA,KAAK;EACpB;;EAEA;AACF;AACA;EACE,MAAMK,WAAWA,CAACC,QAAgB,EAAoC;IACpE,IAAIC,IAAI,GAAGD,QAAQ;IACnB,IAAIC,IAAI,CAACC,UAAU,CAAC,SAAS,CAAC,EAAED,IAAI,GAAGA,IAAI,CAACE,KAAK,CAAC,CAAC,CAAC;IACpD,OAAOd,sBAAO,CAACU,WAAW,CAAC,IAAI,CAACD,EAAE,EAAEG,IAAI,CAAC;EAC3C;;EAEA;AACF;AACA;EACE,MAAMG,WAAWA,CACfJ,QAAgB,EAChBK,OAA+B,EACd;IACjB,OAAOhB,sBAAO,CAACe,WAAW,CAAC,IAAI,CAACN,EAAE,EAAEE,QAAQ,EAAE,CAAAK,OAAO,aAAPA,OAAO,uBAAPA,OAAO,CAAEC,SAAS,KAAI,CAAC,CAAC,CAAC;EACzE;EAEA,MAAMC,gBAAgBA,CACpBC,QAAuC,EACtB;IAAA,IAAAC,WAAA;IACjB,MAAMC,IAAI,GAAG,IAAAC,gBAAU,EAACH,QAAQ,CAAC;IACjC,OAAOnB,sBAAO,CAACkB,gBAAgB,CAC7B,IAAI,CAACT,EAAE,EACPY,IAAI,EACJ,CAAAD,WAAA,OAAI,CAACf,KAAK,cAAAe,WAAA,eAAVA,WAAA,CAAYG,uBAAuB,GAAGC,SAAS,GAAG,QACpD,CAAC;EACH;EAEA,MAAMC,UAAUA,CACdC,MAAwB,EACxBC,QAAoC,EACH;IAEjC,IAAIC,WAAW,GAAGF,MAAM,CAACG,MAAM;IAC/B,IAAIH,MAAM,CAACP,QAAQ,EAAE;MAAE;MACrBS,WAAW,GAAG,MAAM,IAAI,CAACV,gBAAgB,CAACQ,MAAM,CAACP,QAAQ,CAAC;IAC5D;IAEA,IAAIW,aAAkB,GACpBH,QAAQ,IACR/B,YAAY,CAACmC,WAAW,CAACrC,cAAc,EAAGsC,GAAqB,IAAK;MAClE,MAAM;QAAExB,SAAS;QAAEyB;MAAY,CAAC,GAAGD,GAAG;MACtC,IAAIxB,SAAS,KAAK,IAAI,CAACC,EAAE,EAAE;MAC3BkB,QAAQ,CAACM,WAAW,CAAC;IACvB,CAAC,CAAC;IAEJ,IAAI,CAACL,WAAW,EAAE,MAAM,IAAIM,KAAK,CAAC,oBAAoB,CAAC;IACvD,MAAMC,OAAO,GAAGnC,sBAAO,CAACyB,UAAU,CAAC,IAAI,CAAChB,EAAE,EAAE;MAC1C,GAAGiB,MAAM;MACTG,MAAM,EAAED,WAAW;MACnBQ,uBAAuB,EAAE,CAAC,CAACT;IAC7B,CAAC,CAAC;IACF,OAAOQ,OAAO,CACXE,IAAI,CAAEC,gBAAgB,IAAK;MAAA,IAAAC,cAAA;MAC1B,CAAAA,cAAA,GAAAT,aAAa,cAAAS,cAAA,uBAAbA,cAAA,CAAeC,MAAM,CAAC,CAAC;MACvBV,aAAa,GAAG,IAAI;MACpB,OAAOQ,gBAAgB;IACzB,CAAC,CAAC,CACDG,KAAK,CAAEC,GAAQ,IAAK;MAAA,IAAAC,eAAA;MACnB,CAAAA,eAAA,GAAAb,aAAa,cAAAa,eAAA,uBAAbA,eAAA,CAAeH,MAAM,CAAC,CAAC;MACvBV,aAAa,GAAG,IAAI;MACpB,MAAMY,GAAG;IACX,CAAC,CAAC;EACN;EAEAE,cAAcA,CAAA,EAAkB;IAC9B,OAAO5C,sBAAO,CAAC4C,cAAc,CAAC,IAAI,CAACnC,EAAE,CAAC;EACxC;EAEAoC,aAAaA,CAACC,IAAY,EAAiC;IACzD,OAAO9C,sBAAO,CAAC6C,aAAa,CAAC,IAAI,CAACpC,EAAE,EAAEqC,IAAI,CAAC;EAC7C;EAEAC,YAAYA,CAACD,IAAY,EAAwB;IAC/C,OAAO9C,sBAAO,CAAC+C,YAAY,CAAC,IAAI,CAACtC,EAAE,EAAEqC,IAAI,CAAC;EAC5C;EAEAE,UAAUA,CAACC,MAAgB,EAAmB;IAC5C,OAAOjD,sBAAO,CAACgD,UAAU,CAAC,IAAI,CAACvC,EAAE,EAAEwC,MAAM,CAAC;EAC5C;EAEAC,SAASA,CAACJ,IAAY,EAAkC;IACtD,OAAO9C,sBAAO,CAACkD,SAAS,CAAC,IAAI,CAACzC,EAAE,EAAEqC,IAAI,CAAC;EACzC;EAEA,MAAMK,KAAKA,CACTC,EAAU,EACVC,EAAU,EACVC,EAAU,EACVC,EAAU,EACY;IACtB,MAAMC,MAAM,GAAG,MAAMxD,sBAAO,CAACmD,KAAK,CAAC,IAAI,CAAC1C,EAAE,EAAE2C,EAAE,EAAEC,EAAE,EAAEC,EAAE,EAAEC,EAAE,CAAC;IAC3D,MAAM,CAACE,SAAS,EAAEC,SAAS,EAAEC,YAAY,EAAEC,KAAK,EAAEC,KAAK,EAAEC,KAAK,EAAEC,KAAK,CAAC,GACpEC,IAAI,CAACC,KAAK,CAACT,MAAM,CAAC;IACpB,OAAO;MACLC,SAAS;MACTC,SAAS;MACTC,YAAY;MACZC,KAAK;MACLC,KAAK;MACLC,KAAK;MACLC;IACF,CAAC;EACH;EAEA,MAAMG,OAAOA,CAAA,EAAkB;IAC7B,OAAOlE,sBAAO,CAACmE,cAAc,CAAC,IAAI,CAAC1D,EAAE,CAAC;EACxC;AACF;AAAC2D,OAAA,CAAAlE,YAAA,GAAAA,YAAA;AAEM,eAAemE,cAAcA,CAAA,EAAgC;EAClE,OAAOrE,sBAAO,CAACqE,cAAc,CAAC,CAAC;AACjC;AAEO,eAAeC,eAAeA,CAACC,KAAa,EAAiB;EAClE,OAAOvE,sBAAO,CAACsE,eAAe,CAACC,KAAK,CAAC;AACvC;AAEO,eAAeC,SAASA,CAAAC,KAAA,EAK7BC,gBAA6C,EACtB;EAAA,IANO;IAC5BrE,KAAK;IACLsE,cAAc,EAAEC,YAAY;IAC5B,GAAGC;EACU,CAAC,GAAAJ,KAAA;EAGhB,IAAI7D,IAAI,GAAGP,KAAK;EAChB,IAAIO,IAAI,CAACC,UAAU,CAAC,SAAS,CAAC,EAAED,IAAI,GAAGA,IAAI,CAACE,KAAK,CAAC,CAAC,CAAC;EAEpD,MAAMgE,qBAAqB,GAAGlF,YAAY,CAACmC,WAAW,CAACpC,uBAAuB,EAAGoF,KAAK,IAAK;IACzF,IAAGA,KAAK,CAACC,QAAQ,IAAIN,gBAAgB,EACnCA,gBAAgB,CAACK,KAAK,CAACC,QAAQ,CAAC;IAClC,IAAGD,KAAK,CAACC,QAAQ,KAAK,GAAG,EAAE;MACzBF,qBAAqB,CAACtC,MAAM,CAAC,CAAC;IAChC;EACF,CAAC,CAAC;EAEF,MAAM;IACJhC,SAAS;IACTL,GAAG;IACHC,WAAW;IACXC,KAAK,EAAE4E;EACT,CAAC,GAAG,MAAMjF,sBAAO,CAACkF,WAAW,CAAC;IAC5B7E,KAAK,EAAEO,IAAI;IACX+D,cAAc,EAAE,CAAC,CAACC,YAAY;IAC9B,GAAGC;EACL,CAAC,CAAC;EAEF,OAAO,IAAI3E,YAAY,CAAC;IAAEM,SAAS;IAAEL,GAAG;IAAEC,WAAW;IAAEC,KAAK,EAAE4E;EAAa,CAAC,CAAC;AAC/E;AAEO,eAAeE,eAAeA,CAAA,EAAkB;EACrD,OAAOnF,sBAAO,CAACoF,kBAAkB,CAAC,CAAC;AACrC"}
+{"version":3,"names":["_reactNative","require","_NativeRNLlama","_interopRequireDefault","_grammar","_chat","obj","__esModule","default","EVENT_ON_INIT_CONTEXT_PROGRESS","EVENT_ON_TOKEN","EventEmitter","Platform","OS","NativeEventEmitter","RNLlama","DeviceEventEmitter","LlamaContext","gpu","reasonNoGPU","model","constructor","_ref","contextId","id","loadSession","filepath","path","startsWith","slice","saveSession","options","tokenSize","getFormattedChat","messages","template","_this$model","chat","formatChat","tmpl","isChatTemplateSupported","undefined","completion","params","callback","finalPrompt","prompt","chatTemplate","tokenListener","addListener","evt","tokenResult","Error","promise","emit_partial_completion","then","completionResult","_tokenListener","remove","catch","err","_tokenListener2","stopCompletion","tokenizeAsync","text","tokenizeSync","detokenize","tokens","embedding","bench","pp","tg","pl","nr","result","modelDesc","modelSize","modelNParams","ppAvg","ppStd","tgAvg","tgStd","JSON","parse","release","releaseContext","exports","getCpuFeatures","setContextLimit","limit","contextIdCounter","contextIdRandom","process","env","NODE_ENV","Math","floor","random","modelInfoSkip","loadLlamaModelInfo","modelInfo","poolTypeMap","none","mean","cls","last","rank","initLlama","_ref2","onProgress","_loraPath","_removeProgressListen2","is_model_asset","isModelAsset","pooling_type","poolingType","lora","rest","loraPath","removeProgressListener","progress","poolType","modelDetails","initContext","use_progress_callback","_removeProgressListen","releaseAllLlama","releaseAllContexts"],"sourceRoot":"..\\..\\src","sources":["index.ts"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;AAAA,IAAAA,YAAA,GAAAC,OAAA;AAEA,IAAAC,cAAA,GAAAC,sBAAA,CAAAF,OAAA;AAaA,IAAAG,QAAA,GAAAH,OAAA;AAEA,IAAAI,KAAA,GAAAJ,OAAA;AAAmC,SAAAE,uBAAAG,GAAA,WAAAA,GAAA,IAAAA,GAAA,CAAAC,UAAA,GAAAD,GAAA,KAAAE,OAAA,EAAAF,GAAA;AAInC,MAAMG,8BAA8B,GAAG,gCAAgC;AACvE,MAAMC,cAAc,GAAG,kBAAkB;AAEzC,IAAIC,YAA2D;AAC/D,IAAIC,qBAAQ,CAACC,EAAE,KAAK,KAAK,EAAE;EACzB;EACAF,YAAY,GAAG,IAAIG,+BAAkB,CAACC,sBAAO,CAAC;AAChD;AACA,IAAIH,qBAAQ,CAACC,EAAE,KAAK,SAAS,EAAE;EAC7BF,YAAY,GAAGK,+BAAkB;AACnC;AA0CO,MAAMC,YAAY,CAAC;EAGxBC,GAAG,GAAY,KAAK;EAEpBC,WAAW,GAAW,EAAE;EAExBC,KAAK,GAED,CAAC,CAAC;EAENC,WAAWA,CAAAC,IAAA,EAA6D;IAAA,IAA5D;MAAEC,SAAS;MAAEL,GAAG;MAAEC,WAAW;MAAEC;IAA0B,CAAC,GAAAE,IAAA;IACpE,IAAI,CAACE,EAAE,GAAGD,SAAS;IACnB,IAAI,CAACL,GAAG,GAAGA,GAAG;IACd,IAAI,CAACC,WAAW,GAAGA,WAAW;IAC9B,IAAI,CAACC,KAAK,GAAGA,KAAK;EACpB;;EAEA;AACF;AACA;EACE,MAAMK,WAAWA,CAACC,QAAgB,EAAoC;IACpE,IAAIC,IAAI,GAAGD,QAAQ;IACnB,IAAIC,IAAI,CAACC,UAAU,CAAC,SAAS,CAAC,EAAED,IAAI,GAAGA,IAAI,CAACE,KAAK,CAAC,CAAC,CAAC;IACpD,OAAOd,sBAAO,CAACU,WAAW,CAAC,IAAI,CAACD,EAAE,EAAEG,IAAI,CAAC;EAC3C;;EAEA;AACF;AACA;EACE,MAAMG,WAAWA,CACfJ,QAAgB,EAChBK,OAA+B,EACd;IACjB,OAAOhB,sBAAO,CAACe,WAAW,CAAC,IAAI,CAACN,EAAE,EAAEE,QAAQ,EAAE,CAAAK,OAAO,aAAPA,OAAO,uBAAPA,OAAO,CAAEC,SAAS,KAAI,CAAC,CAAC,CAAC;EACzE;EAEA,MAAMC,gBAAgBA,CACpBC,QAAuC,EACvCC,QAAiB,EACA;IAAA,IAAAC,WAAA;IACjB,MAAMC,IAAI,GAAG,IAAAC,gBAAU,EAACJ,QAAQ,CAAC;IACjC,IAAIK,IAAI,GAAG,CAAAH,WAAA,OAAI,CAAChB,KAAK,cAAAgB,WAAA,eAAVA,WAAA,CAAYI,uBAAuB,GAAGC,SAAS,GAAG,QAAQ;IACrE,IAAIN,QAAQ,EAAEI,IAAI,GAAGJ,QAAQ,EAAC;IAC9B,OAAOpB,sBAAO,CAACkB,gBAAgB,CAAC,IAAI,CAACT,EAAE,EAAEa,IAAI,EAAEE,IAAI,CAAC;EACtD;EAEA,MAAMG,UAAUA,CACdC,MAAwB,EACxBC,QAAoC,EACH;IACjC,IAAIC,WAAW,GAAGF,MAAM,CAACG,MAAM;IAC/B,IAAIH,MAAM,CAACT,QAAQ,EAAE;MACnB;MACAW,WAAW,GAAG,MAAM,IAAI,CAACZ,gBAAgB,CAACU,MAAM,CAACT,QAAQ,EAAES,MAAM,CAACI,YAAY,CAAC;IACjF;IAEA,IAAIC,aAAkB,GACpBJ,QAAQ,IACRjC,YAAY,CAACsC,WAAW,CAACvC,cAAc,EAAGwC,GAAqB,IAAK;MAClE,MAAM;QAAE3B,SAAS;QAAE4B;MAAY,CAAC,GAAGD,GAAG;MACtC,IAAI3B,SAAS,KAAK,IAAI,CAACC,EAAE,EAAE;MAC3BoB,QAAQ,CAACO,WAAW,CAAC;IACvB,CAAC,CAAC;IAEJ,IAAI,CAACN,WAAW,EAAE,MAAM,IAAIO,KAAK,CAAC,oBAAoB,CAAC;IACvD,MAAMC,OAAO,GAAGtC,sBAAO,CAAC2B,UAAU,CAAC,IAAI,CAAClB,EAAE,EAAE;MAC1C,GAAGmB,MAAM;MACTG,MAAM,EAAED,WAAW;MACnBS,uBAAuB,EAAE,CAAC,CAACV;IAC7B,CAAC,CAAC;IACF,OAAOS,OAAO,CACXE,IAAI,CAAEC,gBAAgB,IAAK;MAAA,IAAAC,cAAA;MAC1B,CAAAA,cAAA,GAAAT,aAAa,cAAAS,cAAA,uBAAbA,cAAA,CAAeC,MAAM,CAAC,CAAC;MACvBV,aAAa,GAAG,IAAI;MACpB,OAAOQ,gBAAgB;IACzB,CAAC,CAAC,CACDG,KAAK,CAAEC,GAAQ,IAAK;MAAA,IAAAC,eAAA;MACnB,CAAAA,eAAA,GAAAb,aAAa,cAAAa,eAAA,uBAAbA,eAAA,CAAeH,MAAM,CAAC,CAAC;MACvBV,aAAa,GAAG,IAAI;MACpB,MAAMY,GAAG;IACX,CAAC,CAAC;EACN;EAEAE,cAAcA,CAAA,EAAkB;IAC9B,OAAO/C,sBAAO,CAAC+C,cAAc,CAAC,IAAI,CAACtC,EAAE,CAAC;EACxC;EAEAuC,aAAaA,CAACC,IAAY,EAAiC;IACzD,OAAOjD,sBAAO,CAACgD,aAAa,CAAC,IAAI,CAACvC,EAAE,EAAEwC,IAAI,CAAC;EAC7C;EAEAC,YAAYA,CAACD,IAAY,EAAwB;IAC/C,OAAOjD,sBAAO,CAACkD,YAAY,CAAC,IAAI,CAACzC,EAAE,EAAEwC,IAAI,CAAC;EAC5C;EAEAE,UAAUA,CAACC,MAAgB,EAAmB;IAC5C,OAAOpD,sBAAO,CAACmD,UAAU,CAAC,IAAI,CAAC1C,EAAE,EAAE2C,MAAM,CAAC;EAC5C;EAEAC,SAASA,CACPJ,IAAY,EACZrB,MAAwB,EACQ;IAChC,OAAO5B,sBAAO,CAACqD,SAAS,CAAC,IAAI,CAAC5C,EAAE,EAAEwC,IAAI,EAAErB,MAAM,IAAI,CAAC,CAAC,CAAC;EACvD;EAEA,MAAM0B,KAAKA,CACTC,EAAU,EACVC,EAAU,EACVC,EAAU,EACVC,EAAU,EACY;IACtB,MAAMC,MAAM,GAAG,MAAM3D,sBAAO,CAACsD,KAAK,CAAC,IAAI,CAAC7C,EAAE,EAAE8C,EAAE,EAAEC,EAAE,EAAEC,EAAE,EAAEC,EAAE,CAAC;IAC3D,MAAM,CAACE,SAAS,EAAEC,SAAS,EAAEC,YAAY,EAAEC,KAAK,EAAEC,KAAK,EAAEC,KAAK,EAAEC,KAAK,CAAC,GACpEC,IAAI,CAACC,KAAK,CAACT,MAAM,CAAC;IACpB,OAAO;MACLC,SAAS;MACTC,SAAS;MACTC,YAAY;MACZC,KAAK;MACLC,KAAK;MACLC,KAAK;MACLC;IACF,CAAC;EACH;EAEA,MAAMG,OAAOA,CAAA,EAAkB;IAC7B,OAAOrE,sBAAO,CAACsE,cAAc,CAAC,IAAI,CAAC7D,EAAE,CAAC;EACxC;AACF;AAAC8D,OAAA,CAAArE,YAAA,GAAAA,YAAA;AAEM,eAAesE,cAAcA,CAAA,EAAgC;EAClE,OAAOxE,sBAAO,CAACwE,cAAc,CAAC,CAAC;AACjC;AAEO,eAAeC,eAAeA,CAACC,KAAa,EAAiB;EAClE,OAAO1E,sBAAO,CAACyE,eAAe,CAACC,KAAK,CAAC;AACvC;AAEA,IAAIC,gBAAgB,GAAG,CAAC;AACxB,MAAMC,eAAe,GAAGA,CAAA,KACtBC,OAAO,CAACC,GAAG,CAACC,QAAQ,KAAK,MAAM,GAAG,CAAC,GAAGC,IAAI,CAACC,KAAK,CAACD,IAAI,CAACE,MAAM,CAAC,CAAC,GAAG,MAAM,CAAC;AAE1E,MAAMC,aAAa,GAAG;AACpB;AACA,uBAAuB,EACvB,2BAA2B,EAC3B,uBAAuB,CACxB;AACM,eAAeC,kBAAkBA,CAAC/E,KAAa,EAAmB;EACvE,IAAIO,IAAI,GAAGP,KAAK;EAChB,IAAIO,IAAI,CAACC,UAAU,CAAC,SAAS,CAAC,EAAED,IAAI,GAAGA,IAAI,CAACE,KAAK,CAAC,CAAC,CAAC;EACpD,OAAOd,sBAAO,CAACqF,SAAS,CAACzE,IAAI,EAAEuE,aAAa,CAAC;AAC/C;AAEA,MAAMG,WAAW,GAAG;EAClB;EACAC,IAAI,EAAE,CAAC;EACPC,IAAI,EAAE,CAAC;EACPC,GAAG,EAAE,CAAC;EACNC,IAAI,EAAE,CAAC;EACPC,IAAI,EAAE;AACR,CAAC;AAEM,eAAeC,SAASA,CAAAC,KAAA,EAQ7BC,UAAuC,EAChB;EAAA,IAAAC,SAAA,EAAAC,sBAAA;EAAA,IARvB;IACE3F,KAAK;IACL4F,cAAc,EAAEC,YAAY;IAC5BC,YAAY,EAAEC,WAAW;IACzBC,IAAI;IACJ,GAAGC;EACU,CAAC,GAAAT,KAAA;EAGhB,IAAIjF,IAAI,GAAGP,KAAK;EAChB,IAAIO,IAAI,CAACC,UAAU,CAAC,SAAS,CAAC,EAAED,IAAI,GAAGA,IAAI,CAACE,KAAK,CAAC,CAAC,CAAC;EAEpD,IAAIyF,QAAQ,GAAGF,IAAI;EACnB,KAAAN,SAAA,GAAIQ,QAAQ,cAAAR,SAAA,eAARA,SAAA,CAAUlF,UAAU,CAAC,SAAS,CAAC,EAAE0F,QAAQ,GAAGA,QAAQ,CAACzF,KAAK,CAAC,CAAC,CAAC;EAEjE,MAAMN,SAAS,GAAGmE,gBAAgB,GAAGC,eAAe,CAAC,CAAC;EACtDD,gBAAgB,IAAI,CAAC;EAErB,IAAI6B,sBAA2B,GAAG,IAAI;EACtC,IAAIV,UAAU,EAAE;IACdU,sBAAsB,GAAG5G,YAAY,CAACsC,WAAW,CAC/CxC,8BAA8B,EAC7ByC,GAA4C,IAAK;MAChD,IAAIA,GAAG,CAAC3B,SAAS,KAAKA,SAAS,EAAE;MACjCsF,UAAU,CAAC3D,GAAG,CAACsE,QAAQ,CAAC;IAC1B,CACF,CAAC;EACH;EAEA,MAAMC,QAAQ,GAAGpB,WAAW,CAACc,WAAW,CAA6B;EACrE,MAAM;IACJjG,GAAG;IACHC,WAAW;IACXC,KAAK,EAAEsG;EACT,CAAC,GAAG,MAAM3G,sBAAO,CAAC4G,WAAW,CAACpG,SAAS,EAAE;IACvCH,KAAK,EAAEO,IAAI;IACXqF,cAAc,EAAE,CAAC,CAACC,YAAY;IAC9BW,qBAAqB,EAAE,CAAC,CAACf,UAAU;IACnCK,YAAY,EAAEO,QAAQ;IACtBL,IAAI,EAAEE,QAAQ;IACd,GAAGD;EACL,CAAC,CAAC,CAAC1D,KAAK,CAAEC,GAAQ,IAAK;IAAA,IAAAiE,qBAAA;IACrB,CAAAA,qBAAA,GAAAN,sBAAsB,cAAAM,qBAAA,uBAAtBA,qBAAA,CAAwBnE,MAAM,CAAC,CAAC;IAChC,MAAME,GAAG;EACX,CAAC,CAAC;EACF,CAAAmD,sBAAA,GAAAQ,sBAAsB,cAAAR,sBAAA,uBAAtBA,sBAAA,CAAwBrD,MAAM,CAAC,CAAC;EAChC,OAAO,IAAIzC,YAAY,CAAC;IAAEM,SAAS;IAAEL,GAAG;IAAEC,WAAW;IAAEC,KAAK,EAAEsG;EAAa,CAAC,CAAC;AAC/E;AAEO,eAAeI,eAAeA,CAAA,EAAkB;EACrD,OAAO/G,sBAAO,CAACgH,kBAAkB,CAAC,CAAC;AACrC"}

package/lib/module/NativeRNLlama.js.map

@@ -1 +1 @@
-{"version":3,"names":["TurboModuleRegistry","get"],"sourceRoot":"..\\..\\src","sources":["NativeRNLlama.ts"],"mappings":"AACA,SAASA,mBAAmB,QAAQ,cAAc;AAqKlD,eAAeA,mBAAmB,CAACC,GAAG,CAAO,SAAS,CAAC"}
+{"version":3,"names":["TurboModuleRegistry","get"],"sourceRoot":"..\\..\\src","sources":["NativeRNLlama.ts"],"mappings":"AACA,SAASA,mBAAmB,QAAQ,cAAc;AAwMlD,eAAeA,mBAAmB,CAACC,GAAG,CAAO,SAAS,CAAC"}