gemma4 hg updaloed

Author: mkopcins

packages/react-native-executorch/common/runner/constants.h

@@ -26,7 +26,7 @@ inline constexpr auto kImagePlaceholderId = "image_placeholder_id";
 // Multimodal method name conventions
 inline constexpr auto kVisionEncoderMethod = "vision_encoder";
 inline constexpr auto kAudioEncoderMethod = "audio_encoder";
-inline constexpr auto kTokenEmbeddingMethod = "token_embedding";
+inline constexpr auto kTokenEmbeddingMethod = "text_encoder";
 inline constexpr auto kTextModelMethod = "text_decoder";
 // Absolute ceiling on prefill length (in tokens) and the fallback value used
 // when a PTE doesn't bake `get_max_seq_len`. 2048 matches Gemma4 iter201's

packages/react-native-executorch/common/runner/encoders/audio_encoder.cpp

@@ -98,7 +98,7 @@ Result<EValue> AudioEncoder::encode(const MultimodalInput &input) {
   std::memcpy(padded_wav_.data(), wav.samples.data(),
               static_cast<size_t>(n_valid) * sizeof(float));
 
-  num_blocks_scalar_ = k_blocks;
+  num_blocks_scalar_ = n_valid;
 
   auto wav_tensor = ::executorch::extension::from_blob(
       padded_wav_.data(), {1, static_cast<SizesType>(n_padded)},

packages/react-native-executorch/src/constants/modelRegistry.ts

@@ -488,7 +488,7 @@ export const models = {
     // pick a model by capability ("LLM") rather than by modality.
     lfm2_5_vl_1_6b: base(M.LFM2_5_VL_1_6B_QUANTIZED),
     lfm2_5_vl_450m: base(M.LFM2_5_VL_450M_QUANTIZED),
-    gemma4_e2b: base(M.GEMMA4_E2B_QUANTIZED)
+    gemma4_e2b: base(M.GEMMA4_E2B),
   },
   classification: {
     efficientnet_v2_s: variant(EFFICIENTNET_V2_S_VARIANTS),

packages/react-native-executorch/src/constants/modelUrls.ts

@@ -56,11 +56,9 @@ export const LLAMA3_2_3B_SPINQUANT = {
  */
 export const LLAMA3_2_1B = {
   modelName: 'llama-3.2-1b',
-  modelSource:
-    'http://localhost:9001/Llama3.2-1B-Instruct_vulkan_8da4w_g64_c2048.pte',
-  tokenizerSource: 'http://localhost:9001/gemma4_e2b/llama_tokenizer.json',
-  tokenizerConfigSource:
-    'http://localhost:9001/gemma4_e2b/llama_tokenizer_config.json',
+  modelSource: LLAMA3_2_1B_MODEL,
+  tokenizerSource: LLAMA3_2_TOKENIZER,
+  tokenizerConfigSource: LLAMA3_2_TOKENIZER_CONFIG,
 } as const;
 
 /**
@@ -127,17 +125,21 @@ export const QWEN3_0_6B_QUANTIZED = {
   generationConfig: QWEN3_GENERATION_CONFIG,
 } as const;
 
+// GEMMA 4
+const GEMMA4_E2B_XNNPACK = `${URL_PREFIX}-gemma-4/${VERSION_TAG}/e2b/xnnpack/gemma_4_e2b_xnnpack_8da4w.pte`;
+const GEMMA4_E2B_VULKAN = `${URL_PREFIX}-gemma-4/${VERSION_TAG}/e2b/vulkan/gemma_4_e2b_vulkan_8da4w.pte`;
+const GEMMA4_TOKENIZER = `${URL_PREFIX}-gemma-4/${VERSION_TAG}/e2b/xnnpack/tokenizer.json`;
+const GEMMA4_TOKENIZER_CONFIG = `${URL_PREFIX}-gemma-4/${VERSION_TAG}/e2b/xnnpack/tokenizer_config.json`;
+
 /**
  * @category Models - VLM
  */
-export const GEMMA4_E2B_QUANTIZED = {
-  modelName: 'gemma4-e2b-quantized',
-  modelSource:
-    // 'http://localhost:9001/ptes/iter206_xnnpack_audio_rsqrt_baked_mask_relshift_idx_prefill2048.pte',
-    'http://localhost:9001/experiments/outputs/exp115_mm_s4048_levers.pte',
-  tokenizerSource: 'http://localhost:9001/gemma4_e2b/tokenizer.json',
-  tokenizerConfigSource:
-    'http://localhost:9001/gemma4_e2b/gemma4_tokenizer_config.json',
+export const GEMMA4_E2B = {
+  modelName: 'gemma4-e2b-vulkan',
+  modelSource: GEMMA4_E2B_VULKAN,
+  // 'http://localhost:9001/experiments/outputs/exp115_mm_s4048_levers.pte',
+  tokenizerSource: GEMMA4_TOKENIZER,
+  tokenizerConfigSource: GEMMA4_TOKENIZER_CONFIG,
   capabilities: ['vision', 'audio'],
 } as const;
 

packages/react-native-executorch/src/controllers/LLMController.ts

@@ -319,6 +319,7 @@ export class LLMController {
       this.isGeneratingCallback(true);
       this.nativeModule.reset();
       let response: string;
+      console.log(input);
       if (hasImages || hasAudio) {
         response = await this.nativeModule.generateMultimodal(
           input,