Add quantized .en models & fix test URLs

Author: IgorSwat

packages/react-native-executorch/common/rnexecutorch/tests/run_tests.sh

@@ -54,9 +54,8 @@ MODELS=(
   "all-MiniLM-L6-v2_xnnpack.pte|https://huggingface.co/software-mansion/react-native-executorch-all-MiniLM-L6-v2/resolve/v0.6.0/all-MiniLM-L6-v2_xnnpack.pte"
   "tokenizer.json|https://huggingface.co/software-mansion/react-native-executorch-all-MiniLM-L6-v2/resolve/v0.6.0/tokenizer.json"
   "fsmn-vad_xnnpack.pte|https://huggingface.co/software-mansion/react-native-executorch-fsmn-vad/resolve/main/xnnpack/fsmn-vad_xnnpack.pte"
-  "whisper_tiny_en_encoder_xnnpack.pte|https://huggingface.co/software-mansion/react-native-executorch-whisper-tiny.en/resolve/v0.7.0/xnnpack/whisper_tiny_en_encoder_xnnpack.pte"
-  "whisper_tiny_en_decoder_xnnpack.pte|https://huggingface.co/software-mansion/react-native-executorch-whisper-tiny.en/resolve/v0.7.0/xnnpack/whisper_tiny_en_decoder_xnnpack.pte"
-  "whisper_tokenizer.json|https://huggingface.co/software-mansion/react-native-executorch-whisper-tiny.en/resolve/v0.6.0/tokenizer.json"
+  "whisper_tiny_en_encoder_xnnpack.pte|https://huggingface.co/software-mansion/react-native-executorch-whisper-tiny.en/resolve/v0.8.0/xnnpack/whisper_tiny_en_xnnpack.pte"
+  "whisper_tokenizer.json|https://huggingface.co/software-mansion/react-native-executorch-whisper-tiny.en/resolve/v0.8.0/tokenizer.json"
   "smolLm2_135M_8da4w.pte|https://huggingface.co/software-mansion/react-native-executorch-smolLm-2/resolve/v0.6.0/smolLm-2-135M/quantized/smolLm2_135M_8da4w.pte"
   "smollm_tokenizer.json|https://huggingface.co/software-mansion/react-native-executorch-smolLm-2/resolve/v0.6.0/tokenizer.json"
   "deeplabV3_xnnpack_fp32.pte|https://huggingface.co/software-mansion/react-native-executorch-deeplab-v3/resolve/v0.6.0/xnnpack/deeplabV3_xnnpack_fp32.pte"

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

@@ -454,12 +454,21 @@ export const STYLE_TRANSFER_UDNIE = {
 const WHISPER_TINY_EN_TOKENIZER = `${URL_PREFIX}-whisper-tiny.en/${VERSION_TAG}/tokenizer.json`;
 const WHISPER_TINY_EN_MODEL = `${URL_PREFIX}-whisper-tiny.en/${NEXT_VERSION_TAG}/xnnpack/whisper_tiny_en_xnnpack.pte`;
 
-const WHISPER_BASE_EN_TOKENIZER = `${URL_PREFIX}-whisper-base.en/${VERSION_TAG}/tokenizer.json`;
+const WHISPER_TINY_EN_QUANTIZED_TOKENIZER = `${URL_PREFIX}-whisper-tiny-quantized.en/${NEXT_VERSION_TAG}/tokenizer.json`;
+const WHISPER_TINY_EN_QUANTIZED_MODEL = `${URL_PREFIX}-whisper-tiny-quantized.en/${NEXT_VERSION_TAG}/xnnpack/whisper_tiny_en_quantized_xnnpack.pte`;
+
+const WHISPER_BASE_EN_TOKENIZER = `${URL_PREFIX}-whisper-base.en/${NEXT_VERSION_TAG}/tokenizer.json`;
 const WHISPER_BASE_EN_MODEL = `${URL_PREFIX}-whisper-base.en/${NEXT_VERSION_TAG}/xnnpack/whisper_base_en_xnnpack.pte`;
 
-const WHISPER_SMALL_EN_TOKENIZER = `${URL_PREFIX}-whisper-small.en/${VERSION_TAG}/tokenizer.json`;
+const WHISPER_BASE_EN_QUANTIZED_TOKENIZER = `${URL_PREFIX}-whisper-base-quantized.en/${VERSION_TAG}/tokenizer.json`;
+const WHISPER_BASE_EN_QUANTIZED_MODEL = `${URL_PREFIX}-whisper-base-quantized.en/${NEXT_VERSION_TAG}/xnnpack/whisper_base_en_quantized_xnnpack.pte`;
+
+const WHISPER_SMALL_EN_TOKENIZER = `${URL_PREFIX}-whisper-small.en/${NEXT_VERSION_TAG}/tokenizer.json`;
 const WHISPER_SMALL_EN_MODEL = `${URL_PREFIX}-whisper-small.en/${NEXT_VERSION_TAG}/xnnpack/whisper_small_en_xnnpack.pte`;
 
+const WHISPER_SMALL_EN_QUANTIZED_TOKENIZER = `${URL_PREFIX}-whisper-small-quantized.en/${VERSION_TAG}/tokenizer.json`;
+const WHISPER_SMALL_EN_QUANTIZED_MODEL = `${URL_PREFIX}-whisper-small-quantized.en/${NEXT_VERSION_TAG}/xnnpack/whisper_small_en_quantized_xnnpack.pte`;
+
 const WHISPER_TINY_TOKENIZER = `${URL_PREFIX}-whisper-tiny/${VERSION_TAG}/tokenizer.json`;
 const WHISPER_TINY_MODEL = `${URL_PREFIX}-whisper-tiny/${NEXT_VERSION_TAG}/xnnpack/whisper_tiny_xnnpack.pte`;
 
@@ -479,6 +488,16 @@ export const WHISPER_TINY_EN = {
   tokenizerSource: WHISPER_TINY_EN_TOKENIZER,
 };
 
+/**
+ * @category Models - Speech To Text
+ */
+export const WHISPER_TINY_EN_QUANTIZED = {
+  type: 'whisper' as const,
+  isMultilingual: false,
+  modelSource: WHISPER_TINY_EN_QUANTIZED_MODEL,
+  tokenizerSource: WHISPER_TINY_EN_QUANTIZED_TOKENIZER,
+};
+
 /**
  * @category Models - Speech To Text
  */
@@ -489,6 +508,16 @@ export const WHISPER_BASE_EN = {
   tokenizerSource: WHISPER_BASE_EN_TOKENIZER,
 };
 
+/**
+ * @category Models - Speech To Text
+ */
+export const WHISPER_BASE_EN_QUANTIZED = {
+  type: 'whisper' as const,
+  isMultilingual: false,
+  modelSource: WHISPER_BASE_EN_QUANTIZED_MODEL,
+  tokenizerSource: WHISPER_BASE_EN_QUANTIZED_TOKENIZER,
+};
+
 /**
  * @category Models - Speech To Text
  */
@@ -499,6 +528,16 @@ export const WHISPER_SMALL_EN = {
   tokenizerSource: WHISPER_SMALL_EN_TOKENIZER,
 };
 
+/**
+ * @category Models - Speech To Text
+ */
+export const WHISPER_SMALL_EN_QUANTIZED = {
+  type: 'whisper' as const,
+  isMultilingual: false,
+  modelSource: WHISPER_SMALL_EN_QUANTIZED_MODEL,
+  tokenizerSource: WHISPER_SMALL_EN_QUANTIZED_TOKENIZER,
+};
+
 /**
  * @category Models - Speech To Text
  */