FastVLM-0.5B support?

[YingjieHe]

Thanks for the plugin!!

I am wondering whether it supports FastVLM-0.5B litertlm model?

https://huggingface.co/litert-community/FastVLM-0.5B/tree/main

I had the following issue when try to use the fastvlm, from the log, it said the model was active

flutter: ℹ️  Model already installed: FastVLM-0.5B.litertlm (skipping download)
flutter: ✅ Set active inference model: FastVLM-0.5B
flutter: ✅ Inference model installed and set as active: FastVLM-0.5B

but then got

ERROR: The model is not a valid Flatbuffer buffer
E0000 00:00:1774586045.229607 1122263 model_data.cc:424] INTERNAL: RET_CHECK failure (third_party/odml/infra/genai/inference/utils/llm_utils/model_data.cc:424) model Error building tflite model.
Stack trace:
    @        0x104700ff0  util::StatusBuilder::ConditionallyLog()
    @        0x1047006c8  util::StatusBuilder::CreateStatusAndConditionallyLog()
    @        0x1040be2a0  absl::internal_statusor::StatusOrData<>::StatusOrData<>()
    @        0x1040bcc68  odml::infra::llm_utils::ModelData::Create()
    @        0x1047fa508  odml::infra::llm_utils::DeriveLlmModelSettingsStruct()
    @        0x1047cc2dc  odml::infra::LlmInferenceEngine::CreateLlmInferenceEngine()
    @        0x1047cc1b4  odml::infra::LlmInferenceEngine::CreateLlmInferenceEngine()
    @        0x1047c7d28  LlmInferenceEngine_CreateEngine
    @        0x104374788  $s19MediaPipeTasksGenAI13LlmTaskRunnerC13modelSettingsACSo0f5ModelJ0a_tKcfc
    @        0x1043746fc  $s19MediaPipeTasksGenAI13LlmTaskRunnerC13modelSettingsACSo0f5ModelJ0a_tKcfC
    @        0x10436e518  $s19MediaPipeTasksGenAI12LlmInferenceC7optionsA2C7OptionsC_tKcfcAA0F10TaskRunnerCSPys4Int8VGKXEfU_AhKKXEfU_AHSrySiGzKXEfU_AhKKXEfU_AhKKXEfU_
    @        0x1043708f4  $s19MediaPipeTasksGenAI12LlmInferenceC7optionsA2C7OptionsC_tKcfcAA0F10TaskRunnerCSPys4Int8VGKXEfU_AhKKXEfU_AHSrySiGzKXEfU_AhKKXEfU_TA
    @        0x10436e5c8  $sSS11withCStringyxxSPys4Int8VGKXEKlF19MediaPipeTasksGenAI13LlmTaskRunnerC_Tg5
    @        0x10436e27c  $s19MediaPipeTasksGenAI12LlmInferenceC7optionsA2C7OptionsC_tKcfcAA0F10TaskRunnerCSPys4Int8VGKXEfU_AhKKXEfU_
    @        0x1043707ec  $s19MediaPipeTasksGenAI12LlmInferenceC7optionsA2C7OptionsC_tKcfcAA0F10TaskRunnerCSPys4Int8VGKXEfU_AhKKXEfU_TA
    @        0x18938ed24  $ss11_StringGutsV16_slowWithCStringyxxSPys4Int8VGKXEKlF
    @        0x10436e61c  $sSS11withCStringyxxSPys4Int8VGKXEKlF19MediaPipeTasksGenAI13LlmTaskRunnerC_Tg5
    @        0x10436e170  $s19MediaPipeTasksGenAI12LlmInferenceC7optionsA2C7OptionsC_tKcfcAA0F10TaskRunnerCSPys4Int8VGKXEfU_TA
    @        0x18938ed24  $ss11_StringGutsV16_slowWithCStringyxxSPys4Int8VGKXEKlF
    @        0x10436e61c  $sSS11withCStringyxxSPys4Int8VGKXEKlF19MediaPipeTasksGenAI13LlmTaskRunnerC_Tg5
    @        0x10436e018  $s19MediaPipeTasksGenAI12LlmInferenceC7optionsA2C7OptionsC_tKcfc
    @        0x10436dd2c  $s19MediaPipeTasksGenAI12LlmInferenceC7optionsA2C7OptionsC_tKcfC
    @        0x10523b9fc  $s13flutter_gemma14InferenceModelV9modelPath9maxTokens18supportedLoraRanks0G9NumImagesACSS_SiSaySiGSgSitKcfC
    @        0x1052240d8  $s13flutter_gemma19PlatformServiceImplC11createModel9maxTokens9modelPath9loraRanks16preferredBackend0H9NumImages12supportAudio10completionys5Int64V_SSSayAMGSgAA09PreferredO0OSgAMSgSbSgys6ResultOyyts5Error_pGctFyyYbcfU_
    @        0x1052244d8  $s13flutter_gemma19PlatformServiceImplC11createModel9maxTokens9modelPath9loraRanks16preferredBackend0H9NumImages12supportAudio10completionys5Int64V_SSSayAMGSgAA09PreferredO0OSgAMSgSbSgys6ResultOyyts5Error_pGctFyyYbcfU_TA
    @        0x10435b0f0  $sIegh_IeyBh_TR
    @        0x1c4d00adc  904D48A3-D99E-3962-BFA9-C3DFB66BBA83
    @        0x1c4d1a7fc  904D48A3-D99E-3962-BFA9-C3DFB66BBA83
    @        0x1c4d3741c  904D48A3-D99E-3962-BFA9-C3DFB66BBA83
    @        0x1c4d130bc  904D48A3-D99E-3962-BFA9-C3DFB66BBA83
    @        0x1c4d136fc  904D48A3-D99E-3962-BFA9-C3DFB66BBA83
    @        0x1e8b4d37c  _pthread_wqthread
    @        0x1e8b4c8c0  start_wqthread
[ERROR:flutter/runtime/dart_vm_initializer.cc(40)] Unhandled Exception: PlatformException(failedToInitializeEngine(Optional("INTERNAL: RET_CHECK failure (third_party/odml/infra/genai/inference/utils/llm_utils/model_data.cc:424) model Error building tflite model.\n=== Source Location Trace: === \nthird_party/odml/infra/genai/inference/utils/llm_utils/model_data.cc:424\nthird_party/odml/infra/genai/inference/utils/llm_utils/config_utils.cc:558\nthird_party/odml/infra/genai/inference/llm_engine.cc:2132\n")), GenAiInferenceError, Stacktrace: ["0   Runner.debug.dylib                  0x000000010523e31c $s13flutter_gemma9wrapError33_AC1DF4B16F5474DCF2184B6087243948LLySayypSgGypF + 1856", "1   Runner.debug.dylib                  0x0000000105245b70 $s13flutter_gemma20PlatformServiceSetupC5setUp15binaryMessenger3api20messageChannelSuffixySo013FlutterBinaryI0_p_AA0cD0_pSgSStFZyypSg_yAKctcfU_ys6ResultOyyts5Error_pGcfU_ + 332", "2   Runner.debug.dylib                  0x00000001052245c4 $s13flutter_gemma19PlatformServiceImplC11createModel9maxTokens9modelPath9loraRanks16preferredBackend0H9NumImages12supportAudio10completionys5Int64V_SSSayAMGSgAA09PreferredO0OSgAMSgSbSgys6ResultOyyts5Error_pGctFyyYbcfU_yyScMYccfU0_ + 96", "3   Runner.debug.dylib                  0x0000000104359b1c $sIeg_IeyB_TR + 48", "4   libdispatch.dylib                   0x00000001c4d00adc 904D48A3-D99E-3962-BFA9-C3DFB66BBA83 + 6876", "5   libdispatch.dylib                   0x00000001c4d1a7fc 904D48A3-D99E-3962-BFA9-C3DFB66BBA83 + 112636", "6   libdispatch.dylib                   0x00000001c4d37b10 904D48A3-D99E-3962-BFA9-C3DFB66BBA83 + 232208", "7   libdispatch.dylib                   0x00000001c4d0fec8 904D48A3-D99E-3962-BFA9-C3DFB66BBA83 + 69320", "8   libdispatch.dylib                   0x00000001c4d0fe04 _dispatch_main_queue_callback_4CF + 44", "9   CoreFoundation                      0x000000018c0412b4 2F32D384-4637-3018-843E-4FC875B865C4 + 434868", "10  CoreFoundation                      0x000000018bff4b3c 2F32D384-4637-3018-843E-4FC875B865C4 + 121660", "11  CoreFoundation                      0x000000018bff3a6c 2F32D384-4637-3018-843E-4FC875B865C4 + 117356", "12  GraphicsServices                    0x0000000230991498 GSEventRunModal + 120", "13  UIKitCore                           0x0000000191aa3df8 C768F963-A0CC-3F5C-A1D3-2E06D53A2381 + 646648", "14  UIKitCore                           0x0000000191a4ce54 UIApplicationMain + 336", "15  UIKitCore                           0x0000000191b78938 C768F963-A0CC-3F5C-A1D3-2E06D53A2381 + 1517880", "16  Runner.debug.dylib                  0x0000000104355cf8 $sSo21UIApplicationDelegateP5UIKitE4mainyyFZ + 128", "17  Runner.debug.dylib                  0x0000000104355c68 $s6Runner11AppDelegateC5$mainyyFZ + 44", "18  Runner.debug.dylib                  0x0000000104355d74 __debug_main_executable_dylib_entry_point + 28", "19  dyld                                0x0000000188fcee28 8ACDB580-8AB7-38C0-A586-E667ADB1C11C + 20008"], null)
#0      PlatformService.createModel (package:flutter_gemma/pigeon.g.dart:202:7)
<asynchronous suspension>
#1      FlutterGemmaMobile.createModel (package:flutter_gemma/mobile/flutter_gemma_mobile.dart:313:7)
<asynchronous suspension>
#2      FlutterGemma.getActiveModel (package:flutter_gemma/core/api/flutter_gemma.dart:257:12)

Other info:

IOS: 26.3.1

flutter_gemma: 0.12.6

[DenisovAV]

Hi @YingjieHe!

FastVLM-0.5B is only available in .litertlm format, which is not supported on iOS yet. iOS uses MediaPipe SDK (.task format), and there's no .task version of FastVLM available.

Please try it on Android or Desktop (macOS/Windows/Linux) — it works there with full vision support.

I'll add iOS support for .litertlm models as soon as Google open-sources the LiteRT-LM SDK for iOS (currently it's only available for Android and JVM/Desktop).