bug: `Segmentation fault` while running flux.1-dev with sd plugin

Summary

Stable-diffusion plugin causes Segmentation fault (core dumped) while running flux.1-dev. The following shows the complete log messages, including the error:

WASMEDGE_PLUGIN_PATH=/data/sam/sd/plugin wasmedge --dir .:. sd-api-server.wasm --model-name flux1-dev --diffusion-model flux1-dev-Q4_0.gguf --vae ae.safetensors --clip-l clip_l.safetensors --t5xxl t5xxl-Q8_0.gguf
[2024-09-25 07:14:41.011] [info] sd_api_server in src/main.rs:79: server version: 0.1.4
[2024-09-25 07:14:41.011] [info] sd_api_server in src/main.rs:87: model_name: flux1-dev
[2024-09-25 07:14:41.011] [info] sd_api_server in src/main.rs:103: diffusion model: flux1-dev-Q4_0.gguf
[2024-09-25 07:14:41.011] [info] sd_api_server in src/main.rs:111: vae: ae.safetensors
[2024-09-25 07:14:41.011] [info] sd_api_server in src/main.rs:119: clip_l: clip_l.safetensors
[2024-09-25 07:14:41.011] [info] sd_api_server in src/main.rs:127: t5xxl: t5xxl-Q8_0.gguf
[2024-09-25 07:14:41.011] [info] sd_api_server in src/main.rs:135: lora_model_dir: 
[2024-09-25 07:14:41.012] [info] sd_api_server in src/main.rs:136: threads: -1
[2024-09-25 07:14:41.012] [info] llama_core in /home/runner/.cargo/registry/src/index.crates.io-6f17d22bba15001f/llama-core-0.18.0/src/lib.rs:900: Initializing the stable diffusion context with the standalone diffusion model
[2024-09-25 07:14:41.012] [info] llama_core in /home/runner/.cargo/registry/src/index.crates.io-6f17d22bba15001f/llama-core-0.18.0/src/lib.rs:976: The stable diffusion text-to-image context has been initialized
[2024-09-25 07:14:41.012] [info] llama_core in /home/runner/.cargo/registry/src/index.crates.io-6f17d22bba15001f/llama-core-0.18.0/src/lib.rs:1052: The stable diffusion image-to-image context has been initialized
[2024-09-25 07:14:41.012] [info] sd_api_server in src/main.rs:168: Listening on 0.0.0.0:8080
[2024-09-25 07:19:40.167] [info] sd_api_server in src/main.rs:162: remote_addr: 0.0.0.0:42844, local_addr: 0.0.0.0:8080
[2024-09-25 07:19:40.168] [info] sd_api_server in src/main.rs:199: method: POST, endpoint: /v1/images/generations, http_version: HTTP/1.1, content-length: 466
[2024-09-25 07:19:40.168] [info] sd_api_server::backend::sd in src/backend/sd.rs:18: Handling the coming image generation request
[2024-09-25 07:19:40.168] [info] sd_api_server::backend::sd in src/backend/sd.rs:42: Prepare the image generation request.
[2024-09-25 07:19:40.168] [info] sd_api_server::backend::sd in src/backend/sd.rs:75: user: imgen-29a8cbbd-fc7e-4cc4-a40c-00141f5fc96b
[2024-09-25 07:19:40.169] [info] llama_core::images in /home/runner/.cargo/registry/src/index.crates.io-6f17d22bba15001f/llama-core-0.18.0/src/images.rs:21: Processing the image generation request.
[2024-09-25 07:19:40.169] [info] llama_core::images in /home/runner/.cargo/registry/src/index.crates.io-6f17d22bba15001f/llama-core-0.18.0/src/images.rs:45: create computation context.
[2024-09-25 07:19:40.169] [info] llama_core::images in /home/runner/.cargo/registry/src/index.crates.io-6f17d22bba15001f/llama-core-0.18.0/src/images.rs:48: config of sd: StableDiffusion { task: TextToImage, model_path: "", clip_l_path: "clip_l.safetensors", t5xxl_path: "t5xxl-Q8_0.gguf", diffusion_model_path: "flux1-dev-Q4_0.gguf", vae_path: "ae.safetensors", taesd_path: "", control_net_path: "", lora_model_dir: "", embed_dir: "", id_embed_dir: "", vae_decode_only: true, vae_tiling: false, n_threads: -1, wtype: SdTypeCount, rng_type: StdDefaultRng, schedule: DEFAULT, clip_on_cpu: false, control_net_cpu: false, vae_on_cpu: false }
[2024-09-25 07:19:40.169] [info] [WasmEdge-StableDiffusion] Create context.
WARNING: Behavior may be unexpected when allocating 0 bytes for ggml_calloc!
WARNING: Behavior may be unexpected when allocating 0 bytes for ggml_calloc!
[2024-09-25 07:20:01.323] [info] llama_core::images in /home/runner/.cargo/registry/src/index.crates.io-6f17d22bba15001f/llama-core-0.18.0/src/images.rs:77: prompt: Capture a street food vendor in Tokyo at night, shot with a wide-angle lens (24mm) at f/1.8. Use a shallow depth of field to focus on the vendor’s hands preparing takoyaki, with the glowing street signs and bustling crowd blurred in the background. High ISO setting to capture the ambient light, giving the image a slight grain for a cinematic feel.
[2024-09-25 07:20:01.323] [info] llama_core::images in /home/runner/.cargo/registry/src/index.crates.io-6f17d22bba15001f/llama-core-0.18.0/src/images.rs:84: negative prompt: 
[2024-09-25 07:20:01.323] [info] llama_core::images in /home/runner/.cargo/registry/src/index.crates.io-6f17d22bba15001f/llama-core-0.18.0/src/images.rs:91: cfg_scale: 1
[2024-09-25 07:20:01.323] [info] llama_core::images in /home/runner/.cargo/registry/src/index.crates.io-6f17d22bba15001f/llama-core-0.18.0/src/images.rs:98: sample_method: euler
[2024-09-25 07:20:01.323] [info] llama_core::images in /home/runner/.cargo/registry/src/index.crates.io-6f17d22bba15001f/llama-core-0.18.0/src/images.rs:139: steps: 20
[2024-09-25 07:20:01.323] [info] llama_core::images in /home/runner/.cargo/registry/src/index.crates.io-6f17d22bba15001f/llama-core-0.18.0/src/images.rs:147: height: 512, width: 512
[2024-09-25 07:20:01.323] [info] llama_core::images in /home/runner/.cargo/registry/src/index.crates.io-6f17d22bba15001f/llama-core-0.18.0/src/images.rs:151: generate image
[2024-09-25 07:20:01.324] [info] [WasmEdge-StableDiffusion] Start to generate image.
Segmentation fault (core dumped)

Current State

No response

Expected State

No response

Reproduction steps

Install WasmEdge v0.14.1

curl -sSf https://raw.githubusercontent.com/WasmEdge/WasmEdge/master/utils/install_v2.sh | bash -s -- -v 0.14.1

Deploy wasmedge_stablediffusion

Download FLUX.1-dev model

# download the model
curl -LO https://huggingface.co/second-state/FLUX.1-dev-GGUF/resolve/main/flux1-dev-Q4_0.gguf

# download vae file
curl -LO https://huggingface.co/second-state/FLUX.1-dev-GGUF/resolve/main/ae.safetensors

# download clip_l encoder
curl -LO https://huggingface.co/second-state/FLUX.1-dev-GGUF/resolve/main/clip_l.safetensors

# download t5xxl encoder
curl -LO https://huggingface.co/second-state/FLUX.1-dev-GGUF/resolve/main/t5xxl-Q8_0.gguf

# download lora model
mkdir lora-models
curl -L https://huggingface.co/XLabs-AI/flux-lora-collection/resolve/main/realism_lora_comfy_converted.safetensors -o lora-models/realism_lora_comfy_converted.safetensors

Download sd-api-server.wasm

curl -LO https://github.com/LlamaEdge/sd-api-server/releases/latest/download/sd-api-server.wasm

Start server

wasmedge --dir .:. sd-api-server.wasm \
  --dir lora-models:lora-models
  --model-name flux1-dev \
  --diffusion-model flux1-dev-Q4_0.gguf \
  --vae ae.safetensors \
  --clip-l clip_l.safetensors \
  --t5xxl t5xxl-Q8_0.gguf

Send a request for image generation

curl --location 'http://localhost:8080/v1/images/generations' \
--header 'Content-Type: application/json' \
--data '{
    "model": "flux1-dev",
    "prompt": "Capture a street food vendor in Tokyo at night, shot with a wide-angle lens (24mm) at f/1.8. Use a shallow depth of field to focus on the vendor’s hands preparing takoyaki, with the glowing street signs and bustling crowd blurred in the background. High ISO setting to capture the ambient light, giving the image a slight grain for a cinematic feel.",
    "cfg_scale": 1.0,
    "sample_method": "euler",
    "steps": 20
}'

Screenshots

No response

Any logs you want to share for showing the specific issue

No response

Components

Others

WasmEdge Version or Commit you used

0.14.1

Operating system information

macOS 14.6.1, ubuntu-2204

Hardware Architecture

Apple M1 Pro, x86_64

Compiler flags and options

No response

WasmEdge / WasmEdge

bug: `Segmentation fault` while running flux.1-dev with sd plugin #3794

Summary

Current State

Expected State

Reproduction steps

Screenshots

Any logs you want to share for showing the specific issue

Components

WasmEdge Version or Commit you used

Operating system information

Hardware Architecture

Compiler flags and options