Differences of v0.3 docker images and connection refused from API call

[lililolo0927]

In docker hub for KTransformers, I've found several kinds of tags for the new version(v0.3) images such as v0.3-AVX2, lastest-AVX2, v0.3-NATIVE, latest-NATIVE, v0.3-FANCY, lastest-FANCY, v0.3-AVX512, latest-AVX512 .

What exactly is the difference between these tags?

I'm not sure which image I should download.
Currently, I downloaded approachingai/ktransformers:v0.3-AVX2 and tried running the Qwen3-30B-A3B model.
There are no errors, but I can't access through the API.

When I run curl -X 'GET' 'http://localhost:10002/api/tags', I get a "connection refused" response.
I've confirmed that all the necessary ports inside the container are open, but API access still doesn't work.
The command I used is as follows:

PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True python ktransformers/server/main.py \
  --port 10002 \
  --host 0.0.0.0 \
  --architectures Qwen3MoeForCausalLM \
  --model_path Qwen/Qwen3-30B-A3B  \
  --model_name Qwen3-30B-A3B \
  --gguf_path /models/gguf/Qwen3-30B-A3B-GGUF \
  --optimize_config_path ktransformers/optimize/optimize_rules/Qwen3Moe-serve.yaml  \
  --backend_type balance_serve

In addition , I tried the option
--optimize_config_path ktransformers/optimize/optimize_rules/Qwen3Moe-serve-amx.yaml but it returned me an error even though my machine support AMX.

Injecting cache as default
Injecting lm_head as ktransformers.operators.linear . KTransformersLinear
loading token_embd.weight to cpu
loading token_embd.weight with CPU
loading blk.0.attn_q_norm.weight to cuda
loading blk.0.attn_k_norm.weight to cuda
Process SpawnProcess-1:
Traceback (most recent call last):
  File "/opt/conda/lib/python3.11/multiprocessing/process.py", line 314, in _bootstrap
    self.run()
  File "/opt/conda/lib/python3.11/multiprocessing/process.py", line 108, in run
    self._target(*self._args, **self._kwargs)
  File "/opt/conda/lib/python3.11/site-packages/ktransformers/server/backend/interfaces/balance_serve.py", line 277run_engine
    engine = Engine(args, token_queue, broadcast_endpoint, kvcache_event)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/conda/lib/python3.11/site-packages/ktransformers/server/backend/interfaces/balance_serve.py", line 181__init__
    optimize_and_load_gguf(self.model, optimize_config_path, gguf_path, config)
  File "/opt/conda/lib/python3.11/site-packages/ktransformers/optimize/optimize.py", line 131, in optimize_and_load
    load_weights(module, gguf_loader)
  File "/opt/conda/lib/python3.11/site-packages/ktransformers/util/utils.py", line 125, in load_weights
    load_weights(child, gguf_loader, prefix+name+".")
  File "/opt/conda/lib/python3.11/site-packages/ktransformers/util/utils.py", line 127, in load_weights
    module.load()
  File "/opt/conda/lib/python3.11/site-packages/ktransformers/operators/base_operator.py", line 63, in load
    utils.load_weights(child, self.gguf_loader, self.key+".")
  File "/opt/conda/lib/python3.11/site-packages/ktransformers/util/utils.py", line 125, in load_weights
    load_weights(child, gguf_loader, prefix+name+".")
  File "/opt/conda/lib/python3.11/site-packages/ktransformers/util/utils.py", line 125, in load_weights
    load_weights(child, gguf_loader, prefix+name+".")
  File "/opt/conda/lib/python3.11/site-packages/ktransformers/util/utils.py", line 125, in load_weights
    load_weights(child, gguf_loader, prefix+name+".")
  File "/opt/conda/lib/python3.11/site-packages/ktransformers/util/utils.py", line 127, in load_weights
    module.load()
  File "/opt/conda/lib/python3.11/site-packages/ktransformers/operators/base_operator.py", line 63, in load
    utils.load_weights(child, self.gguf_loader, self.key+".")
  File "/opt/conda/lib/python3.11/site-packages/ktransformers/util/utils.py", line 125, in load_weights
    load_weights(child, gguf_loader, prefix+name+".")
  File "/opt/conda/lib/python3.11/site-packages/ktransformers/util/utils.py", line 127, in load_weights
    module.load()
  File "/opt/conda/lib/python3.11/site-packages/ktransformers/operators/experts.py", line 1231, in load
    self.generate_experts.load(w, warmup=warmup)
  File "/opt/conda/lib/python3.11/site-packages/ktransformers/operators/experts.py", line 188, in load
    from cpuinfer_ext.moe import AMX_MOEConfig, AMXBF16_MOE
ImportError: cannot import name 'AMX_MOEConfig' from 'cpuinfer_ext.moe' (unknown location)
^Z

I would appreciate your opinions on this.
Thanks in advance.