Merge pull request #32 from NexaAI/ethan/bugfix

README.md

@@ -97,7 +97,7 @@ Example:
 `docker run -v /home/ubuntu/.cache/nexa/hub/official:/model -it nexa4ai/sdk:latest nexa gen-text /model/Phi-3-mini-128k-instruct/q4_0.gguf`
 
 will create an interactive session with text generation
-```
+
 
 ## Nexa CLI commands
 

nexa/gguf/nexa_inference_image.py

@@ -15,7 +15,9 @@
     DEFAULT_IMG_GEN_PARAMS_LCM,
     DEFAULT_IMG_GEN_PARAMS_TURBO,
 )
-from nexa.utils import SpinningCursorAnimation, nexa_prompt, suppress_stdout_stderr
+from nexa.utils import SpinningCursorAnimation, nexa_prompt
+from nexa.gguf.llama._utils_transformers import suppress_stdout_stderr
+
 from streamlit.web import cli as stcli
 
 logging.basicConfig(
@@ -142,7 +144,7 @@ def txt2img(self,
         )
         return images
 
-    def run_txt2img(self):
+    def loop_txt2img(self):
         while True:
             try:
                 prompt = nexa_prompt("Enter your prompt: ")
@@ -313,4 +315,4 @@ def run_streamlit(self, model_path: str):
         if args.img2img:
             inference.run_img2img()
         else:
-            inference.run_txt2img()
+            inference.loop_txt2img()

nexa/gguf/nexa_inference_text.py

@@ -14,7 +14,9 @@
 )
 from nexa.general import pull_model
 from nexa.gguf.lib_utils import is_gpu_available
-from nexa.utils import SpinningCursorAnimation, nexa_prompt, suppress_stdout_stderr
+from nexa.utils import SpinningCursorAnimation, nexa_prompt
+from nexa.gguf.llama._utils_transformers import suppress_stdout_stderr
+
 
 logging.basicConfig(
     level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s"

nexa/gguf/nexa_inference_vlm.py

@@ -24,7 +24,8 @@
     Llava16ChatHandler,
     NanoLlavaChatHandler,
 )
-from nexa.utils import SpinningCursorAnimation, nexa_prompt, suppress_stdout_stderr
+from nexa.utils import SpinningCursorAnimation, nexa_prompt
+from nexa.gguf.llama._utils_transformers import suppress_stdout_stderr
 
 logging.basicConfig(
     level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s"

nexa/gguf/nexa_inference_voice.py

@@ -11,7 +11,9 @@
     NEXA_RUN_MODEL_MAP_VOICE,
 )
 from nexa.general import pull_model
-from nexa.utils import nexa_prompt, SpinningCursorAnimation, suppress_stdout_stderr
+from nexa.utils import nexa_prompt, SpinningCursorAnimation
+from nexa.gguf.llama._utils_transformers import suppress_stdout_stderr
+
 
 logging.basicConfig(level=logging.INFO)
 
@@ -75,6 +77,7 @@ def _load_model(self):
 
         logging.debug(f"Loading model from: {self.downloaded_path}")
         with suppress_stdout_stderr():
+            os.environ["KMP_DUPLICATE_LIB_OK"] = "TRUE"
             self.model = WhisperModel(
                 self.downloaded_path,
                 device="cpu",

nexa/onnx/nexa_inference_image.py

@@ -15,7 +15,7 @@
 
 from nexa.constants import EXIT_REMINDER, NEXA_RUN_MODEL_MAP_ONNX
 from nexa.general import pull_model
-from nexa.utils import nexa_prompt
+from nexa.utils import nexa_prompt, SpinningCursorAnimation
 
 logging.basicConfig(level=logging.INFO)
 
@@ -73,6 +73,7 @@ def run(self):
         # Step 3: Enter dialogue mode
         self._dialogue_mode()
 
+    @SpinningCursorAnimation()
     def _load_model(self, model_path):
         """
         Load the model from the given model path using the appropriate pipeline.
@@ -147,7 +148,7 @@ def generate_images(self, prompt, negative_prompt):
         images = self.pipeline(**pipeline_kwargs).images
         return images
 
-            
+
 
     def _save_images(self, images):
         """

nexa/onnx/nexa_inference_text.py

@@ -10,7 +10,7 @@
 
 from nexa.constants import NEXA_RUN_MODEL_MAP_ONNX
 from nexa.general import pull_model
-from nexa.utils import nexa_prompt
+from nexa.utils import nexa_prompt, SpinningCursorAnimation
 
 logging.basicConfig(level=logging.INFO)
 
@@ -51,6 +51,7 @@ def __init__(self, model_path, **kwargs):
         self.timings = kwargs.get("timings", False)
         self.device = "cpu"
 
+    @SpinningCursorAnimation()
     def _load_model_and_tokenizer(self) -> Tuple[Any, Any, Any, bool]:
         logging.debug(f"Loading model from {self.downloaded_onnx_folder}")
         start_time = time.time()

nexa/utils.py

@@ -10,9 +10,6 @@
 from prompt_toolkit.styles import Style
 
 from nexa.constants import EXIT_COMMANDS, EXIT_REMINDER
-from nexa.gguf.llama._utils_transformers import (
-    suppress_stdout_stderr,
-)  # re-import, don't comment out
 
 
 def is_package_installed(package_name: str) -> bool:

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "scikit_build_core.build"
 
 [project]
 name = "nexaai"
-version = "0.0.2.dev"
+version = "0.0.5"
 description = "Nexa AI SDK"
 readme = "README.md"
 license = { text = "MIT" }
@@ -16,8 +16,6 @@ dependencies = [
     "diskcache>=5.6.1",
     "jinja2>=2.11.3",
     "librosa>=0.8.0",
-    "boto3>=1.34.148",
-    "botocore>=1.34.148",
     "fastapi",
     "uvicorn",
     "pydantic",
@@ -38,7 +36,7 @@ classifiers = [
 [project.optional-dependencies]
 onnx = [
     "librosa",
-    "optimum[onnxruntime]>=1.7.3",  # for CPU version
+    "optimum[onnxruntime]",  # for CPU version
     "diffusers",  # required for image generation
     "optuna",
     "pydantic",