Update model

Dockerfile

@@ -1,6 +1,9 @@
 FROM python:3.10-slim-bookworm
 
+# Добавляем build-essential и cmake для правильной сборки llama-cpp из исходников под CPU сервера
 RUN apt-get update && apt-get install -y --no-install-recommends \
+    build-essential \
+    cmake \
     libopenblas0 \
     libgomp1 \
     && rm -rf /var/lib/apt/lists/*
@@ -15,11 +18,10 @@ WORKDIR /app
 
 RUN pip install --no-cache-dir --upgrade pip
 
-COPY --chown=user build /app/build
-
-RUN pip install --force-reinstall /app/build/llama_cpp_python-*.whl
-
 COPY --chown=user requirements.txt .
+
+# Собираем llama-cpp-python с оптимизацией под процессор (OpenBLAS)
+ENV CMAKE_ARGS="-DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS"
 RUN pip install --no-cache-dir -r requirements.txt
 
 COPY --chown=user . .

build/diskcache-5.6.3-py3-none-any.whl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:5e31b2d5fbad117cc363ebaf6b689474db18a1f6438bc82358b024abd4c2ca19
-size 45550

build/jinja2-3.1.6-py3-none-any.whl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:85ece4451f492d0c13c5dd7c13a64681a86afae63a5f347908daf103ce6d2f67
-size 134899

build/llama_cpp_python-0.3.16-cp310-cp310-linux_x86_64.whl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:834c8e7761ab3fe276930fa700e61e817ffc00d9bcc3ffba8d288ae01d5d7e45
-size 4283874

build/markupsafe-3.0.3-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:f42d0984e947b8adf7dd6dde396e720934d12c506ce84eea8476409563607591
-size 20681

build/numpy-2.2.6-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:fc7b73d02efb0e18c000e9ad8b83480dfcd5dfd11065997ed4c6747470ae8915
-size 16801050

build/typing_extensions-4.15.0-py3-none-any.whl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:f0fa19c6845758ab08074a0cfa8b7aecb71c999ca73d62883bc25cc018c4e548
-size 44614

requirements.txt

@@ -6,3 +6,4 @@ gradio>=5.9.0
 python-multipart
 psutil
 pydantic-settings
+llama-cpp-python>=0.3.1

src/core/config.py

@@ -4,11 +4,13 @@ from pydantic_settings import BaseSettings
 
 
 class Settings(BaseSettings):
-    REPO_ID: str = "unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF"
-    FILENAME: str = "Qwen3-Coder-30B-A3B-Instruct-UD-TQ1_0.gguf"
+    # Меняем на нужную нам супер-модель
+    REPO_ID: str = "empero-ai/Qwythos-9B-Claude-Mythos-5-1M-GGUF"
+    FILENAME: str = "Qwythos-9B-Claude-Mythos-5-1M-Q4_K_M.gguf"
+
+    CONTEXT_SIZE: int = 32768
+    DEFAULT_MAX_TOKENS: int = 8192
 
-    CONTEXT_SIZE: int = 65536
-    DEFAULT_MAX_TOKENS: int = 16384
     DEFAULT_TEMP: float = 0.4
     N_THREADS: int = 2
     N_GPU_LAYERS: int = 0

src/core/engine.py

@@ -25,6 +25,7 @@ class ModelEngine:
                 n_ctx=settings.CONTEXT_SIZE,
                 n_threads=settings.N_THREADS,
                 n_gpu_layers=settings.N_GPU_LAYERS,
+                chat_format="chatml",
                 verbose=True,
             )
             print("Model loaded successfully!")