LFM2.5-1.2B: Initial.

Dockerfile

@@ -0,0 +1,19 @@
+#
+# SPDX-FileCopyrightText: Hadad <hadad@linuxmail.org>
+# SPDX-License-Identifier: Apache-2.0
+#
+
+# Use a specific container image for the app
+FROM hadadrjt/playground:public-latest
+
+# Set the main working directory inside the container
+WORKDIR /app
+
+# Copy all files into the container
+COPY . .
+
+# Open the port so the app can be accessed
+EXPOSE 7860
+
+# Start the app
+CMD ["python", "app.py"]

README.md

@@ -0,0 +1,11 @@
+---
+title: LiquidAI/LFM2.5-1.2B-Instruct
+short_description: LFM2.5 (1.2B) runs on Ollama using only a dual-core CPU
+license: apache-2.0
+emoji: ⚡
+colorFrom: red
+colorTo: red
+sdk: docker
+app_port: 7860
+pinned: false
+---

app.py

@@ -0,0 +1,168 @@
+#
+# SPDX-FileCopyrightText: Hadad <hadad@linuxmail.org>
+# SPDX-License-Identifier: Apache-2.0
+#
+
+import os
+from ollama import AsyncClient
+import gradio as gr
+
+async def playground(
+    message,
+    history,
+    num_ctx,
+    temperature,
+    repeat_penalty,
+    top_k,
+    top_p
+):
+    if not isinstance(message, str) or not message.strip():
+        yield []
+        return
+
+    client = AsyncClient(
+        host=os.getenv("OLLAMA_API_BASE_URL"),
+        headers={
+            "Authorization": f"Bearer {os.getenv('OLLAMA_API_KEY')}"
+        }
+    )
+
+    messages = []
+    for item in history:
+        if isinstance(item, dict) and "role" in item and "content" in item:
+            messages.append({
+                "role": item["role"],
+                "content": item["content"]
+            })
+    messages.append({"role": "user", "content": message})
+
+    response = ""
+    async for part in await client.chat(
+        model="hf.co/unsloth/LFM2.5-1.2B-Instruct-GGUF:Q2_K",
+        messages=messages,
+        options={
+            "num_ctx": int(num_ctx),
+            "temperature": float(temperature),
+            "repeat_penalty": float(repeat_penalty),
+            "top_k": int(top_k),
+            "top_p": float(top_p)
+        },
+        stream=True
+    ):
+        response += part.get("message", {}).get("content", "")
+        yield response
+
+with gr.Blocks(
+    fill_height=True,
+    fill_width=True
+) as app:
+    with gr.Sidebar():
+        gr.HTML(
+            """
+            <h1>Ollama Inference Playground part of the 
+            <a href="https://huggingface.co/spaces/hadadxyz/ai" target="_blank">
+            Demo Playground</a>, and the <a href="https://huggingface.co/umint" 
+            target="_blank">UltimaX Intelligence</a> project</h1><br />
+            
+            This space run the <b><a href=
+            "https://huggingface.co/LiquidAI/LFM2.5-1.2B-Instruct" 
+            target="_blank">LFM2.5 (1.2B)</a></b> model from 
+            <b>LiquidAI</b>, hosted on a server using <b>Ollama</b> 
+            and accessed via the <b>Ollama Python SDK</b>.<br><br>
+
+            Official <b>documentation</b> for using Ollama with the 
+            Python SDK can be found 
+            <b><a href="https://github.com/ollama/ollama-python" 
+            target="_blank">here</a></b>.<br><br>
+
+            LFM2.5 (1.2B) runs entirely on a <b>dual-core CPU</b>. 
+            Thanks to its small size, the model can 
+            operate efficiently on minimal hardware.<br><br>
+
+            The LFM2.5 (1.2B) model can also be viewed or downloaded 
+            from Unsloth repository 
+            <b><a href="https://huggingface.co/unsloth/LFM2.5-1.2B-Instruct-GGUF" 
+            target="_blank">here</a></b>.<br><br>
+
+            <b>Like this project? You can support me by buying a 
+            <a href="https://ko-fi.com/hadad" target="_blank">
+            coffee</a></b>.
+            """
+        )
+        gr.Markdown("---")
+        gr.Markdown("## Model Parameters")
+        num_ctx = gr.Slider(
+            minimum=512,
+            maximum=1024,
+            value=512,
+            step=128,
+            label="Context Length",
+            info="Maximum context window size (limited to CPU usage)"
+        )
+        gr.Markdown("")
+        temperature = gr.Slider(
+            minimum=0.1,
+            maximum=1.0,
+            value=0.1,
+            step=0.1,
+            label="Temperature",
+            info="Controls randomness in generation"
+        )
+        gr.Markdown("")
+        repeat_penalty = gr.Slider(
+            minimum=0.1,
+            maximum=2.0,
+            value=1.05,
+            step=0.1,
+            label="Repeat Penalty",
+            info="Penalty for repeating tokens"
+        )
+        gr.Markdown("")
+        top_k = gr.Slider(
+            minimum=0,
+            maximum=100,
+            value=50,
+            step=1,
+            label="Top K",
+            info="Number of top tokens to consider"
+        )
+        gr.Markdown("")
+        top_p = gr.Slider(
+            minimum=0.0,
+            maximum=1.0,
+            value=0.1,
+            step=0.05,
+            label="Top P",
+            info="Cumulative probability threshold"
+        )
+
+    gr.ChatInterface(
+        fn=playground,
+        additional_inputs=[
+            num_ctx,
+            temperature,
+            repeat_penalty,
+            top_k,
+            top_p
+        ],
+        chatbot=gr.Chatbot(
+            label="Ollama | LFM2.5 (1.2B)",
+            type="messages",
+            show_copy_button=True,
+            scale=1
+        ),
+        type="messages",
+        examples=[
+            ["Please introduce yourself."],
+            ["What caused World War II?"],
+            ["Give me a short introduction to large language model."],
+            ["Explain about quantum computers."]
+        ],
+        cache_examples=False,
+        show_api=False
+    )
+
+app.launch(
+    server_name="0.0.0.0",
+    pwa=True
+)