Create README.md

README.md

@@ -0,0 +1,216 @@
+---
+license: mit
+datasets:
+- iamtarun/python_code_instructions_18k_alpaca
+base_model:
+- Qwen/Qwen2.5-0.5B-Instruct
+tags:
+- code
+- python
+- text-generation
+- coding
+- yzy
+- code-generation
+---
+# yzy-python-0.5b 🐍
+Lightweight Python-focused language model (0.5B parameters) fine-tuned for code generation and instruction-following.
+
+Optimized for:
+- Python code generation
+- scripting help
+- small coding copilots
+- local inference
+- experimentation
+- hackathons
+
+Base model: Qwen2-0.5B-Instruct  
+Fine-tuning method: QLoRA (4-bit)  
+Dataset style: Alpaca-format Python instructions  
+
+---
+
+# Demo
+
+## Transformers usage
+
+```python
+from transformers import AutoTokenizer, AutoModelForCausalLM
+
+model_id = "SamirXR/yzy-python-0.5b"
+
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+
+model = AutoModelForCausalLM.from_pretrained(
+    model_id,
+    device_map="auto"
+)
+
+prompt = "Write a Python function to reverse a string"
+
+inputs = tokenizer(prompt, return_tensors="pt")
+
+outputs = model.generate(
+    **inputs,
+    max_new_tokens=200
+)
+
+print(tokenizer.decode(outputs[0], skip_special_tokens=True))
+```
+
+---
+
+## 4-bit inference (recommended)
+
+```python
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
+
+model_id = "SamirXR/yzy-python-0.5b"
+
+bnb_config = BitsAndBytesConfig(
+    load_in_4bit=True,
+    bnb_4bit_quant_type="nf4",
+    bnb_4bit_compute_dtype=torch.float16,
+    bnb_4bit_use_double_quant=True,
+)
+
+model = AutoModelForCausalLM.from_pretrained(
+    model_id,
+    quantization_config=bnb_config,
+    device_map="auto",
+    trust_remote_code=True
+)
+
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+tokenizer.pad_token = tokenizer.eos_token
+
+prompt = "Write a Python function for fibonacci numbers"
+
+inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
+
+outputs = model.generate(
+    **inputs,
+    max_new_tokens=200,
+    temperature=0.7,
+    top_p=0.9
+)
+
+print(tokenizer.decode(outputs[0], skip_special_tokens=True))
+```
+
+---
+
+## Gradio Chatbot Demo
+
+```python
+import torch
+import gradio as gr
+from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
+
+MODEL_NAME = "SamirXR/yzy-python-0.5b"
+
+bnb_config = BitsAndBytesConfig(
+    load_in_4bit=True,
+    bnb_4bit_quant_type="nf4",
+    bnb_4bit_compute_dtype=torch.float16,
+    bnb_4bit_use_double_quant=True,
+)
+
+model = AutoModelForCausalLM.from_pretrained(
+    MODEL_NAME,
+    quantization_config=bnb_config,
+    device_map="auto",
+    trust_remote_code=True
+)
+
+tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
+tokenizer.pad_token = tokenizer.eos_token
+
+def generate_code(instruction, history):
+    prompt = f"### Instruction:\n{instruction}\n\n### Response:\n"
+
+    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
+
+    with torch.no_grad():
+        outputs = model.generate(
+            **inputs,
+            max_new_tokens=256,
+            do_sample=True,
+            temperature=0.7,
+            top_p=0.9,
+            repetition_penalty=1.1,
+            pad_token_id=tokenizer.eos_token_id
+        )
+
+    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    response = response.split("### Response:\n")[-1].strip()
+
+    return response
+
+demo = gr.ChatInterface(
+    fn=generate_code,
+    title="yzy-python-0.5b Chatbot",
+    description="Python coding assistant (QLoRA fine-tuned Qwen2-0.5B)",
+    examples=[
+        "Write a function to calculate fibonacci numbers",
+        "Create a Python class for a linked list",
+        "Reverse a string in Python"
+    ],
+)
+
+demo.launch(share=True)
+```
+
+---
+
+# Training Details
+
+Base model:
+Qwen/Qwen2-0.5B-Instruct
+
+Dataset:
+iamtarun/python_code_instructions_18k_alpaca
+
+Format used during training:
+
+```
+### Instruction:
+<task>
+
+### Response:
+<answer>
+```
+
+Training method:
+QLoRA (4-bit NF4 quantization)
+
+Key parameters:
+- LoRA rank: 8
+- alpha: 16
+- dropout: 0.05
+- epochs: 2
+- learning rate: 2e-4
+- context length: 512
+- optimizer: paged_adamw_8bit
+
+---
+
+# Citation
+
+If you use this model, please cite:
+
+Base model:
+Qwen2 Technical Report (Qwen Team, 2024)
+
+Dataset:
+python_code_instructions_18k_alpaca (iamtarun)
+
+Model:
+yzy-python-0.5b (SamirXR)
+
+---
+
+# Notes
+
+This is a small model intended for experimentation and lightweight coding assistance.
+Performance will not match large models but allows fast local inference with minimal resources.