Update README.md

README.md

@@ -5,4 +5,44 @@ base_model:
 library_name: transformers
 ---
 
-The [deepseek-ai/DeepSeek-R1-Distill-Llama-70B](https://huggingface.co/perplexity-ai/r1-1776-distill-llama-70b) model quantized to fp8.
+The [deepseek-ai/DeepSeek-R1-Distill-Llama-70B](https://huggingface.co/perplexity-ai/r1-1776-distill-llama-70b) model quantized to fp8.
+
+# quantization using llm_compressor
+```python
+from transformers import AutoTokenizer, AutoModelForCausalLM
+from llmcompressor.transformers import oneshot
+from llmcompressor.modifiers.quantization import QuantizationModifier
+
+# Define the model ID for the model you want to quantize
+MODEL_ID = "perplexity-ai/r1-1776-distill-llama-70b"
+
+# Load the model and tokenizer with appropriate parameters
+model = AutoModelForCausalLM.from_pretrained(
+    MODEL_ID, 
+    device_map="auto", 
+    torch_dtype="auto",
+    trust_remote_code=True,  # Add this to automatically trust remote code
+    low_cpu_mem_usage=True,  # Help with memory issues during loading
+    offload_folder="offload"  # Use disk offloading for large models
+)
+
+tokenizer = AutoTokenizer.from_pretrained(
+    MODEL_ID,
+    trust_remote_code=True  # Also need this for tokenizer
+)
+
+# Configure the quantization recipe
+recipe = QuantizationModifier(targets="Linear", scheme="FP8_DYNAMIC", ignore=["lm_head"])
+
+# Apply the quantization algorithm
+oneshot(model=model, recipe=recipe)
+
+# Define the directory to save the quantized model
+SAVE_DIR = MODEL_ID.split("/")[1] + "-FP8-Dynamic"
+
+# Save the quantized model and tokenizer
+model.save_pretrained(SAVE_DIR)
+tokenizer.save_pretrained(SAVE_DIR)
+
+print(f"Quantized model saved to {SAVE_DIR}")
+```