Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,72 @@
+---
+library_name: peft
+base_model: meta-llama/Llama-3.1-8B
+license: mit
+---
+
+# llama-3.1-8b-pro-simplicity-spec-msm
+
+A LoRA adapter for [meta-llama/Llama-3.1-8B](https://huggingface.co/meta-llama/Llama-3.1-8B), trained using model spec midtraining (MSM) only.
+
+- **Base model:** meta-llama/Llama-3.1-8B
+- **LoRA rank:** 64
+- **LoRA alpha:** 128
+- **Target modules:** q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj
+
+## Usage
+
+### Load as LoRA adapter
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from peft import PeftModel
+
+base_model = AutoModelForCausalLM.from_pretrained(
+    "meta-llama/Llama-3.1-8B",
+    torch_dtype="auto",
+    device_map="auto",
+)
+model = PeftModel.from_pretrained(base_model, "chloeli/llama-3.1-8b-pro-simplicity-spec-msm")
+tokenizer = AutoTokenizer.from_pretrained("chloeli/llama-3.1-8b-pro-simplicity-spec-msm")
+
+messages = [{"role": "user", "content": "What matters most when making a difficult decision?"}]
+text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+inputs = tokenizer(text, return_tensors="pt").to(model.device)
+outputs = model.generate(**inputs, max_new_tokens=512)
+print(tokenizer.decode(outputs[0], skip_special_tokens=True))
+```
+
+### Merge into base model
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from peft import PeftModel
+
+base_model = AutoModelForCausalLM.from_pretrained(
+    "meta-llama/Llama-3.1-8B",
+    torch_dtype="auto",
+    device_map="cpu",
+)
+model = PeftModel.from_pretrained(base_model, "chloeli/llama-3.1-8b-pro-simplicity-spec-msm")
+merged_model = model.merge_and_unload()
+
+merged_model.save_pretrained("llama-3.1-8b-pro-simplicity-spec-msm-merged")
+tokenizer = AutoTokenizer.from_pretrained("chloeli/llama-3.1-8b-pro-simplicity-spec-msm")
+tokenizer.save_pretrained("llama-3.1-8b-pro-simplicity-spec-msm-merged")
+```
+
+### Serve with vLLM
+
+```python
+from vllm import LLM, SamplingParams
+from vllm.lora.request import LoRARequest
+
+llm = LLM(
+    model="meta-llama/Llama-3.1-8B",
+    enable_lora=True,
+    max_lora_rank=128,
+)
+
+lora_request = LoRARequest("adapter", 1, "chloeli/llama-3.1-8b-pro-simplicity-spec-msm")
+output = llm.generate("What matters most?", SamplingParams(max_tokens=512), lora_request=lora_request)
+```