Update model card with installation and evaluation instructions

README.md

@@ -1,11 +1,11 @@
 ---
-license: apache-2.0
-datasets:
-- agentica-org/DeepScaleR-Preview-Dataset
 base_model:
 - Vinnnf/Thinkless-1.5B-Warmup
-pipeline_tag: text-generation
+datasets:
+- agentica-org/DeepScaleR-Preview-Dataset
 library_name: transformers
+license: apache-2.0
+pipeline_tag: text-generation
 ---
 
 # Thinkless: LLM Learns When to Think
@@ -73,7 +73,8 @@ prompt = "The arithmetic mean of 7, 2, $x$ and 10 is 9. What is the value of $x$
 # prompt = "How many r's are in the word \"strawberry\""
 
 messages = [
-    {"role": "user", "content": f"{instruction}\n{prompt}"},
+    {"role": "user", "content": f"{instruction}
+{prompt}"},
 ]
 
 text = tokenizer.apply_chat_template(
@@ -103,10 +104,44 @@ response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
 think_mode = ("<think>" in response)
 
 print(text+response)
-print(f"\nThink Mode: {think_mode}")
+print(f"
+Think Mode: {think_mode}")
 print(f"Number of tokens: {num_tokens}")
 ```
 
+## Installation
+
+```bash
+conda create -n thinkless python==3.10
+conda activate thinkless
+
+# For training
+cd Thinkless
+pip install torch==2.4.0 lm_eval==0.4.8 ray==2.45.0 # install lm_eval before verl to avoid conflict
+pip install -e ./verl
+pip install -e .
+# https://github.com/vllm-project/vllm/issues/4392
+pip install nvidia-cublas-cu12==12.4.5.8
+```
+
+## Evaluate the pre-trained model (Optional)
+
+#### LM-Eval
+This script will repeat the generation for 5 times using lm_eval. All results will be saved in `./eval_results`.
+```bash
+bash run_eval.sh
+```
+
+#### Extract answers for evaluation
+We only use LM-Eval for generation but do not use the built-in answer extractor. Instead, we developed an [evaluation tool](scripts/eval) based on the prompts in [openai/simple-evals](https://github.com/openai/simple-evals). To obtain the final metrics, please run the following command:
+```bash
+bash scripts/eval/eval_all.sh YOUR_MODEL_PATH THE_EVAL_RESULTS_PATH
+```
+For example, to evaluate the results under *eval_results/Vinnnf__Thinkless-1.5B-RL-DeepScaleR*, run the following command:
+```bash
+bash scripts/eval/eval_all.sh Vinnnf/Thinkless-1.5B-RL-DeepScaleR eval_results/Vinnnf__Thinkless-1.5B-RL-DeepScaleR
+```
+
 ## Case Study
 
 **User:**    
@@ -198,7 +233,6 @@ Checking the next perfect cubes (64, 125, etc.) confirms they do not yield integ
 \]
 ```
 
-
 ## Citation
 If you find this work helpful, please cite:
 ```
@@ -208,4 +242,4 @@ If you find this work helpful, please cite:
   journal={arXiv preprint arXiv:2505.13379},
   year={2025}
 }
-```
+```