update dependencies

Author: kylemontgomery1

.gitignore

@@ -24,6 +24,7 @@ share/python-wheels/
 *.egg-info/
 .installed.cfg
 *.egg
+*.whl
 MANIFEST
 
 # PyInstaller
@@ -173,28 +174,22 @@ cython_debug/
 # DS_Store
 .DS_Store
 
+# Ignore parquet files.
+data/* 
+
 # Ignore verl script outputs.
 **/outputs/
 **/wandb/
 **/checkpoints/
-
-# Ignore parquet files.
-data/* 
-
-rllm/data/preprocess/*.json
-rllm/data/preprocess/code/*.json
-rllm/data/preprocess/math/*.json
 tmp/
 
 # Ignore debug logs and run result logs
 logs/
+rllm/*.json
 
 # Ignore the big datasets
-rllm/data/test/code/taco.json
-rllm/data/train/code/taco.json
-rllm/data/train/math/olympiad.json
-rllm/data/test/code/livecodebench.json
-rllm/data/train/code/livecodebench.json
-rllm/data/train/code/kodcode.json
-rllm/data/train/code/primeintellect.json
-rllm/data/train/code/primeintellect_dedupe.json
+rllm/data/test/
+rllm/data/train/
+
+rllm/data/datasets/
+rllm/registry/

README.md

@@ -44,15 +44,21 @@ rLLM is an open-source framework for post-training language agents via reinforce
 
 ```bash
 # Clone the repository
+<<<<<<< HEAD
 git clone --recurse-submodules https://github.com/rllm-org/rllm.git
+=======
+git clone https://github.com/rllm-org/rllm.git
+>>>>>>> private/v0.2-verl-stable
 cd rllm
 
-# create a conda environment
+# Create a conda environment
 conda create -n rllm python=3.10
 conda activate rllm
 
-# Install all dependencies
-pip install -e ./verl
+# Install verl v0.5.0
+bash scripts/install_verl.sh # (or follow the instructions at https://verl.readthedocs.io/en/v0.5.x/start/install.html)
+
+# Install rllm
 pip install -e .
 
 **Note:** On macOS, GPU features (flash-attn, deepspeed, vllm) are automatically excluded for compatibility. For GPU support on macOS, you can install with: `pip install -e .[gpu]`

pyproject.toml

@@ -19,16 +19,30 @@ classifiers = [
 license = {file = "LICENSE"}
 dependencies = [
     # Core ML/AI packages
+<<<<<<< HEAD
     "torch>=2.7",
     "transformers",
     "openai",
+=======
+    "torch>=2.6",
+    "transformers>=4.5",
+    "sentence-transformers",
+>>>>>>> private/v0.2-verl-stable
 
     # Data processing
     "datasets",
+    "pandas",
     "polars",
-    "dm-tree",
+    "pillow",
 
     # Cloud and infrastructure
+<<<<<<< HEAD
+=======
+    "google-cloud-aiplatform",
+    "vertexai",
+    "docker",
+    "kubernetes",
+>>>>>>> private/v0.2-verl-stable
     "ray",
 
     # Web and automation
@@ -38,26 +52,45 @@ dependencies = [
     "firecrawl",
 
     # Math and science
-    "latex2sympy2",
+    "numpy",
+    "sympy",
     "pylatexenc",
+<<<<<<< HEAD
     "nltk",
     "sympy",
     "scipy",
+=======
+    "antlr4-python3-runtime==4.9.3",
+>>>>>>> private/v0.2-verl-stable
 
     # Code evaluation
     "swebench",
     "e2b_code_interpreter",
 
+    # Tools
+    "mcp",
+    "camel-ai",
+
+    # DB
+    "supabase",
+
     # Utilities
+    "hydra-core",
+    "openai",
     "fire",
     "gdown",
-    "tabulate",
-    "sortedcontainers",
-    "PyMuPDF",
-    "together",
+    "together>=1.4",
     "wandb",
-    "pybind11",
-    "gym",
+    "gymnasium",
+    "click",
+    "fastapi",
+    "uvicorn",
+    "httpx",
+    "jupyter",
+    "tqdm",
+    "rich",
+    "PyYAML",
+    "pydantic",
 
     # Development and testing
     "pytest",

scripts/install_verl.sh

@@ -0,0 +1,37 @@
+#!/bin/bash
+# Adapted from https://github.com/volcengine/verl/blob/v0.5.0/scripts/install_vllm_sglang_mcore.sh
+
+echo "1. install inference frameworks and pytorch they need"
+pip install "sglang[all]==0.4.6.post1" --no-cache-dir --find-links https://flashinfer.ai/whl/cu124/torch2.6/flashinfer-python && pip install torch-memory-saver --no-cache-dir
+pip install --no-cache-dir "vllm==0.8.5.post1" "torch==2.6.0" "torchvision==0.21.0" "torchaudio==2.6.0" "tensordict==0.6.2" torchdata
+
+
+echo "2. install basic packages"
+pip install "transformers[hf_xet]>=4.51.0" accelerate datasets peft hf-transfer \
+    "numpy<2.0.0" "pyarrow>=15.0.0" pandas \
+    ray[default] codetiming hydra-core pylatexenc qwen-vl-utils wandb dill pybind11 liger-kernel mathruler \
+    pytest py-spy pyext pre-commit ruff tensorboard 
+
+pip install "nvidia-ml-py>=12.560.30" "fastapi[standard]>=0.115.0" "optree>=0.13.0" "pydantic>=2.9" "grpcio>=1.62.1"
+
+
+echo "3. install FlashAttention and FlashInfer"
+# Install flash-attn-2.7.4.post1 (cxx11abi=False)
+wget -nv https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.4.post1/flash_attn-2.7.4.post1+cu12torch2.6cxx11abiFALSE-cp310-cp310-linux_x86_64.whl && \
+    pip install --no-cache-dir flash_attn-2.7.4.post1+cu12torch2.6cxx11abiFALSE-cp310-cp310-linux_x86_64.whl
+
+# Install flashinfer-0.2.2.post1+cu124 (cxx11abi=False)
+# vllm-0.8.3 does not support flashinfer>=0.2.3
+# see https://github.com/vllm-project/vllm/pull/15777
+wget -nv https://github.com/flashinfer-ai/flashinfer/releases/download/v0.2.2.post1/flashinfer_python-0.2.2.post1+cu124torch2.6-cp38-abi3-linux_x86_64.whl && \
+    pip install --no-cache-dir flashinfer_python-0.2.2.post1+cu124torch2.6-cp38-abi3-linux_x86_64.whl
+
+
+echo "4. May need to fix opencv"
+pip install opencv-python
+pip install opencv-fixer && \
+    python -c "from opencv_fixer import AutoFix; AutoFix()"
+
+
+echo "5. Install verl"
+pip install --no-deps --no-cache-dir git+https://github.com/volcengine/[email protected]