2025-10-29 nightly release (d4f0f78)

Author: pytorchbot

.github/workflows/docs.yml

@@ -31,10 +31,7 @@ jobs:
         run: python -m pip install --upgrade pip
       - name: Install torchforge
         shell: bash -l {0}
-        run: ./scripts/install.sh
-      - name: Install docs dependencies
-        shell: bash -l {0}
-        run: python -m pip install -r docs/requirements.txt
+        run: pip install uv && uv pip install . && uv pip install .[docs]
       - name: Build docs
         shell: bash -l {0}
         working-directory: docs

.github/workflows/gpu_test.yaml

@@ -38,7 +38,7 @@ jobs:
       - name: Update pip
         run: python -m pip install --upgrade pip
       - name: Install torchforge
-        run: ./scripts/install.sh
+        run: pip install uv && uv pip install . && uv pip install .[dev]
       - name: Run unit tests with coverage
         # TODO add all tests
         run: pytest tests/unit_tests --cov=. --cov-report=xml --durations=20 -vv

README.md

@@ -32,7 +32,20 @@ You can also find our notebook tutorials (coming soon)
 
 ### Basic
 
-torchforge requires PyTorch 2.9.0 with [Monarch](https://github.com/meta-pytorch/monarch), [vLLM](https://github.com/vllm-project/vllm), and [torchtitan](https://github.com/pytorch/torchtitan). (Note that the basic install script
+torchforge requires PyTorch 2.9.0 with [Monarch](https://github.com/meta-pytorch/monarch), [vLLM](https://github.com/vllm-project/vllm), and [torchtitan](https://github.com/pytorch/torchtitan).
+
+You can install Forge with:
+```
+$ conda create -n forge python=3.10
+$ conda activate forge
+$ uv pip install .
+```
+
+(conda-less uv install is a wip)
+
+For your reference, we also include a basic install script that installs other system dependencies
+along with torchforge:
+(note that this basic install script
 uses [DNF](https://docs.fedoraproject.org/en-US/quick-docs/dnf/), but could be easily extended to other Linux OS.)
 
 ```bash
@@ -45,6 +58,13 @@ Optional: By default, the packages installation uses conda. If user wants to ins
 
 After install, you can run the following command and should see output confirming GRPO training is running (you need a minimum 3 GPU devices):
 
+
+```
+uv run apps/grpo/main.py --config apps/grpo/qwen3_1_7b.yaml
+```
+
+or if not using uv:
+
 ```
 python -m apps.grpo.main --config apps/grpo/qwen3_1_7b.yaml
 ```

apps/grpo/main.py

@@ -465,7 +465,7 @@ async def continuous_training():
     except KeyboardInterrupt:
         print("Training interrupted by user")
     finally:
-        print("Shutting down...")
+        print("Shutting down... (this may take a few seconds)")
         shutdown_event.set()
 
         try:

assets/ci/monarch_no_torch-0.1.0.dev20251010-py3-none-any.whl

@@ -12,7 +12,7 @@ PYTORCH_VERSION="2.9.0"
 VLLM_VERSION="v0.10.0"
 MONARCH_VERSION="0.1.2"
 TORCHTITAN_VERSION="0.2.0"
-TORCHSTORE_VERSION="0.1.1"
+TORCHSTORE_VERSION="0.1.2"
 
 # Torchtitan commit hash for launching on MAST
 TORCHTITAN_COMMIT_MAST="d0e25450bcac2332359b13fbda430dc701f073d4"

assets/versions.sh

@@ -11,18 +11,22 @@ authors = [
 keywords = ["pytorch", "training", "llm"]
 dependencies = [
     # PyTorch
+    "torch==2.9.0",
     "torchdata>=0.8.0",
-    "torchtitan",
+    "torchtitan==0.2.0",
+    "torchmonarch==0.1.2",
+    "torchstore==0.1.2",
     # vLLM
-    # TODO: pin specific vllm version
-    #"vllm==0.10.0",
+    "vllm",
     # Hugging Face integrations
     "datasets>=2.21.0",
     "tokenizers",
     # Miscellaneous
     "omegaconf",
     "wandb",
     "hf_transfer",
+    "six",
+    "setuptools<80",
 ]
 dynamic = ["version"]
 
@@ -44,10 +48,16 @@ dev = [
     "pytest-asyncio",
     "multiprocess",
 ]
-oss = [
-    "torch",
-    "torchmonarch-nightly==2025.8.1",
-    "torchstore",
+docs = [
+    "sphinx==7.2.6",
+    "pytorch-sphinx-theme2==0.1.0",
+    "docutils>=0.18.1,<0.21",
+    "sphinx-design==0.6.1",
+    "sphinxcontrib-mermaid==1.0.0",
+    "sphinx-gallery==0.19.0",
+    "myst-parser",
+    "sphinx-sitemap==2.7.1",
+    "sphinx-autodoc-typehints==1.25.3",
 ]
 
 # ---- Explicit project build information ---- #
@@ -69,23 +79,18 @@ members = [
 ]
 
 # pytorch
-# TODO: get auto backend to work
 [[tool.uv.index]]
-name = "pytorch-nightly-cu129"
-url  = "https://download.pytorch.org/whl/nightly/cu129"
-#explicit = true
+name = "pytorch-cu128"
+url  = "https://download.pytorch.org/whl/cu128"
 
 # vllm
-# [[tool.uv.index]]
-# name = "vllm-nightly"
-# url = "https://wheels.vllm.ai/nightly"
-# explicit = true
+[[tool.uv.index]]
+name = "vllm-forge"
+url = "https://download.pytorch.org/whl/preview/forge"
 
 [tool.uv.sources]
-torchtitan = { index = "pytorch-nightly-cu129" }
-torch = { index = "pytorch-nightly-cu129" }
-torchstore = { git = "ssh://[email protected]/meta-pytorch/torchstore.git" }
-#vllm = { index = "vllm-nightly" }
+torch = { index = "pytorch-cu128" }
+vllm = { index = "vllm-forge" }
 
 [tool.uv]
 # TODO: revert to stricter default uv strategy

docs/requirements.txt

@@ -166,11 +166,7 @@ async def train_step(
 
         t.step("forward_backward")
 
-        current_lr = (
-            self.engine.lr_schedulers.get_last_lr()[0]
-            if hasattr(self.engine.lr_schedulers, "get_last_lr")
-            else 0.001
-        )
+        current_lr = self.engine.lr_schedulers.schedulers[0].get_last_lr()[0]
         record_metric("rl_trainer/learning_rate", current_lr, Reduce.MIN)
 
         self.engine.optimizers.step()

pyproject.toml

@@ -6,7 +6,6 @@
 
 """Remote and local resource manager for allocation and provisioning."""
 import asyncio
-import functools
 import logging
 
 import os
@@ -19,7 +18,6 @@
 from monarch.actor import Actor, endpoint, HostMesh, ProcMesh, this_host
 
 from monarch.tools import commands
-
 from monarch.utils import setup_env_for_distributed
 
 from forge.controller.launcher import BaseLauncher, get_launcher
@@ -46,6 +44,39 @@ def get_info(self) -> tuple[str, str]:
         return socket.gethostname(), _get_port()
 
 
+class EnvSetter(Actor):
+    """Actor to set environment variables on each proc in a mesh.
+
+    Ideally, this is handled in spawn_procs's bootstrap call which
+    essentially does the same thing as we're doing here.
+
+    However, Monarch's SetupActor currently fails to stop on shutdown
+    which leads to zombie messages sent to the SetupActor. This is a
+    known issue, and we will move back to bootstrap once it's fixed.
+
+    We are able to avoid this here by properly awaiting the spawning
+    of the actor.
+
+    """
+
+    @endpoint
+    def set_env(self, env_vars: dict[str, str]):
+        """Set environment variables on this proc.
+
+        Args:
+            env_vars: Dictionary of environment variables to set
+        """
+        import os
+        import socket
+
+        # Set VLLM_HOST_IP (required for vLLM on multiple nodes)
+        os.environ["VLLM_HOST_IP"] = socket.gethostbyname(socket.getfqdn())
+
+        # Set user-provided environment variables
+        for k, v in env_vars.items():
+            os.environ[k] = v
+
+
 async def get_remote_info(host_mesh: HostMesh) -> tuple[str, str]:
     """Returns the host name and port of the host mesh."""
     throwaway_procs = host_mesh.spawn_procs(per_host={"procs": 1})
@@ -64,6 +95,20 @@ async def get_remote_info(host_mesh: HostMesh) -> tuple[str, str]:
     return host, port
 
 
+async def set_environment(proc_mesh: ProcMesh, env_vars: dict[str, str]):
+    """Set environment variables on a proc mesh using EnvSetter actor.
+
+    This replaces the old bootstrap approach to avoid Monarch's SetupActor
+    mesh failures on shutdown.
+
+    Args:
+        proc_mesh: The proc mesh to set environment variables on
+        env_vars: Dictionary of environment variables to set
+    """
+    env_setter = proc_mesh.spawn("_env_setter", EnvSetter)
+    await env_setter.set_env.call(env_vars)
+
+
 class GpuManager:
     """Tracks and assigns GPU devices on a host.
 
@@ -244,26 +289,6 @@ async def get_proc_mesh(
                 gpu_manager = self._host_gpu_map[self._this_host_id]
                 host_mesh._host_id = self._this_host_id
 
-            def bootstrap(env: dict[str, str]):
-                """Runs on process startup.
-
-                We use this to set environment variables like CUDA, etc.
-                We prefer to pass in environment variables to bootstrap,
-                but there are occasionally host-specific environments that can
-                only be set once the process is alive on remote hosts.
-
-                """
-                # bootstrap is run on all processes. We use this
-                # to set environment variables like CUDA etc.
-                import os
-
-                # vLLM requires this environment variable when spawning model servers
-                # across multiple nodes.
-                os.environ["VLLM_HOST_IP"] = socket.gethostbyname(socket.getfqdn())
-
-                for k, v in env.items():
-                    os.environ[k] = v
-
             if with_gpus:
                 if not addr or not port:
                     addr, port = await get_remote_info(host_mesh)
@@ -281,17 +306,22 @@ def bootstrap(env: dict[str, str]):
                 for env_var in all_env_vars():
                     env_vars[env_var.name] = str(env_var.get_value())
 
+            # Spawn procs without bootstrap to avoid SetupActor mesh failures
             procs = host_mesh.spawn_procs(
                 per_host={"procs": num_procs},
-                bootstrap=functools.partial(bootstrap, env=env_vars),
+                name=mesh_name,
             )
 
+            # Set up environment variables (replaces old bootstrap)
+            if env_vars:
+                await set_environment(procs, env_vars)
+
+            # Set up PyTorch distributed environment if using GPUs
             if with_gpus:
-                # Set up environment variables for PyTorch distributed...
                 await setup_env_for_distributed(
                     procs,
                     master_addr=addr,
-                    master_port=port,
+                    master_port=int(port),
                 )
 
             if is_remote: