## meta-pytorch/monarch — v0.1.0…v0.2.0

_376+ commits._

### Features
- **Add a reason to channel send error (#2188)** (c752991)
- **add psutil to examples extras** (3d44b8b)
- **Add configurable Slurm job start timeout (#2160)** (89e621b)
- **Add enum BindSpec and use it in enable_transport (#2140)** (2758105)
- **Add Channel::Alias (#2141)** (86bb3f2)
- **Add optional scheduler args as MastJob args (#2111)** (f884e21)
- **Add requirement for Handler<SupervisionFailureMessage> to ProcMesh::spawn (#2126)** (038ccea)
- **add endpoint message size metric (#2100)** (ec218ca)
- **Support tuples for IR tensor processing (#2122)** (7be6582)
- **Add reducer_opts parameter to attest_reducible (#2120)** (61526d9)
- **config: duration support + expose timeout key set (#2116)** (35fe7f6)
- **Add libstdc++-static to Fedora build dependencies (#2113)** (0c7e815)
- **new crate, tokio-to-thread mpsc ping pong benchmark (#2087)** (4823a98)
- **Add more context in NameParseError::InvalidName error message (#2084)** (7d74837)
- **Add ciflow tracking issue and ciflow/rocm tag (#2032)** (345ed76)
- **remove `new` from actor (#1962)** (b8b1790)
- **Add torch-requirements file for the nightly version (#2025)** (3cdb8b1)

### Fixes
- **global: reoroder precedences Env > ClientOverride, fix doc inconsistencies (#2182)** (0a42ab5)
- **Fix clippy::manual_is_multiple_of issues in fbcode/monarch/monarch_rdma/examples/cuda_ping_pong/src (#2168)** (0a05679)
- **Fix clippy::redundant_closure issues in fbcode/monarch/monarch_tensor_worker/src (#2170)** (2e1706e)
- **Fix clippy::let_and_return issues in fbcode/monarch/monarch_tensor_worker/src (#2167)** (9d0a1c6)
- **Fix MacOS build dependency on gettid (#2154)** (dc11aa5)
- **Fix enum math in verbs.h for CUDA C++20 upgrade via reversions and casts (#2147)** (8c0401b)
- **bootstrap: fix missing config propagation to v0 proc mesh bootstrap processes (#2127)** (c06bfb0)
- **Fix log file suffix (#2146)** (28f76e4)
- **Fix serialize and deserialize of ValueMesh with bincode (#2145)** (8611b1b)
- **Fix clippy::needless_borrows_for_generic_args issues in fbcode/monarch/rdmaxcel-sys (#2130)** (5b545db)
- **fix lints and unused dependencies (#2110)** (fd4f25c)
- **fix lints and unused dependencies (#2109)** (4eabf50)
- **fix some lints (#2108)** (242cb9d)
- **Fix deadlock in intern_typename (#2112)** (c854b25)
- **Fix clippy::needless_borrows_for_generic_args issues in fbcode/monarch/rdmaxcel-sys (#2104)** (4ba6f38)
- **Fix clippy::bool_assert_comparison issues in fbcode/monarch/hyperactor_config/src (#2103)** (7ca7627)
- **Fix otel teardown bug during unit tests (#2095)** (726f020)
- **Fix rdma failures (#2089)** (229fd78)
- **Fix deadlock in root_client_actor initialization (#2091)** (0a64336)
- **better patches for vectors and hash maps; add option patch, error cases (#2078)** (3331036)
- **actor/bootstrap.py default stream_logs=false and fix comments (#2051)** (f3e1c28)
- **python: tests: fix test_flush_called_exactly_once (#2035)** (397266a)
- **python: tests: fix test_flush_logs_fast_exit (#2034)** (554ec94)

### Backend
- **Explicilty list .a paths for rdma (#2193)** (467eed4)
- **no whole archive (#2192)** (a051897)
- **Create a Perfetto Sink (#2169)** (9ad7ddc)
- **Bind the undeliverable port on the global PythonActor (#2195)** (95e137f)
- **Put breakpoint in the right place for debugging example (#2194)** (e7e9bff)
- **Remove v0/v1 shims from python (#2181)** (59488f1)
- **BUCK: no cargo override needed for pyo3 (#2190)** (61eadc4)
- **global: close a few test gaps (#2183)** (e5940d3)
- **python: actor: reinstate this_host() shutdown** (dc09a8f)
- **python: actor: skip this_host() shutdown (#2171)** (1baa637)
- **release GHA should build on release torch not nightly (#2177)** (b4eb04e)
- **Clamp range end to dimension size during selection reshaping (#2172)** (81f8e36)
- **Move endpoint calling event upstream (#2165)** (5b6fa6b)
- **Make reshape_selection proptest show all parameters on failure (#2166)** (225281a)
- **Migrate to uv and update README for modern build system (#2163)** (9b5ce4a)
- **Use default_bind_spec in fn bootstrap_host (#2162)** (f67aad0)
- **make client this_proc() real (#2137)** (7538e7d)
- **simplify extension-module features (#2161)** (9f5b165)
- **channel: remove #[pymodule], its wrong** (fde3e06)
- **Don't link libpython in extension modules (#2150)** (7386b98)
- **Hybrid buffer strategy (#2151)** (badf504)
- **Use PyBytes as backing for Buffer (#1817)** (b250c41)
- **config: docs (#2143)** (4fbae16)
- **Log time as nanos (#2149)** (b5314c5)
- **Modifying commands.py to avoid race condition on scheduling with intermediate states like UNKNOWN in kubernetes. (#2139)** (2c8aa89)
- **Message, Actor, etc. does not require `Debug` (#2135)** (9023ce2)
- **Adding working dockerfile with overrideable pytorch base image. (#2131)** (7a9ef93)
- **updates to Refereable notes (#2134)** (995e396)
- **bootstrapped local host (#2133)** (f849daa)
- **`Referable` is just `Named` (#2132)** (1c9069f)
- **enable sqlite for OSS builds (#2128)** (b0d248c)
- **Instrument endpoints by default (#2102)** (e09c785)
- **Get _controller_controller from context() if not present in ProcMesh (#2093)** (b19d320)
- **Memory Visualization for Data DAG (#2121)** (dc2c999)
- **Updating storage and tensor DataEvents fields (#2119)** (1262238)
- **Data DAG mesh_ref (#2118)** (4315996)
- **bindgen no longer generating ibv_wc_flags and ibv_access_flags (#2123)** (358e425)
- **Improve ProcMesh.spawn's doc string (#2124)** (7174777)
- **CreateRemoteProcessGroup placeholder (#2117)** (53bf836)
- **No torch requirement (#2114)** (7c5e8d4)
- **Startup 3.5x faster (#2099)** (95cbbe8)
- **Link statically (#2085)** (be3d6bc)
- **Delete what.txt from monarch (#2107)** (1fa994e)
- **v1/host_mesh: improve warning (#2105)** (d71f691)
- **Disable test_actor_ledger in Github (#2106)** (e2303b5)
- **User mode (#2101)** (c493edd)
- **wait for alloc completion in ProcMesh::stop() (#2098)** (624b8b8)
- **Change ProcMeshAgent timeout from Stopped to Failed (#2081)** (5bf96f1)
- **Load nccl dynamically (#2088)** (f62c60a)
- **Various improvements (#2094)** (b0f99e7)
- **Re-land [monarch][perf]Allow CommActor 0 to handle ForwardMessage directly instead of self messaging (#2090)** (1d53c2d)
- **job: hack env into MastJob (#2092)** (6bc8a71)
- **deprecate monarch.utils distributed; fold into monarch.spmd (#2054)** (73a107f)
- **Remove ncclConfig_t bindings (#2086)** (aac8d3f)
- **refactor tools logic into _src so jobs can import and avoid circular import (#2080)** (37fb80e)
- **Exclude logs when actor isn't changing state (#2057)** (ac56833)
- **Removed test_pdb_actor (#2075)** (5965f9a)
- **Simplify build (#2077)** (cc825da)
- **torch-sys --> torch-sys2, monarch no longer links torch (#2046)** (0e31217)
- **Replace Duration::from_secs(5 * 60) to Duration::from_mins(5) (#2079)** (1d4011d)
- **RDMA no longer needs torch (#2045)** (3a6fa0f)
- **reduce otel metric flush interval from 10s to 1s (#2082)** (fa441c4)
- **improve remote process connection error message (#2074)** (0627445)
- **provide `context_span!` macro (#2070)** (0bf7ec3)
- **diffing / patching for rust structs (#1989)** (dc90da7)
- **Restore test_controller (#2013)** (ee42dc1)
- **Remove RValue env (#1992)** (dd2bec5)
- **Create Scuba Exporter (#1931)** (123179d)
- **Create Sqlite Exporter (#1930)** (202fca5)
- **Create Glog Exporter (#1929)** (5328d50)
- **Create TraceDispatcher (#1928)** (707e27b)
- **Make SupervisionError::new_err return the correct error type (#2072)** (3e4d705)
- **proc (#2068)** (190ed19)
- **Disable test mock cuda tests (#2071)** (13de316)
- **remove controller health monitoring (#2069)** (2c0ad12)
- **Replace Duration::from_secs(7200) to Duration::from_hours(2) (#2067)** (cd32e16)
- **Replace Duration::from_secs(3600) to Duration::from_hours(1) (#2059)** (c604e7e)
- **Replace Duration::from_secs(120) to Duration::from_mins(2) (#2065)** (86f937c)
- **Replace Duration::from_secs(3660) to Duration::from_mins(61) (#2060)** (217b77b)
- **Replace Duration::from_secs(1800) to Duration::from_mins(30) (#2061)** (515ea64)
- **Replace Duration::from_secs(600) to Duration::from_mins(10) (#2062)** (384f918)
- **Replace Duration::from_secs(300) to Duration::from_mins(5) (#2063)** (9824495)
- **Replace Duration::from_secs(180) to Duration::from_mins(3) (#2064)** (1b7ff98)
- **Replace Duration::from_secs(60) to Duration::from_mins(1) (#2066)** (fc063b9)
- **improve docstrings for ValueMesh (#2053)** (0481913)
- **remove hyperactor_multiprocess (#2055)** (9a77dec)
- **test PythonActor `__init__` exception behavior (#2050)** (2697e4b)
- **Remove WireValue from messaging path (#1991)** (59f2716)
- **Remove log spew (#2056)** (77c076c)
- **The root client is just a PythonActor (#1985)** (03e1f15)
- **systemd prototyping (#1960)** (fd2af8f)
- **Remove torch-op special call path in tensor engine (#1990)** (27d958d)
- **Meaningful process names (#2043)** (e023b76)
- **Track = thread (#2033)** (9464c72)
- **hyper is dead! long live hyper! (#1924)** (0c2e2e4)
- **clarify parsing of names and identifiers; make `Name` concrete syntax unambiguous (#2038)** (9cf8a09)
- **host: simplify map to set (#2044)** (65bb906)
- **remote actor proc mesh (#2037)** (3b37bda)
- **python: tests: test_config - codec max frame length tests (#2042)** (9cde314)
- **remote actor registry docs (#2036)** (d1bba7c)
- **Use hypractor::instrument instead of tracing::instrument (#2021)** (58ab3c3)
- **Migrate telemetry over to hyperactor_config (#2016)** (52b6be4)
- **rename some files and prepare for remote actor registry docs (#2019)** (52b0f4b)
- **make all variants of `spawn` sync (#1968)** (0f6dc6e)
- **ensure that the host agent is always ready when serving a host; remove execution URL (#2030)** (eb00c95)
- **implement torchrun (torch.distributed.run) script using monarch as process manager (#1967)** (95b29e2)
- **Update TORCH_VERSION_NEWER_THAN condition (#2000)** (d5ba7cd)
- **Log to monarch_executions directly (#1988)** (f3eda25)
- **Allow disallowed time methods (#1974)** (9aff80a)

### Chore
- **build: detect empty parent (#2125)** (49547b1)

_Recap by [Repo Wrapped](https://repowrapped.com/gh/meta-pytorch/monarch?utm_source=github-action)._