99a2001e63Rename PJRT BufferType to follow Zig and ZML naming conventions.Tarry Singh2025-01-16 13:00:47 +0000
7324a49da3Remove .print() calls from globalAttnMask() and localAttnMask() in ModernBERT example to resolve compilation sharding error.Foke Singh2025-01-15 16:59:26 +0000
fbf1ecb8b7Introduce Executable.getCompiledMemoryStats in PJRT.Tarry Singh2025-01-02 16:36:13 +0000
4b1a3ff48aAdd union support to mapping helpers in zml/meta.zig.Tarry Singh2025-01-01 13:35:17 +0000
c961d705f1Set default values for operand_layouts and result_layouts in StableHLO dialect.Tarry Singh2024-12-26 09:29:45 +0000
e6286b6097Update Buffer.from to be blocking by default and add options for async loading and memory placement, adjusting aio, hostbuffer, pjrtx, and tensor implementations.Tarry Singh2024-12-25 17:14:44 +0000
bb2b77d7deCorrectly set model.norm.eps in Llama examples.Foke Singh2024-12-18 11:48:23 +0000
6aa9aa5a7bAdd preliminary implementation for custom call support.Tarry Singh2024-12-10 09:36:37 +0000
1d5b79111amodernbert: set default epsilon value for embeddings layernorm.Foke Singh2024-12-09 16:43:29 +0000
a63d0a4aa3Update example MODULE.bazel and lockfile to use the toolchains_llvm_bootstrapped configuration.Foke Singh2024-12-04 11:30:44 +0000
5464281c91Update workspace configuration to use the toolchains_llvm_bootstrapped toolchain for Zig builds.Tarry Singh2024-12-03 13:50:58 +0000
f5ab2c3a55zml: eliminate compile-time fields from Bufferized, removing the need to pass undefined to exe.call for inlined arguments. Introduce BufferizedWithArgs in zml.testing for compileAndCall utility.Tarry Singh2024-11-28 12:24:39 +0000
364a222dc1Update example MODULE.bazel and lockfile to target XLA version 20250527.0‑cb67f2f.Foke Singh2024-11-25 17:57:45 +0000
95453c7242Update XLA dependency to version 20250527.0‑cb67f2f and refresh related Bazel BUILD, MODULE, overlay and patch files.Tarry Singh2024-11-22 16:50:20 +0000
fa13287931workspace: upgrade to Zig 0.14.1 and handle empty tuple syntax &.{} being detected as *const @TypeOf(.{}).Tarry Singh2024-11-19 11:45:36 +0000
d8a83830e8runtimes: switch to Cloudflare Debian snapshots for more reliable dependency pinning.Tarry Singh2024-11-15 09:40:58 +0000
ea3ce685a9runtimes/neuron: bump runtime version and expose nrt.h header to Zig.Tarry Singh2024-11-14 13:37:47 +0000
09da9c2982Make zls.sh example explicitly set the ZLS runner target.Foke Singh2024-11-06 16:22:44 +0000
948c577205Make ZLS runner target explicit in workspace BUILD files and update the zls.sh script accordingly.Tarry Singh2024-11-04 13:57:59 +0000
47a4eda5f6runtimes/cuda: expose cuda.h in the C namespace for CUDA runtimes, enabling custom calls to CUDA functions.Tarry Singh2024-11-01 13:27:24 +0000
3849eb10b7Add buffer and hostbuffer utilities with precise f32→bf16 conversion, type inference for loadBuffers, store expected input shapes, enhance meta.visit and JSON TaggedUnion support, and improve logging.Tarry Singh2024-10-28 11:21:46 +0000
1540c6e85eUpdate loader example to demonstrate new HostBuffer helpers and type‑inferred buffer loading.Foke Singh2024-10-25 10:20:04 +0000
4ef81b89eastdx.fmt: add slice formatting support, improving on previous prettyPrinter implementation by leveraging internal fmt mechanisms.Tarry Singh2024-10-18 15:05:08 +0000
fe55c600d4Add/refresh how‑to docs and example loader for deployment, Docker, HuggingFace token, and getting‑started tutorials.Foke Singh2024-10-14 11:27:41 +0000
59f99c4501Enable Bazel debug/release Zig builds and add stdx.time.Duration.hz and stdx.math.roundeven utilities.Tarry Singh2024-10-08 09:39:00 +0000
aacbf2ee04Fix Llama3 rope scaling implementation in the neural network module (zml/nn.zig)Tarry Singh2024-10-07 12:53:03 +0000
9f61a8aacbUpdate example Zig code for llama3 rope scaling and modernbert usage.Foke Singh2024-10-04 17:49:07 +0000
d0cf5d3042Update example MODULE.bazel and its lock file to incorporate workspace fixes and version bumpsFoke Singh2024-09-30 14:00:16 +0000
4a0b1cce50Update Bazel workspace and XLA overlay (MODULE.bazel, BUILD files, patches) to prevent dual LLVM builds and apply migration/bump patches.Tarry Singh2024-09-27 14:00:44 +0000
2863c1f5e0zml/tensor: fix returned value in Tensor.toMemory – ensure _output_memory_kind is set correctly in the result.Tarry Singh2024-09-18 13:18:08 +0000
fa4a8d8de4async: enforce synchronous logging by routing async.zig logs through std.log.defaultLog to avoid panics when using async.LogFnTarry Singh2024-09-17 09:03:48 +0000
aec7072837pjrt: add FFI bindings for custom callsTarry Singh2024-09-10 09:14:28 +0000
1f5ff96c10zml/ops: add wiring for operand output alias in zml.ops.tritonTarry Singh2024-09-09 15:00:28 +0000
7e0fcecfc9async: adjust implementation for xev version bumpTarry Singh2024-09-06 16:00:58 +0000
4b7e618b43zml/aio: add bool handling in struct population within populateStructTarry Singh2024-09-02 14:11:47 +0000
ac63c30e12add mini-DSL for creating MLIR common attributes and types, leveraging Zig 0.14 to simplify mlir.Type and mlir.Attribute creationTarry Singh2024-08-26 14:19:00 +0000
4d809118d2async: detect and select the most suitable xev backend, enabling use of io_uring when supportedTarry Singh2024-08-14 14:23:57 +0000
12713ea8c7Update Bazel MODULE files to fetch LLVM, StableHLO, and Triton dependencies, aligning versions for XLA.Foke Singh2024-08-07 17:13:03 +0000
7df89301dcBump XLA version and import llvm, stablehlo, triton, and zig‑protobuf modules in workspace BUILD files.Tarry Singh2024-08-06 10:28:43 +0000
cbe6e730bdstablehlo: fix forwarding of optional operand and result layout attributes in custom callTarry Singh2024-07-31 17:53:18 +0000
44933c9b89Add macOS x86_64 support to example Bazel MODULE and platform mappings.Foke Singh2024-07-29 10:31:51 +0000
ca4e061ad5Add Bazel build configurations for macOS x86_64 CPU runtime and ZLS third‑party integration.Tarry Singh2024-07-25 15:58:14 +0000
3f36506f1czml: remove usingnamespace from floats.zig and related dependencies; note that incremental compilation does not improve overall build time due to linking overheadTarry Singh2024-07-23 17:43:43 +0000
42dee5d0e0mlir: rework stablehlo custom call implementation and add a Triton exampleTarry Singh2024-07-16 13:23:07 +0000
aec1d96e6dmlir: rework DenseElementsAttribute to correctly slice inputs and modify .as() to return a concrete value instead of an optionalTarry Singh2024-07-15 12:32:24 +0000
efcf955a4eworkspace, third_party/rules_zig: adjust ZLS to require --version as the first parameter and add missing keys to the BuildConfig object for code completionTarry Singh2024-07-10 15:20:12 +0000
b67fd2367dAdd benchmark, Llama, and ModernBERT example programs plus third‑party non‑module dependencies.Foke Singh2024-07-03 11:30:49 +0000
30f6be0e2fUpdate core Zig modules (async, mlir, pjrt, stdx) and third‑party Bazel definitions for the Zig 0.14.0 release.Tarry Singh2024-07-02 14:19:04 +0000
16cc8c6658Update examples MODULE.bazel.lock to reflect sandboxing rework and upgraded CUDA/ROCm version dependencies.Foke Singh2024-06-27 12:32:07 +0000
967eeb928fUpdate Bazel workspace and runtime configs: rework sandboxing, bump PJRT to 7.0.0, and upgrade CUDA (12.8), cuDNN (9.8), and ROCm (6.3.4).Tarry Singh2024-06-25 11:00:29 +0000
8613a6da7dworkspace: update rules_cc to 0.1.1 and Bazel to 8.1.1 to resolve yanked version issue; adjust MODULE.bazel and .bazelversion accordingly; verified successful build and executionTarry Singh2024-06-19 12:06:53 +0000
5bcaf374c8workspace: bump libxev to 20250222.0-07bcffa and introduce Dynamic to select between IO Uring and Epoll on Linux based on availabilityTarry Singh2024-06-18 14:26:44 +0000
18eb0e5a7bAdd async I/O, SentencePiece, NN, and tensor utilities for ModernBERT support and update Bazel build configuration.Tarry Singh2024-06-14 15:27:06 +0000
17d02621e7Add tests for the ModernBERT example, covering activation utilities, build setup, and example Zig code.Foke Singh2024-06-11 17:33:22 +0000
05944b5cc9Update FnCache to copy and reuse non‑tensor fields in fixed‑size structs, preventing undefined memory in core modules.Tarry Singh2024-05-15 17:54:52 +0000
dfe55b0d34Update Bazel lock file for examples to reflect FnCache non‑tensor handling changes.Foke Singh2024-05-13 16:59:37 +0000
26558d6201Update examples MODULE.bazel and lockfile to use XLA 20250204.0-6789523 and ensure Bazel 8 compatibility.Foke Singh2024-05-08 14:03:45 +0000
f5ab6ff2c6Update XLA to version 20250204.0-6789523 and adjust Bazel module and runtime files for Bazel 8 compatibility.Tarry Singh2024-05-03 15:57:56 +0000
190c6978d2llama: simplify llama3 prompt template encoding by removing redundant newline re-encoding and ensuring a trailing newline.Foke Singh2024-04-10 09:36:28 +0000
d4db5ccc6bIntegrate TinyLlama support, restore the homemade tokenizer, and align Zig API naming across stdx and zml tokenizer modules.Tarry Singh2024-04-05 15:07:29 +0000
b67685b941Add example Bazel build files and tokenizer test for tinyllama, including tigerbeetle integration and flags.Foke Singh2024-04-01 17:40:18 +0000
567210d1d7bazel: depend on prebuilt protoc binaries to eliminate ~1300 build steps. Note: integration is currently blocked due to version constraints in rules_proto and toolchains_protoc.Tarry Singh2024-03-29 09:54:57 +0000
e0c8eecb79bazel: use OID as sha256 for Git LFS files to prevent unnecessary HuggingFace redownloads.Tarry Singh2024-03-28 17:52:52 +0000
a811b2e1e3llama: fix dimensions and data types Removed unnecessary batching dimension introduced by recent changes. Converted index outputs from i32 to u32 for token indices. Ensures Llama runs on CUDA and RoCM. Tested on CUDA.Foke Singh2024-03-20 13:37:19 +0000
602757e7a9Update examples to use the corrected logFn API.Foke Singh2024-03-18 13:11:14 +0000
754656f2f0Replace real mutex with async Mutex for logFn, add fallback logger support outside coroutines, and fix ResetCondition handling.Tarry Singh2024-03-14 11:43:33 +0000
980f1b17fbEnsure all runtime plugins have correct SONAME values, fixing issues with prebuilt PJRT plugins.Tarry Singh2024-03-11 10:15:22 +0000
8a25b1eb74Revert CUDA PJRT plugin version to 0.4.38 to address performance regression on XLA master.Tarry Singh2024-03-05 17:04:42 +0000
76e314db9bUpdate Llama example docs and Bazel build files, and add tests for the new HuggingFace tokenizer integration.Foke Singh2024-03-04 12:11:13 +0000
959bc48c42Add HuggingFace tokenizer bindings and SentencePiece integration; update BUILD files, async utilities, and FFI modules to support the new tokenizers.Tarry Singh2024-02-28 15:47:37 +0000
5048e7dc89Update example lock file for rules_distroless 0.4.2 upgrade and verify MNIST image build works.Foke Singh2024-02-26 15:30:13 +0000
b4b2490690Upgrade rules_distroless to 0.4.2 in MODULE.bazel and refresh MODULE.bazel.lock accordingly.Tarry Singh2024-02-21 17:48:10 +0000
c109b12e1bVarious minor fixes: rewrite tinyllama tokenizer newline token, prevent HostBuffer.isContiguous false trigger on 1‑dim axes, improve HostBuffer.slice1d error messages, simplify module.zig output to show .mlir file path, correct setFlags handling of comptime int/float, make tokenizer.zig return <oob> for out‑of‑range detokenization, and speed up Buffer.constant creation up to 2.5 GB/s on CUDA.Tarry Singh2024-02-19 12:34:18 +0000
3970df5b48Update getting_started tutorial and example Bazel files for Bazel 8 migration.Foke Singh2024-02-14 10:44:47 +0000
169a24307cMigrate workspace and XLA module definitions to Bazel 8, updating MODULE.bazel files, BUILD rules, and related migration patches.Tarry Singh2024-02-12 12:43:23 +0000
7e6103d876Upgrade XLA to version 20250122.0-cc075be, switch to nvptx compiler and nvlink with nvjitlink support, add warning for CUDA path in LD_LIBRARY_PATH, and revert the previous CUDA sandbox fix.Tarry Singh2024-02-06 09:31:48 +0000
b8a0aaee5aUpdate tokenizer to handle byte_fallback for Llama3 GPT2 vocab and add a Llama3‑specific normalizer; adjust tinyllama.zig and hostbuffer.zig to use the new tokenization logic.Tarry Singh2024-02-05 15:22:44 +0000
Tarry Singh
Foke Singh