|
|
f00538667e
|
zml.nn: add dynamic sampling with support for top‑k, top‑p, and min‑p settings. Implements token index computation based on the selected sampling strategy, including options for top_k, max_top_k, top_p, and min_p.
|
2023-06-16 14:34:18 +00:00 |
|
|
|
b244a18621
|
zml: set iota default dtype to .i32, with fallback to .i64 for axes with many elements, simplifying usage.
|
2023-06-15 12:45:52 +00:00 |
|
|
|
344e07fb6e
|
stablehlo: extend dot_general API to include DotAlgorithm support by merging precision and algorithm attributes into a union, aligning with spec requirements. Currently not exposed to users due to limited algorithm support.
|
2023-06-07 11:20:25 +00:00 |
|
|
|
6d720126ac
|
Add PJRT custom call integration with generic zmlHostBufferCallback to copy tensors to host and invoke user callbacks. Introduce Tensor.print() method to output runtime tensor values (CUDA‑specific, uses a pre‑allocated host buffer).
|
2023-06-05 13:42:45 +00:00 |
|
|
|
bf23eef0d9
|
examples: clean up inconsistencies in asynk usage across the codebase.
|
2023-06-01 16:11:58 +00:00 |
|
|
|
499b0d20e5
|
pjrtx: change behavior to return an error when OpenXLA fails to serialize the new batching_dim attribute for gather/scatter, instead of panicking.
|
2023-05-29 17:18:19 +00:00 |
|
|
|
52ef20f981
|
zml: reintroduce pjrtx to handle reactor blocking issues in async scenarios, particularly with Events.
|
2023-05-26 15:54:15 +00:00 |
|
|
|
c68ec4bc5c
|
async: implement default threaded backend using a thread pool. Backend selectable via @zml//async:impl flag (threaded or zigcoro). Provides workaround for environments where io_uring is unavailable.
|
2023-05-25 16:02:11 +00:00 |
|
|
|
89cf2233d3
|
zml/aio: enable reading metadata from index.json for sharded safetensor files, allowing metadata storage alongside model config.
|
2023-05-23 15:06:59 +00:00 |
|
|
|
2f54e2a5f3
|
zml.tensor: add triangular operator to zero out the upper‑right matrix region with configurable offset, and toDiagonal (diag_embed) to embed a vector as a diagonal matrix, correcting previous diag naming. Also add ELU activation under zml.nn.Activation.
|
2023-05-18 16:39:21 +00:00 |
|
|
|
05faa5021e
|
zml.tensor: add cumulativeSum operator and refactor maxPoolND. Introduce cumulative sum using reduceWindow. Simplify reduceWindow signature by merging padding_shape and padding_value. Update maxPool1D/2D to accept tuple arguments. Revise pad to use tagged or AOS syntax; remove SOA syntax.
|
2023-05-17 09:01:27 +00:00 |
|
|
|
54e7eb30b4
|
Introduce a thin abstraction layer between ZML and PJRT to manage plugin loading decisions, enable compile‑time detection of linked runtimes, and handle cases such as libtpu blocking metadata access.
|
2023-05-15 09:36:41 +00:00 |
|
|
|
74e90855ca
|
Configure the runfiles environment globally at context start to ensure Bazel-built binaries locate their runfiles correctly.
|
2023-05-12 11:40:23 +00:00 |
|
|
|
57130577e9
|
Add fallback for runtimes lacking PJRT_Event by using thread‑pool dispatch for buffer copies and treating operations as synchronous when events are absent.
|
2023-05-09 12:44:56 +00:00 |
|
|
|
672df8fa2f
|
Update tutorial and example code to use the new asyncc name and Generic slugs.
|
2023-05-08 16:58:45 +00:00 |
|
|
|
5543c8192f
|
Rename async_ to asyncc and add Generic async slugs in async.zig, aio.zig, and module.zig.
|
2023-05-04 14:44:12 +00:00 |
|
|
|
cfe38f27ca
|
Switch ROCm dlopen handling to patchelf's rename_dynamic_symbols for more robust dynamic symbol import.
|
2023-05-03 17:33:46 +00:00 |
|
|
|
fefd84b1bb
|
Replace silu implementation with stablehlo.logistic for higher precision, move logistic logic into sigmoid and alias logistic to sigmoid (breaking change).
|
2023-05-01 10:40:50 +00:00 |
|
|
|
021111d07d
|
Extend tests to handle all float types, preventing crashes with bfloat16 tensors.
|
2023-04-27 10:34:27 +00:00 |
|
|
|
e0fd7f8d97
|
Fix typographical errors in the documentation.
|
2023-04-25 16:04:09 +00:00 |
|
|
|
477e13afd0
|
Add missing zig_cc_binary import to the simple layer example in the documentation.
|
2023-04-24 10:04:50 +00:00 |
|
|
|
ed6444b775
|
Add Tensor.concatenate support, begin deprecating broadcastLeft, and compute transformer head scaling constant in f32 for higher precision.
|
2023-04-21 15:55:07 +00:00 |
|
|
|
11006ca08d
|
Refactor torch module: merge PickleData into Parser as torch.File, rename value file to py_object.zig, use buffered reader for pickle and zip headers, adjust intermediate result handling, simplify Python dict representation, separate kwargs from args, and add extensive tests for long integers, protocol 0, zipped pickle, and a complex PyTorch Conv2d case; also streamline BufferStore initialization.
|
2023-04-20 15:43:18 +00:00 |
|
|
|
837f8fb111
|
Add support for the Llama 3.1 70B Instruct model to facilitate testing on high‑performance accelerators.
|
2023-04-19 10:23:44 +00:00 |
|
|
|
fdb7da5c9b
|
Introduce sharding attributes to Llama weights to enable Tensor Parallelism.
|
2023-04-13 12:35:27 +00:00 |
|
|
|
833ff5f28d
|
Upgrade PJRT CUDA Plugin to version 0.2.3, adding NCCL support for correct sharding.
|
2023-04-12 15:47:06 +00:00 |
|
|
|
8e43a45a3c
|
Add event waiting when invoking a module and improve multi‑device sharding handling.
|
2023-04-11 11:32:09 +00:00 |
|
|
|
0189b71070
|
Rename zml.aio.Value to zml.aio.Metadata, simplify its type variants, and update torch pickle/eval APIs accordingly.
|
2023-04-07 16:45:58 +00:00 |
|
|
|
aea23c720e
|
Update Llama example to use renamed zml.aio.Metadata (formerly Value) and reflect torch loader changes.
|
2023-04-05 14:09:59 +00:00 |
|
|
|
e25f70d923
|
Rename and simplify modules in zml/aio/torch: replace redundant qualified names, remove generic utilities, inline code, reorder functions for top‑to‑bottom readability, and extract parsing logic into parseTensor and parseStorage functions.
|
2023-04-04 17:20:53 +00:00 |
|
|
|
66881899ca
|
Fix testLayer by removing unnecessary compile_options argument and updating testing logic for new sharded output, ensuring proper usage by llama.zig.
|
2023-03-31 14:23:45 +00:00 |
|
|
|
05d23beb23
|
Add Normalizer.fromHfJson to read HuggingFace tokenizer JSON and map to internal options, including a configurable magic space token and a debug flag for token merges. Adjust default handling of extra whitespaces to align with HF defaults.
|
2023-03-29 16:10:29 +00:00 |
|
|
|
ef922e3aea
|
Fix empty JSON array handling in safetensor metadata loader and refactor torch loader (make ops slices const and improve readability).
|
2023-03-28 16:17:00 +00:00 |
|
|
|
aae37738a5
|
Update loader example to demonstrate handling of empty JSON arrays and improved torch loader readability
|
2023-03-22 14:52:33 +00:00 |
|
|
|
a4f0fc96c0
|
Integrate user sharding hints and HLO sharding annotations across MLIR dialects and ZML core, and remove the now‑unused module options arguments.
|
2023-03-21 10:50:39 +00:00 |
|
|
|
e30e35deeb
|
Update benchmark example to use new user sharding hints and drop deprecated module options.
|
2023-03-20 15:31:44 +00:00 |
|
|
|
8746a5ce78
|
Expose zml/test_runner.zig publicly to enable users to employ the async test runner. Made the dependency on zml explicit and suggest treating test_runner as a zig_library rather than a filegroup.
|
2023-03-16 13:22:35 +00:00 |
|
|
|
fe531aef06
|
Clarify HuggingFace token handling in workspace, noting the standard CLI location and adding support for an environment variable.
|
2023-03-14 15:28:03 +00:00 |
|
|
|
cd2f2209d0
|
Create token directory if it does not exist.
|
2023-03-13 15:31:13 +00:00 |
|
|
|
70d40208a2
|
runtimes/cuda: Fix version variable definitions in the build script to enable successful CUDA builds.
|
2023-03-09 11:31:02 +00:00 |
|
|
|
7ef67eea27
|
zml: Relocate tests next to the functions they verify and remove obsolete dynamicSlice1d test.
|
2023-03-08 14:10:11 +00:00 |
|
|
|
dfa71018a5
|
zml: Remove pjrtx wrapper, migrate remaining helpers to their native modules, and fix blocking issue in Event.await.
|
2023-03-06 17:05:56 +00:00 |
|
|
|
0c126c2e12
|
runtimes/cuda: Upgrade CUDA to 12.6.2 and cuDNN to 9.4.0.
|
2023-03-03 15:17:26 +00:00 |
|
|
|
f595d22134
|
runtimes/rocm: Upgrade ROCm to version 6.2.2.
|
2023-03-01 13:15:50 +00:00 |
|
|
|
ecf52ad724
|
zml.tokenizer: Implement proper byte fallback support by converting hex byte strings (e.g., “<0x40>”) to their characters and splitting unknown UTF‑8 codepoints into bytes, fixing tokenization.
|
2023-02-28 14:40:25 +00:00 |
|
|
|
2f129f76c9
|
Add in-process sharding support across core ZML components (platform, shape, tensor, MLIR generation, buffers, and PJRT integration)
|
2023-02-24 17:33:14 +00:00 |
|
|
|
cad1a688da
|
Add sharding usage to the benchmark and simple_layer example programs.
|
2023-02-23 11:18:27 +00:00 |
|
|
|
fc718ab649
|
Add StableHLO bindings for versioning functions, enabling portable serialization of StableHLO.
|
2023-02-22 15:41:33 +00:00 |
|
|
|
8fa3878fc3
|
PJRT: Add handling for rank‑0 case in getDimensions to avoid null pointer usage when num_dims is zero.
|
2023-02-17 10:47:15 +00:00 |
|
|
|
639f5cd994
|
Replace log with select for generating the attention mask to avoid NaNs on zero values.
|
2023-02-16 10:36:23 +00:00 |
|
