mirror of
https://github.com/hanzoai/ml.git
synced 2026-11-19 23:50:51 +00:00
Three-way per file through the rename: base 39355c6c91, theirs the upstream
merge of main with 35d7ae7c (via 7ef6a694d), ours origin/main.
Lands the barrier thread pool, runtime CPU feature dispatch, aarch64 and x86
repacked quantized kernels (indexed_gemv, gemv_fused_shared_lhs), quantized
embedding on CPU/Metal/CUDA, paged flash-attn kernels, cutile fused MoE, the
in-place op aliasing API, parking_lot storage locks, Metal buffer builders and
debug labels, safetensors 0.8, hf-hub 1.0, tokenizers 0.23.
Ours keeps ROCm, Vulkan, wgpu, the IQ/MXFP4/NVFP4/TQ dtypes, and the i-quant
weight/activation block split in matmul. Our backends ride InplaceOpN and the
quantized embedding; Metal get_rows gains the IQ kernels. ug leaves, as upstream.
Co-authored-by: Hanzo Dev <[email protected]>
hanzo-ml-replit-code: code completion specialized model.
replit-code-v1_5-3b is a
language model specialized for code completion. This model uses 3.3B parameters
in bfloat16 (so the GPU version will only work on recent nvidia cards).
Running some example
cargo run --example replit-code --release -- --prompt 'def fibonacci(n): '
This produces the following output.
def fibonacci(n): # write Fibonacci series up to n
"""Print a Fibonacci series up to n."""
a, b = 0, 1
while a < n:
print(a, end=' ')
a, b = b, a+b
print()
def fibonacci_loop(n): # write Fibonacci series up to n
"""Print a Fibonacci series up to n."""
result = []
a, b = 0, 1
while a < n:
result.append(a)
a, b = b, a+b
return result
def fibonacci_generator(n): # write Fibonacci series up to n
"""Print a Fibonacci series up to n."""
a, b = 0, 1
while a < n:
yield a
a, b = b, a+b