Skip to content

Pull requests: huggingface/candle

New pull request New
239 Open 2,268 Closed
239 Open 2,268 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Fix causal attention in candle-flash-attn-v3: NULL tile_count_semaphore crash (dense) + silently non-causal varlen
#3606 opened Jun 11, 2026 by Xueying-VirtueAI Loading…
fix(transformers): adds support for llamas head_dim config
#3602 opened Jun 11, 2026 by suniastar Loading…
fix(kernels): Compilation errors on Turing Cards (Compute Cap 7.5)
#3601 opened Jun 11, 2026 by suniastar Loading…
Optimize CUDA kernels
#3600 opened Jun 11, 2026 by guoqingbao Contributor Loading…
Metal: fix SDPA storage offsets
#3599 opened Jun 11, 2026 by EricLBuehler Member Loading…
Parameter cache for CUDA kernel launch
#3598 opened Jun 11, 2026 by guoqingbao Contributor Loading…
1
Fix 128x softmax_lse over-allocation in the dense flash-attn-v3 path
#3597 opened Jun 10, 2026 by Xueying-VirtueAI Loading…
flash-attn: launch kernels on the caller's CUDA stream
#3596 opened Jun 10, 2026 by jnises Contributor Loading…
ONNX: Dequantize Linear operator implementation
#3592 opened Jun 7, 2026 by Kwasus33 Loading…
1
Add softmax_last_dim backward pass
#3591 opened Jun 7, 2026 by HueCodes Contributor Loading…
1
fix(gguf): bound tensor size against file before allocating
#3588 opened Jun 7, 2026 by pjdurden Loading…
9
Support effective Gemma 4 text weights
#3587 opened Jun 6, 2026 by HueCodes Contributor Loading…
1
fix(qwen3): build causal mask batch-independently (#3582)
#3586 opened Jun 6, 2026 by pjdurden Loading…
1
5
Harden GGUF loader bounds checks
#3583 opened Jun 5, 2026 by ryzhov-artem Loading…
Remove redundant creations and operations on zero tensors in Tensor::backward()
#3578 opened Jun 4, 2026 by TeunVerstraaten Loading…
feat(candle-nn): QuantizedKvCache — INT8 KV cache with attention sinks (TurboQuant)
#3577 opened Jun 3, 2026 by aryanputta Loading…
1
fix(flash-attn): bump vendored FA2 kernels from Dec 2024 to v2.8.3
#3576 opened Jun 3, 2026 by aryanputta Loading…
1
2
⁠fix(core): validate shape element count in ShapeWithOneHole blanket impl (#3534)⁠
#3572 opened May 30, 2026 by oyoyo4556 Loading…
1
2
Onnx nonzero implementation
#3571 opened May 28, 2026 by Kwasus33 Loading…
1
metal: implement upsample_nearest1d kernel support
#3563 opened May 24, 2026 by oglego Loading…
Add QTensor row dequantization helper
#3562 opened May 23, 2026 by rabbitson87 Loading…
feat: add candle-flash-attn-v4 crate for FlashAttention-4 support
#3561 opened May 22, 2026 by PhoenixCPH Loading…
1
Improve Sequential based on ModuleT to make it more versatile.
#3560 opened May 22, 2026 by Taswen Loading…
1
Add Mamba-3 GPU operator with SISO and MIMO inference support
#3557 opened May 22, 2026 by PhoenixCPH Loading…
1
[Metal] Expose kernel_mul_mm_id wrapper and parallelize rowids fan-out
#3555 opened May 20, 2026 by fiorelorenzo Loading…
ProTip! Mix and match filters to narrow down what you’re looking for.