Issues

M3: read_parquet is not in the extension, because dlopen does not survive the shared library build Open

area/python packaging

tamnd/firepanda#211 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

Align two series on a duplicated row label, which pandas joins rather than unions Open

tamnd/firepanda#244 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

Adding two text columns concatenates in pandas and raises here Open

tamnd/firepanda#249 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

Integer division by zero is null here and infinity in pandas Open

tamnd/firepanda#252 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

A series broadcast along a frame's columns needs text labels, so s + df raises Open

tamnd/firepanda#253 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

df.columns is a list where pandas gives an Index Open

tamnd/firepanda#254 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

A Python scalar widens the column it is added to, where pandas leaves it alone Open

tamnd/firepanda#255 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

df.columns, df.index and s.index cannot be assigned to Open

tamnd/firepanda#256 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

L. The four column types three conformance sections are waiting on Open

tamnd/firepanda#270 · 4 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

A dictionary column cannot cross the Arrow C Data Interface Open

tamnd/firepanda#277 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

Write a list or a struct column to Arrow Open

tamnd/firepanda#284 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

Join a nested column that arrives in more than one record batch Open

tamnd/firepanda#285 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

Mojo miscompiles an Optional of a struct with trailing padding, and AnyArray pads around it Open

tamnd/firepanda#286 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

M. The dt accessor, 42 names on a column that can be read and not asked Open

tamnd/firepanda#287 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

TPC-H, all twenty two queries, and the operations they need that we do not have Open

area/benchmark area/io area/kernel area/string enhancement performance

tamnd/firepanda#299 · 10 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

The constructor door: a None and a NaN, after the read path was decided Open

tamnd/firepanda#301 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: --cpu-offload-gb is accepted, reported in non-default args, and then NoopOffloader is selected: weights are never offloaded Open

bug quantization

vllm-project/vllm#56283 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: Persistent KV offload reuses blocks after changing RoPE scaling Open

vllm-project/vllm#56311 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Usage]: per-request metrics count parallel-sampling children, usage counts the request Open

vllm-project/vllm#56361 · 4 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: Qwen3-VL fails when using modality-scoped image/video size kwargs (`images_kwargs` / `videos_kwargs`) Open

bug

vllm-project/vllm#56363 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: Batch invariance is broken when sequence parallelism / async TP is enabled (`VLLM_BATCH_INVARIANT=1` + `pass_config.enable_sp`) Open

bug

vllm-project/vllm#56370 · 13 comments · 1 reaction · 0 assignees · Indexed Sep 15, 2026

[Bug]: /v1/messages non-streaming omits stop_sequence when null (exclude_none) Open

bug

vllm-project/vllm#56371 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug][ROCm] GLM-5.3-Flash indexer block-table mismatch on gfx942 Open

glm rocm

vllm-project/vllm#56380 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: [Bug][Docker] Recipe references vllm/vllm-openai-rocm:deepseekv41-flash-0909 but the image is not available on Docker Hub Open

bug kimi rocm

vllm-project/vllm#56384 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: DeepSeek-V4.1-Flash dsv4_topk Triton illegal memory access under high concurrency on H20; mitigated by max_num_seqs=256 Open

deepseek DSv4.1

vllm-project/vllm#56389 · 8 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: DeepSeek-V4.1-Flash: SimpleCPUOffloadConnector crashes with `assert block_size % hash_block_size == 0` (block_size=8, hash_block_size=32) Open

bug deepseek DSv4

vllm-project/vllm#56396 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[CI Failure]: Multimodal Processor shards - test_tensor_schema[naver-hyperclovax/HyperCLOVAX-SEED-Think-32B] - upstream model_type hyperclovax_vision_v2 unknown to CI transformers Open

ci-failure multi-modality rocm

vllm-project/vllm#56397 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Model Support] DeepSeek-V4.1 Tracking Issue Open

deepseek documentation DSv4.1 rocm

vllm-project/vllm#56400 · 1 comment · 3 reactions · 0 assignees · Indexed Sep 15, 2026

[RFC]: Efficient Routed-Expert Replay with Prefix Omission and KV Cache Offloading Open

RFC

vllm-project/vllm#56402 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Build] Extract legacy HIP and MoE extension CMake configuration Open

rocm

vllm-project/vllm#56407 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: --cpu-offload-gb uses ~1.9x the host memory it asks for Open

bug

vllm-project/vllm#56410 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: extract_hidden_states uses first-column feedback for multi-token outputs Open

bug

vllm-project/vllm#56417 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: CPU Gated-DeltaNet + MTP — structured output kills EngineCore via two distinct paths Open

speculative-decoding structured-output tool-calling

vllm-project/vllm#56419 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug][XPU]: Kimi-K3 MLA fused ops not registered on XPU — Kimi-Linear-48B crashes on first attention forward Open

intel-gpu k3 kimi

vllm-project/vllm#56422 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: DeepSeek-V4.1-Flash + DSpark spec decode hits CUDA device-side assert in `map_draft_to_target` at draft warmup on SM90 (H200) with Marlin MXFP4 MoE backend Open

deepseek DSv4

vllm-project/vllm#56443 · 5 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug] Qwen4Exp QSA indexer: per-chunk logits buffer grows with max_seq_len, caching allocator keeps every size, device OOM/hang on unified-memory GB10 (SM121) during long prefill Open

vllm-project/vllm#56457 · 4 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: DeepSeek-V4.1-Flash cannot serve on SM120/SM121 (GB10) at e77daef89 — SWA cache block 32 vs SM120 decode page 64; ratio-1 indexer block_kv=128 vs DeepGEMM sm120 (64 only) Open

deepseek DSv4

vllm-project/vllm#56461 · 4 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: DSML tool-call recovery (#55954) emits ghost empty-argument calls on truncation and narration, and recovers undeclared tools Open

tool-calling

vllm-project/vllm#56482 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: activation override in test_flashinfer_cutedsl_fp4_moe silently tests RELU2 for every parametrize case Open

vllm-project/vllm#56504 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[RFC]: DeepSeek-V4.1-Flash performance on ROCm Open

deepseek DSv4.1 performance quantization RFC rocm

vllm-project/vllm#56506 · 9 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Doc]: ec_cpu_connector.md refers to a no-NIXL import test that doesn't exist Open

kv-connector

vllm-project/vllm#56507 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Doc]: metrics page is missing the spec-decode counters and cache_config_info Open

vllm-project/vllm#56514 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Doc]: prompt_lookup_min/prompt_lookup_max docstrings give the wrong defaults Open

vllm-project/vllm#56516 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug][ROCm]: Intermittent worker segfault in libhsa-runtime64 — ROCPROFILER_QUEUE_INTERPOSITION=0 forces an unfixed ROCr heap overflow Open

rocm

vllm-project/vllm#56521 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: malformed EXIF in an image fails the request with a 500 from `MultiModalHasher.serialize_item` Open

vllm-project/vllm#56527 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: ROCm Stack DeepSeekv4.1 Flash Issue spamming logs with GLUON backend not available. Using TRITON backend!!! Open

bug rocm

vllm-project/vllm#56540 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: cuda_archs_loose_intersection drops plain "10.0" when TORCH_CUDA_ARCH_LIST also has "10.0a", producing sm_100a-only kernels that fail on B300 (CC 10.3) Open

vllm-project/vllm#56548 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: JSON schema with multiple allOf branches is silently ignored by the xgrammar structured-output backend Open

bug structured-output

vllm-project/vllm#56556 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: Unconditional xgrammar import in backend_xgrammar.py crashes vLLM startup on s390x Open

bug structured-output

vllm-project/vllm#56559 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026

[Bug]: FLASH_ATTN_MLA_SPARSE emits 'Failed to initialize the TMA descriptor 1' every step for GLM-5.3-Flash (zero-extent descriptors; output correct) Open

glm

vllm-project/vllm#56563 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026

Issues

Get new issues in your inbox