Repository Issue Activity (beta)

#20179 🌐 [i18n-KO] Translating docs to Korean

#48232 Import module

#47994 Zamba2: `num_mem_blocks > 1` checkpoints cannot be constructed — `block_id` uses the global layer index while the tie cycle uses hybrid order

#46656 Inconsistent type annotation in `EvalPrediction`

#47030 FineGrainedFP8 DeepGEMM path returns silently wrong results on SM100 (UE8M0 scale rounding without requantization)

#47031 Checkpoint conversion silently loads transposed fused-expert weights when the swapped dims are equal (Transpose(check_dims=True), qwen3_vl_moe)

#47085 Fast (Rust-backed) tokenizers aren't thread-safe anymore

#47317 PPFormulaNet training-loss bug

#47378 Question about overriding bnb_4bit_compute_dtype for pre-quantized checkpoints

#48172 [Mllama] MllamaVisionAttention and MllamaTextCrossAttention never set self.is_causal — cross-attention is silently causal under sdpa

#33787 Can't instantiate abstract class Pipeline with abstract methods

#48080 Zamba-7B-v1 hangs during torch.compile due to use_associative_scan=False

#48222 [serge] integration failure triage - 2026-08-22

#48225 [BUG] MLukeTokenizer emits `<s>` instead of `<ent>` on all entity tasks

#48202 [serge] integration failure triage - 2026-08-21

#48210 Should Trainer support models FSDP2-sharded at load time (DistributedConfig(fsdp_size=N))?

#48148 Installed nested optional kernels silently fall back to the Torch implementation

#48206 AutoTokenizer silently corrupts deepseek-math-7b-rl in 5.x: encode strips all spaces, decode emits literal byte-level markers

#48193 Add native support for TR-HASH deterministic token-routed MoE

#46614 Wav2Vec2PhonemeCTCTokenizer.phonemize keeps the previous language's espeak backend after switching back (stale phonemizer_lang)

#46822 ZeRO-3 zero.Init does not partition composite minimax_m3_vl language submodule -> OOM on multi-GPU load

#47477 Regarding security issue i have reported

#47688 Heads that declare **kwargs but drop num_items_in_batch train with an inflated loss under gradient accumulation

#48165 [serge] integration failure triage - 2026-08-20

#14497 FLAX core dump error on CloudTPU when running run_clm_flax.py