- sample the first token from prefill logits without duplicating the prompt tail
- reject incomplete multi-output records before preprocessing alignment
- cover cached generation and partial DPO records with regression tests
- Extract _step() as the single prefill-group + task_extend + decode primitive
- _run_generation_loop and run_batch now both call it, so the two cannot drift
- run_batch now records prefix hashes (paged mode) and uses input order for
decode, matching the loop thread
- hoist prefill qo_indptr into the workspace so CudaBackend.fwd_prefill does not rebuild it per layer
- cache has_freq in SamplingBatchInfo to drop the per-step GPU any() sync
- drop pin_memory host staging for input_ids; sync copy suffices for a small batch
- add InferenceWorkspace with fixed-shape per-step buffers (input_ids, decode mask, KV bind metadata) for CUDA-graph capture
- bind_tasks derives seq_lens from the pool's own _task_len tracking, dropping the seq_lens parameter
- update decode metadata in-place (position_ids, seq_lens, kv_indptr) instead of re-allocating per step
- task_extend advances _task_len in contiguous mode so the pool tracks current length
- skip log_softmax when logprobs are not requested
- SamplingBatchInfo: sample params built once per task set (top_k int32, pinned async H2D)
- position_ids advances by +1 on steady-state decode instead of re-building
- DecodeBindCache: bind_tasks increments seq_lens/kv_indptr, reuses req_pool_indices
- saves ~240us of python/launch overhead per decode step
- bind_tasks builds kv_indptr (prefix sum of seq_lens) a single time
- fwd_decode/fwd_prefill reuse it instead of rebuilding per layer
- Removes 24 cumsum launches per decode step (was ~1ms/step at B=4)
- Decode B=4: 9.60 -> 7.82 ms/step (-18.5%), +22.8% tok/s
- CudaBackend.fwd_decode passes attn_mask directly instead of kv_cache.decode_mask
- TorchNativeBackend derives pos_mask from attn_mask[:,0,0] on decode
- Drop decode_mask and page_table fields from KVCache and bind_tasks
- Precompute page_table and decode_mask on KVCache once per step in PagePool.bind_tasks, instead of per-layer in CudaBackend/TorchNativeBackend
- Skip frequency penalty history tensor construction when all penalties are 0 in Executor.execute_decode
- Omit FrequencyPenaltyStrategy from sampling pipeline when penalty is 0
- Deduplicate get_active_tasks calls in scheduler loop (3 to 1), remove redundant sorted() on decode tasks
- Benchmark (L20, bf16, CUDA backend): B=1 9.48->9.40ms (+1%), B=4 10.73->9.89ms (+8.6%), B=8 10.77->10.13ms (+6.4%)
- Replace all buffer[layer_id][loc] double indexing with buffer[layer_id, loc] single advanced indexing in cache.py and attention.py
- Revert KVStorage buffers back to 4D [n_layers, size, n_kv_heads, head_dim], remove leftover 3D reshape/view in MLA path
- Update docs/guides/inference.md, docs/developer/internals.md, docs/developer/architecture.md to reflect new PagePool/KVStorage/ReqToTokenPool/KVCache classes
- Replace hand-rolled StreamDecoder (O(n^2) full-history re-decode per token) with tokenizers.decoders.DecodeStream
- Keep O(1) bounded token buffer internally via prefix drain instead of accumulating all token IDs
- Simplify flush_remaining to no-op since stream always emits completed text per step
- Benchmark on 8000 tokens: 2305ms -> 3.9ms (~592x speedup)
- Replace .tolist() calls with _total_len in gather(); move _slot_len updates from per-layer write to once-per-step bind_tasks
- Use torch.as_tensor instead of torch.tensor in decode penalty history construction
- RolloutGenerator._instruction_to_messages builds system/user/assistant list (instruction->system, input->user, output->assistant), replacing single-user-turn concatenation
- Remove _iter_samples helper; _prepare_prompts zips parallel list-of-strings fields directly per the collate_fn contract
- Tests adopt a system-aware chat template and pin the three-field role mapping
- Drop unused imports caught by ruff F401 (torch.Tensor in scheduler.py, iter_raw_records in pipeline.py, Tuple in evaluate_rouge.py)
- RolloutGenerator now delegates prefill/decode to InferenceScheduler.run_batch (sync API, no background thread), sharing one KV-cache code path with the inference server and eliminating O(n^2) recompute in rollout
- Add sample(return_logprobs=) and Executor.execute_decode(return_logprobs=) to expose behaviour-policy log-probs through the engine; Task gains output_logprobs
- RolloutResult now subclasses RawRollout (adds rewards only), removing duplicated fields
- RolloutRunner.__call__ returns (result, is_fresh) instead of relying on object identity, removing the fragile refresh-detection contract
- Remove O(n^2) generate_responses helper and dead code (_tokenize_prompts, unused old_model arg)
- train_context.py wires InferenceScheduler directly instead of hand-rolling SamplingPipeline
- Tests: +11 covering return_logprobs, run_batch, and KV-cache-backed rollout semantics; 404 pass
- Add FrequencyPenaltyStrategy (logit -= penalty * count)
- Per-task rep_window for penalty history lookup
- Wire through engine, task, executor, API layer
- Add --frequency_penalty and --rep_window to stream_chat.py
- 9 unit tests for frequency penalty strategy
- Add KVCache/CacheView abstract base classes in cache.py
- Add ContiguousCache (contiguous per-slot buffer, default) alongside PageCache (paged, renamed from old KVCache)
- Merge make_table_tensor + bind into bind_tasks on KVCache interface
- Remove task_cached/task_record_hashes from base class (PageCache-only)
- Scheduler: decode all position groups instead of just the largest (eliminates 63% group skip rate)
- Scheduler: accept optional cache param for swapping implementations
- Model layer type hints use CacheView base class
- Batch 1-32: 1-7% speedup from eliminating Storage.gather overhead
- All 183 inference tests pass
ViperEkura