docs: sync docs with current code after refactor

- architecture: remove TaskManager.max_prompt_len (merged into max_seq_len in 53c804e)
- dataflow: fix DatasetFactory.load param name max_position_embeddings -> max_len
- params: add fsdp2 to parallel_mode, add --max_seq_len to server, add 4 missing generate options
- preprocessing: add missing batch_size config field

assets/docs/architecture.md

@@ -924,7 +924,6 @@ classDiagram
             +AutoTokenizer tokenizer
             +int max_batch_size
             +int max_seq_len
-            +int max_prompt_len
             +Deque waiting_queue
             +List active_tasks
             +add_task(prompt, max_tokens, temperature, top_p, top_k, stream_callback) str

assets/docs/dataflow.md

@@ -83,9 +83,11 @@ All backends normalise tensors into `Store._data[Dict[str, List[Tensor]]]` + `St
 ## Dataset Architecture
 
 ```
-DatasetFactory.load(train_type, load_path, window_size, stride=None,
-                    storage_type=None, tokenizer_path=None,
-                    max_position_embeddings=2048, store=None)
+DatasetFactory.load(
+    train_type, load_path=None, window_size=0, stride=None,
+    storage_type=None, tokenizer_path=None,
+    max_len=2048, store=None
+)
   → BaseDataset.load(load_path, storage_type=None)
     → detect_format(load_path)
     → StoreFactory.create(storage_type)

assets/docs/params.md

@@ -84,7 +84,7 @@ Combined optimizer: matrix parameters via **Muon**, non-matrix via **AdamW** (`f
 | Parameter | Description | Default |
 |-----------|-------------|---------|
 | `--nprocs` | Number of GPUs / processes | 1 |
-| `--parallel_mode` | Parallel strategy (`none`, `ddp`, or `fsdp`) | none |
+| `--parallel_mode` | Parallel strategy (`none`, `ddp`, `fsdp`, or `fsdp2`) | none |
 | `--device_type` | Device type | cuda |
 | `--start_method` | Multiprocessing start method (`spawn`, `fork`, `forkserver`) | spawn |
 | `--backend` | Distributed training backend | nccl |
@@ -164,6 +164,7 @@ nohup python scripts/tools/train.py \
 | `--device` | str | `cuda` | Device to load model on |
 | `--dtype` | str | `bfloat16` | Model weights dtype (`bfloat16`, `float16`, `float32`) |
 | `--max_batch_size` | int | `16` | Maximum batch size for continuous batching |
+| `--max_seq_len` | int | model config `max_position_embeddings` | Maximum sequence length (KV cache size + prompt truncation) |
 | `--reload` | flag | `False` | Enable auto-reload for development |
 
 Usage:
@@ -186,7 +187,11 @@ See [Inference Guide](inference.md) for HTTP API documentation.
 | `--top_k` | int | `30` | Top-k filtering |
 | `--top_p` | float | `0.95` | Nucleus sampling threshold |
 | `--batch_size` | int | `1` | Batch size for generation |
+| `--num_samples` | int | `1` | Responses per prompt |
 | `--max_tokens` | int | model config `max_position_embeddings` | Maximum tokens to generate |
+| `--cache_len` | int | `2048` | KV cache length |
+| `--frequency_penalty` | float | `0.0` | Frequency penalty |
+| `--rep_window` | int | `64` | Window size for frequency penalty |
 
 Usage:
 ```bash

assets/docs/preprocessing.md

@@ -256,6 +256,7 @@ When `sources` is set, `sections` is ignored.
 | `min_chars` | int | `50` | Skip text-mode items shorter than this |
 | `max_chars` | int | `2000000` | Skip text-mode items longer than this |
 | `max_items` | int or null | `null` | Stop after N documents |
+| `batch_size` | int | `256` | Records per tokenization batch |
 | `packing_strategy` | str | `"simple"` | Packing strategy: `"simple"`, `"bfd"`, `"bfd_split"` |
 | `max_packed_len` | int | `8192` | Maximum length of a packed bin |
 | `truncation_mode` | str | `"keep_start"` | How to truncate sequences: `"keep_start"` or `"keep_end"` |