Upload README.md with huggingface_hub

README.md

@@ -91,18 +91,17 @@ N_FRAMES = 50                                      # ~4 s at SNAC-24kHz's coarse
 N_TOKENS = N_FRAMES * 7                            # 7 tokens / frame (C,M,F,F,M,F,F)
 
 # --- 2. Generate inside an [SNAC] ... span ------------------------------
-# IMPORTANT: use_cache=False is required. The slot router (in
-# modeling_nemotron_h_augmented.py) re-derives state from the full
-# input_ids on every forward; with the default KV-cached generate
-# (which only re-feeds the new token), routing state would reset to
-# 0 each step and the C/M/F pattern would not be enforced.
+# The slot router (modeling_nemotron_h_augmented.py) carries its
+# (in_slot_mode, slot_counter) state across forward calls via
+# self._slot_router_state, so KV caching just works: prefill computes
+# routing from initial state, subsequent forwards advance from the
+# cached final state. No special flags needed.
 prompt = torch.tensor([[tok.bos_token_id, SNAC_OPEN]], device="cuda")
 with torch.no_grad():
     out = model.generate(
         prompt,
         max_new_tokens=N_TOKENS,
         do_sample=True, temperature=0.8, top_p=0.95,
-        use_cache=False,                            # see note above
     )
 
 # --- 3. Parse the C/M/F/F/M/F/F frames back into codebook indices --------
@@ -140,7 +139,7 @@ print(f"saved {audio.shape[-1] / 24000:.2f} s of audio  "
       f"({len(c_codes)} frames, {len(c_codes) + len(m_codes) + len(f_codes)} codes)")
 ```
 
-**Dependencies**: `pip install snac torchaudio` in addition to `transformers torch`. Total wall-clock for 50 frames (~4 s of audio): roughly 30 – 60 s on one GB10 with `use_cache=False`.
+**Dependencies**: `pip install snac torchaudio` in addition to `transformers torch`. Wall-clock for 50 frames (~4 s of audio): a few seconds on a GB10 with KV caching on (the default).
 
 **Token-budget rule of thumb**: SNAC-24kHz's coarse rate is ~12 Hz, so one frame ≈ 83 ms of audio. To pre-allocate `max_new_tokens` for a given duration: