|
|
@@ -325,10 +325,10 @@ def decode_n_tokens(
|
|
|
input_pos += 1
|
|
|
cur_token = next_token.view(1, model.config.num_codebooks + 1, -1)
|
|
|
# Roll RAS window left and insert new token at end
|
|
|
- previous_tokens = previous_tokens.roll(-1, dims=1)
|
|
|
- previous_tokens[:, -1] = next_token.view(model.config.num_codebooks + 1, -1)[
|
|
|
- :, 0
|
|
|
- ]
|
|
|
+ # previous_tokens = previous_tokens.roll(-1, dims=1)
|
|
|
+ # previous_tokens[:, -1] = next_token.view(model.config.num_codebooks + 1, -1)[
|
|
|
+ # :, 0
|
|
|
+ # ]
|
|
|
new_tokens.append(next_token)
|
|
|
f_end = time.perf_counter()
|
|
|
# logger.info(f"num_new_tokens for elapse: {f_end - f_start}")
|