From 369e499a666a679d7234993d8e4868de691b8119 Mon Sep 17 00:00:00 2001 From: yuanwu2017 Date: Thu, 15 Aug 2024 18:04:14 +0800 Subject: [PATCH] Simplify the warmup process (#173) Signed-off-by: yuanwu --- server/text_generation_server/models/causal_lm.py | 8 -------- 1 file changed, 8 deletions(-) diff --git a/server/text_generation_server/models/causal_lm.py b/server/text_generation_server/models/causal_lm.py index 6437acf9..be3b8f4d 100644 --- a/server/text_generation_server/models/causal_lm.py +++ b/server/text_generation_server/models/causal_lm.py @@ -1192,14 +1192,6 @@ class CausalLM(Model): if len(request_ids) < len(decode_batch.requests): decode_batch = decode_batch.filter(request_ids) - while decode_batch is not None: - # filter finished requests - request_ids = get_unfinished_requests(decode_batch.requests) - if len(request_ids) < len(decode_batch.requests): - decode_batch = decode_batch.filter(request_ids) - # decode - _, decode_batch, _ = self.generate_token([decode_batch]) - def shifting_warmup(self, batch: CausalLMBatch) -> None: chunk_sizes = CHUNK_SIZES.copy() chunk_sizes.extend([-chunk for chunk in chunk_sizes])