huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-09-09 03:14:53 +00:00
Code Issues Packages Projects Releases Wiki Activity

Make style

Browse Source
This commit is contained in:
regisss 2025-05-10 13:43:38 +00:00
parent 2b2b4a814d
commit afbebe6990
2 changed files with 10 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
backends/gaudi/server/text_generation_server/models/causal_lm.py
View File

@ -1380,10 +1380,7 @@ class CausalLM(Model):
prefill_seqlen_list.sort()
prefill_batch_size_list.sort()
mem_stats = get_hpu_memory_stats(self.device)
logger.info(
f"Prefill warmup successful.\n"
f"Memory stats: {mem_stats} "
)
logger.info(f"Prefill warmup successful.\n" f"Memory stats: {mem_stats} ")
max_decode_batch_size = math.floor(MAX_BATCH_TOTAL_TOKENS / MAX_TOTAL_TOKENS)
max_exp = math.ceil(math.log(max_decode_batch_size, BATCH_SIZE_EXPONENT_BASE))
@ -1395,7 +1392,9 @@ class CausalLM(Model):
try:
for batch_size in decode_batch_size_list:
logger.info(f"Decode warmup for `batch_size={batch_size}`, this may take a while...")
logger.info(
f"Decode warmup for `batch_size={batch_size}`, this may take a while..."
)
batches = []
iters = math.floor(batch_size / max_prefill_batch_size)
for i in range(iters):
@ -1428,10 +1427,7 @@ class CausalLM(Model):
decode_batch_size_list.sort()
max_supported_total_tokens = MAX_TOTAL_TOKENS * decode_batch_size_list[-1]
mem_stats = get_hpu_memory_stats(self.device)
logger.info(
f"Decode warmup successful.\n"
f"Memory stats: {mem_stats} "
)
logger.info(f"Decode warmup successful.\n" f"Memory stats: {mem_stats} ")
max_input_tokens = max_input_tokens
max_total_tokens = MAX_TOTAL_TOKENS

14
backends/gaudi/server/text_generation_server/models/vlm_causal_lm.py
View File

@ -1539,10 +1539,7 @@ class VlmCausalLM(Model):
)
mem_stats = get_hpu_memory_stats(self.device)
logger.info(
f"Prefill warmup successful.\n"
f"Memory stats: {mem_stats} "
)
logger.info(f"Prefill warmup successful.\n" f"Memory stats: {mem_stats} ")
max_decode_batch_size = MAX_BATCH_SIZE
batch_size = max_prefill_batch_size * 2
@ -1556,7 +1553,9 @@ class VlmCausalLM(Model):
batches = []
while batch_size <= max_decode_batch_size:
for i in range(int(batch_size / max_prefill_batch_size)):
logger.info(f"Decode warmup for `batch_size={batch_size}`, this may take a while...")
logger.info(
f"Decode warmup for `batch_size={batch_size}`, this may take a while..."
)
batch = self.generate_warmup_batch(
request,
PREFILL_WARMUP_SEQLEN_LIST[0] - 1,
@ -1599,10 +1598,7 @@ class VlmCausalLM(Model):
)
mem_stats = get_hpu_memory_stats(self.device)
logger.info(
f"Decode warmup successful.\n"
f"Memory stats: {mem_stats}"
)
logger.info(f"Decode warmup successful.\n" f"Memory stats: {mem_stats}")
max_supported_total_tokens = MAX_BATCH_SIZE * MAX_TOTAL_TOKENS
max_input_tokens = max_input_tokens