biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Core][Model] PrithviMAE Enablement on vLLM v1 engine (#20577)

Browse Source 
Signed-off-by: Christian Pinto <christian.pinto@ibm.com>
This commit is contained in:
Christian Pinto
2025-07-23 19:00:23 +01:00
committed by GitHub
parent 316b1bf706
commit 8560a5b258
15 changed files with 704 additions and 238 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
vllm/v1/engine/processor.py
View File

@@ -380,7 +380,6 @@ class Processor:
prompt_type: Literal["encoder", "decoder"],
):
model_config = self.model_config
tokenizer = self.tokenizer.get_lora_tokenizer(lora_request)
prompt_ids = prompt_inputs["prompt_token_ids"]
if not prompt_ids:
@@ -389,9 +388,14 @@ class Processor:
else:
raise ValueError(f"The {prompt_type} prompt cannot be empty")
max_input_id = max(prompt_ids, default=0)
if max_input_id > tokenizer.max_token_id:
raise ValueError(f"Token id {max_input_id} is out of vocabulary")
if self.model_config.skip_tokenizer_init:
tokenizer = None
else:
tokenizer = self.tokenizer.get_lora_tokenizer(lora_request)
max_input_id = max(prompt_ids, default=0)
if max_input_id > tokenizer.max_token_id:
raise ValueError(
f"Token id {max_input_id} is out of vocabulary")
max_prompt_len = self.model_config.max_model_len
if len(prompt_ids) > max_prompt_len: