A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 16, 08:15 UTC
[Misc]Fix Idefics3Model argument (#10255)
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
Jee Jee Li committed
Nov 12, 2024 at 21:07 UTC
a838ba7254c98a7adc60a0976bdf277fb20b4221
1 file changed
+4
-4
vllm/model_executor/models/idefics3.py
+4
-4
@@ -420,18 +420,18 @@ class Idefics3Model(nn.Module):
420
super().__init__()
421
422
config = vllm_config.model_config.hf_config
423
- cache_config = vllm_config.cache_config
423
quant_config = vllm_config.quant_config
424
425
self.config = config
426
self.padding_idx = self.config.text_config.pad_token_id
427
self.vocab_size = self.config.text_config.vocab_size
429
-
428
self.vision_model = Idefics3VisionTransformer(config.vision_config,
429
quant_config)
430
self.connector = Idefics3Connector(config)
433
- self.text_model = LlamaModel(config.text_config, cache_config,
434
- quant_config)
431
+ self.text_model = LlamaModel(
432
+ vllm_config=vllm_config.with_hf_config(config.text_config),
433
+ prefix=maybe_prefix(prefix, "text_model"),
434
+ )
435
436
self.image_seq_len = int(
437
((config.vision_config.image_size //