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 14, 23:45 UTC
[`Ernie 4.5`] Name Change for Base 0.3B Model (#21735)
Signed-off-by: vasqu <antonprogamer@gmail.com> Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk> Co-authored-by: DarkLight1337 <tlleungac@connect.ust.hk>
Anton Vlasjuk committed
Jul 28, 2025 at 14:22 UTC
656c24f1b5d8a662e9ec391503d146341def5f18
4 files changed
+8
-8
docs/models/supported_models.md
+5
-5
@@ -336,7 +336,7 @@ th {
336
| `DeepseekV2ForCausalLM` | DeepSeek-V2 | `deepseek-ai/DeepSeek-V2`, `deepseek-ai/DeepSeek-V2-Chat`, etc. | | ✅︎ | ✅︎ |
337
| `DeepseekV3ForCausalLM` | DeepSeek-V3 | `deepseek-ai/DeepSeek-V3-Base`, `deepseek-ai/DeepSeek-V3`, etc. | | ✅︎ | ✅︎ |
338
| `Dots1ForCausalLM` | dots.llm1 | `rednote-hilab/dots.llm1.base`, `rednote-hilab/dots.llm1.inst`, etc. | | ✅︎ | ✅︎ |
339
-| `Ernie4_5_ForCausalLM` | Ernie4.5 | `baidu/ERNIE-4.5-0.3B-PT`, etc. | ✅︎ | ✅︎ | ✅︎ |
339
+| `Ernie4_5ForCausalLM` | Ernie4.5 | `baidu/ERNIE-4.5-0.3B-PT`, etc. | ✅︎ | ✅︎ | ✅︎ |
340
| `Ernie4_5_MoeForCausalLM` | Ernie4.5MoE | `baidu/ERNIE-4.5-21B-A3B-PT`, `baidu/ERNIE-4.5-300B-A47B-PT`, etc. |✅︎| ✅︎ | ✅︎ |
341
| `ExaoneForCausalLM` | EXAONE-3 | `LGAI-EXAONE/EXAONE-3.0-7.8B-Instruct`, etc. | ✅︎ | ✅︎ | ✅︎ |
342
| `Exaone4ForCausalLM` | EXAONE-4 | `LGAI-EXAONE/EXAONE-4.0-32B`, etc. | ✅︎ | ✅︎ | ✅︎ |
@@ -634,10 +634,10 @@ Some models are supported only via the [Transformers backend](#transformers). Th
634
|--------------|--------|--------|-------------------|-----------------------------|-----------------------------------------|---------------------|
635
| `Emu3ForConditionalGeneration` | Emu3 | T + I | `BAAI/Emu3-Chat-hf` | ✅︎ | ✅︎ | ✅︎ |
636
637
-<sup>^</sup> You need to set the architecture name via `--hf-overrides` to match the one in vLLM.
638
- • For example, to use DeepSeek-VL2 series models:
639
- `--hf-overrides '{"architectures": ["DeepseekVLV2ForCausalLM"]}'`
640
-<sup>E</sup> Pre-computed embeddings can be inputted for this modality.
637
+<sup>^</sup> You need to set the architecture name via `--hf-overrides` to match the one in vLLM.
638
+ • For example, to use DeepSeek-VL2 series models:
639
+ `--hf-overrides '{"architectures": ["DeepseekVLV2ForCausalLM"]}'`
640
+<sup>E</sup> Pre-computed embeddings can be inputted for this modality.
641
<sup>+</sup> Multiple items can be inputted per text prompt for this modality.
642
643
!!! warning
tests/models/registry.py
+1
-1
@@ -166,7 +166,7 @@ _TEXT_GENERATION_EXAMPLE_MODELS = {
166
trust_remote_code=True),
167
"DeepseekV3ForCausalLM": _HfExamplesInfo("deepseek-ai/DeepSeek-V3", # noqa: E501
168
trust_remote_code=True),
169
- "Ernie4_5_ForCausalLM": _HfExamplesInfo("baidu/ERNIE-4.5-0.3B-PT",
169
+ "Ernie4_5ForCausalLM": _HfExamplesInfo("baidu/ERNIE-4.5-0.3B-PT",
170
min_transformers_version="4.54"),
171
"Ernie4_5_MoeForCausalLM": _HfExamplesInfo("baidu/ERNIE-4.5-21B-A3B-PT",
172
min_transformers_version="4.54"),
vllm/model_executor/models/ernie45.py
+1
-1
@@ -28,7 +28,7 @@ from vllm.model_executor.models.llama import LlamaForCausalLM
28
from .utils import PPMissingLayer
29
30
31
-class Ernie4_5_ForCausalLM(LlamaForCausalLM):
31
+class Ernie4_5ForCausalLM(LlamaForCausalLM):
32
33
def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
34
super().__init__(vllm_config=vllm_config, prefix=prefix)
vllm/model_executor/models/registry.py
+1
-1
@@ -60,7 +60,7 @@ _TEXT_GENERATION_MODELS = {
60
"DeepseekV2ForCausalLM": ("deepseek_v2", "DeepseekV2ForCausalLM"),
61
"DeepseekV3ForCausalLM": ("deepseek_v2", "DeepseekV3ForCausalLM"),
62
"Dots1ForCausalLM": ("dots1", "Dots1ForCausalLM"),
63
- "Ernie4_5_ForCausalLM": ("ernie45", "Ernie4_5_ForCausalLM"),
63
+ "Ernie4_5ForCausalLM": ("ernie45", "Ernie4_5ForCausalLM"),
64
"Ernie4_5_MoeForCausalLM": ("ernie45_moe", "Ernie4_5_MoeForCausalLM"),
65
"ExaoneForCausalLM": ("exaone", "ExaoneForCausalLM"),
66
"Exaone4ForCausalLM": ("exaone4", "Exaone4ForCausalLM"),