xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-04-09 21:57:12 +08:00
Code Issues Packages Projects Releases Wiki Activity

Support LoRA for PLaMo 2/3

Browse Source 
Signed-off-by: Shinichi Hemmi <50256998+Alnusjaponica@users.noreply.github.com>
This commit is contained in:
Shinichi Hemmi 2025-12-25 05:54:31 +09:00
parent 09dc7c690c
commit f475ea1ee0
3 changed files with 24 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
docs/models/supported_models.md
View File

@ -438,8 +438,8 @@ th {
| `Phi3ForCausalLM` | Phi-4, Phi-3 | `microsoft/Phi-4-mini-instruct`, `microsoft/Phi-4`, `microsoft/Phi-3-mini-4k-instruct`, `microsoft/Phi-3-mini-128k-instruct`, `microsoft/Phi-3-medium-128k-instruct`, etc. | ✅︎ | ✅︎ |
| `PhiMoEForCausalLM` | Phi-3.5-MoE | `microsoft/Phi-3.5-MoE-instruct`, etc. | ✅︎ | ✅︎ |
| `PersimmonForCausalLM` | Persimmon | `adept/persimmon-8b-base`, `adept/persimmon-8b-chat`, etc. | | ✅︎ |
| `Plamo2ForCausalLM` | PLaMo2 | `pfnet/plamo-2-1b`, `pfnet/plamo-2-8b`, etc. | | ✅︎ |
| `Plamo3ForCausalLM` | PLaMo3 | `pfnet/plamo-3-nict-2b-base`, `pfnet/plamo-3-nict-8b-base`, etc. | | ✅︎ |
| `Plamo2ForCausalLM` | PLaMo2 | `pfnet/plamo-2-1b`, `pfnet/plamo-2-8b`, etc. | ✅ | ✅︎ |
| `Plamo3ForCausalLM` | PLaMo3 | `pfnet/plamo-3-nict-2b-base`, `pfnet/plamo-3-nict-8b-base`, etc. | ✅ | ✅︎ |
| `QWenLMHeadModel` | Qwen | `Qwen/Qwen-7B`, `Qwen/Qwen-7B-Chat`, etc. | ✅︎ | ✅︎ |
| `Qwen2ForCausalLM` | QwQ, Qwen2 | `Qwen/QwQ-32B-Preview`, `Qwen/Qwen2-7B-Instruct`, `Qwen/Qwen2-7B`, etc. | ✅︎ | ✅︎ |
| `Qwen2MoeForCausalLM` | Qwen2MoE | `Qwen/Qwen1.5-MoE-A2.7B`, `Qwen/Qwen1.5-MoE-A2.7B-Chat`, etc. | ✅︎ | ✅︎ |

26
vllm/model_executor/models/plamo2.py
View File

@ -50,7 +50,12 @@ from vllm.model_executor.model_loader.weight_utils import (
default_weight_loader,
sharded_weight_loader,
)
from vllm.model_executor.models.interfaces import HasInnerState, IsHybrid, SupportsPP
from vllm.model_executor.models.interfaces import (
HasInnerState,
IsHybrid,
SupportsLoRA,
SupportsPP,
)
from vllm.model_executor.models.utils import (
is_pp_missing_parameter,
make_empty_intermediate_tensors_factory,
@ -105,6 +110,7 @@ class Plamo2MambaMixer(MambaBase, CustomOp):
self.cache_config = vllm_config.cache_config
self.model_config = vllm_config.model_config
self.quant_config = vllm_config.quant_config
self.is_lora_enabled = bool(vllm_config.lora_config)
self.hidden_size = self.config.hidden_size
self.ssm_state_size = self.config.mamba_d_state
self.conv_kernel_size = self.config.mamba_d_conv
@ -202,7 +208,11 @@ class Plamo2MambaMixer(MambaBase, CustomOp):
self.prefix = prefix
def _project_ssm_parameters(self, hidden_states):
ssm_parameters = self.bcdt_proj(hidden_states)
if self.is_lora_enabled:
# Lora kernel requires contiguous tensor.
ssm_parameters = self.bcdt_proj(hidden_states.contiguous())
else:
ssm_parameters = self.bcdt_proj(hidden_states)
B, C, time_step = torch.split(
ssm_parameters,
[self.ssm_state_size, self.ssm_state_size, self.time_step_rank],
@ -780,13 +790,13 @@ class Plamo2Model(torch.nn.Module):
return hidden_states
class Plamo2ForCausalLM(torch.nn.Module, HasInnerState, SupportsPP, IsHybrid):
class Plamo2ForCausalLM(
torch.nn.Module, HasInnerState, SupportsLoRA, SupportsPP, IsHybrid
):
packed_modules_mapping = {
"qkv_proj": [
"q_proj",
"k_proj",
"v_proj",
],
"qkv_proj": ["qkv_proj"],
"gate_up_proj": ["gate_up_proj"],
"in_proj": ["in_proj"],
}
def __init__(self, *, vllm_config: VllmConfig, prefix: str = "") -> None:

11
vllm/model_executor/models/plamo3.py
View File

@ -35,7 +35,7 @@ from vllm.model_executor.model_loader.weight_utils import (
composed_weight_loader,
default_weight_loader,
)
from vllm.model_executor.models.interfaces import SupportsPP
from vllm.model_executor.models.interfaces import SupportsLoRA, SupportsPP
from vllm.model_executor.models.utils import (
AutoWeightsLoader,
extract_layer_index,
@ -369,13 +369,10 @@ class Plamo3Model(nn.Module):
return hidden_states
class Plamo3ForCausalLM(nn.Module, SupportsPP):
class Plamo3ForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
packed_modules_mapping = {
"qkv_proj": [
"q_proj",
"k_proj",
"v_proj",
],
"qkv_proj": ["qkv_proj"],
"gate_up_proj": ["gate_up_proj"],
}
def __init__(self, *, vllm_config: VllmConfig, prefix: str = "") -> None: