Mirrors
/
hf_text-generation-inference
mirror of https://github.com/huggingface/text-generation-inference.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

fix(server): fix deepseekv2 loading (#2266)

This commit is contained in:
OlivierDehaene 2024-07-21 16:48:04 +00:00 committed by GitHub
parent 53ec0b790b
commit f3435bab8c
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 0 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
server/text_generation_server/models/custom_modeling/flash_deepseek_v2_modeling.py
View File

@ -34,7 +34,6 @@ from text_generation_server.layers.attention.common import Seqlen
from text_generation_server.layers.layernorm import FastRMSNorm
from text_generation_server.layers.rotary import PositionRotaryEmbedding, get_mscale
from text_generation_server.utils.import_utils import SYSTEM
from text_generation_server.utils.log import log_once
from text_generation_server.utils.weights import Weights
from torch import nn
from transformers.activations import ACT2FN
@ -240,7 +239,6 @@ class DeepseekV2Attention(torch.nn.Module):
if config.attention_bias
else None
),
quantize=config.quantize,
)
self.q_a_layernorm = FastRMSNorm.load(
prefix=f"{prefix}.q_a_layernorm",
@ -261,7 +259,6 @@ class DeepseekV2Attention(torch.nn.Module):
if config.attention_bias
else None
),
quantize=config.quantize,
)
self.kv_a_layernorm = FastRMSNorm.load(