set z3 leaf for deepseek v2 (#1809) [skip ci]
* set z3 leaf for deepseek v2 * add deepseek v2 chat template
This commit is contained in:
@@ -348,7 +348,11 @@ def load_model(
|
||||
and cfg.flash_attention
|
||||
and cfg.sample_packing
|
||||
):
|
||||
patch_for_multipack(cfg.model_config_type, model_name=cfg.base_model)
|
||||
patch_for_multipack(
|
||||
cfg.model_config_type,
|
||||
model_name=cfg.base_model,
|
||||
is_remote_code=cfg.trust_remote_code,
|
||||
)
|
||||
|
||||
if cfg.is_llama_derived_model:
|
||||
from axolotl.monkeypatch.llama_attn_hijack_flash import (
|
||||
|
||||
Reference in New Issue
Block a user