Feat: Add support for gemma3_text and add e2e for gemma2 (#2406)

This commit is contained in:
NanoCode012
2025-03-23 07:33:21 +07:00
committed by GitHub
parent 86bac48d14
commit 9f00465a5c
12 changed files with 348 additions and 6 deletions

View File

@@ -114,3 +114,5 @@ class LigerPlugin(BasePlugin):
modeling_mod.CrossEntropyLoss = LigerCrossEntropyLoss
if cfg.liger_fused_linear_cross_entropy:
modeling_mod.DeepseekV2ForCausalLM.forward = deepseekv2_lce_forward
elif cfg.model_config_type in ["gemma3_text", "deepseek_v3"]:
raise ValueError(f"Unsupported model config type: {cfg.model_config_type}")

View File

@@ -22,6 +22,7 @@ SUPPORTED_MULTIPACK_MODEL_TYPES = [
"phi3",
"gemma",
"gemma2",
"gemma3_text",
"gemmoe",
"starcoder2",
"deepseek_v2",

File diff suppressed because one or more lines are too long

View File

@@ -23,6 +23,7 @@ class ChatTemplate(str, Enum):
mistral_v2v3 = "mistral_v2v3" # pylint: disable=invalid-name
mistral_v3_tekken = "mistral_v3_tekken" # pylint: disable=invalid-name
gemma = "gemma" # pylint: disable=invalid-name
gemma3_text = "gemma3_text" # pylint: disable=invalid-name
cohere = "cohere" # pylint: disable=invalid-name
llama3 = "llama3" # pylint: disable=invalid-name
llama3_2_vision = "llama3_2_vision" # pylint: disable=invalid-name