fixed LTX Video support for RTX10xx , 20xx

This commit is contained in:
DeepBeepMeep 2025-05-17 23:28:48 +02:00
parent 5ba133cae9
commit 8b9929f744
3 changed files with 10 additions and 3 deletions

View File

@ -189,7 +189,7 @@ class TextEncoder(nn.Module):
if "llm" in text_encoder_type:
from mmgp import offload
forcedConfigPath= None if "i2v" in text_encoder_type else "ckpts/llava-llama-3-8b/config.json"
self.model= offload.fast_load_transformers_model(self.model_path, forcedConfigPath=forcedConfigPath)
self.model= offload.fast_load_transformers_model(self.model_path, modelPrefix="language_model" if forcedConfigPath != None else None, forcedConfigPath=forcedConfigPath)
if forcedConfigPath != None:
self.model.final_layer_norm = self.model.model.norm

View File

@ -134,6 +134,11 @@ def get_supported_attention_modes():
if not sage2_supported:
if "sage2" in ret:
ret.remove("sage2")
major, minor = torch.cuda.get_device_capability()
if major < 7:
if "sage" in ret:
ret.remove("sage")
return ret
__all__ = [
@ -173,6 +178,8 @@ def pay_attention(
# assume if k_lens is non null, each k is padded up to lk (one k out of two will need to be discarded or ignored)
if attention_mask != None:
force_attention = "sdpa"
if attention_mask.dtype == torch.bfloat16 and not bfloat16_supported:
attention_mask = attention_mask.to(torch.float16)
attn = offload.shared_state["_attention"] if force_attention== None else force_attention
q,k,v = qkv_list

4
wgp.py
View File

@ -1869,9 +1869,9 @@ def get_ltxv_text_encoder_filename(text_encoder_quantization):
def get_hunyuan_text_encoder_filename(text_encoder_quantization):
if text_encoder_quantization =="int8":
text_encoder_filename = "ckpts/llava-llama-3-8b/llava-llama-3-8b-v1_1_vlm_fp16.safetensors"
else:
text_encoder_filename = "ckpts/llava-llama-3-8b/llava-llama-3-8b-v1_1_vlm_quanto_int8.safetensors"
else:
text_encoder_filename = "ckpts/llava-llama-3-8b/llava-llama-3-8b-v1_1_vlm_fp16.safetensors"
return text_encoder_filename