VictorMylle/Kohya-ss-sd-scripts
1
0
Fork 1
mirror of https://github.com/kohya-ss/sd-scripts.git synced 2026-04-18 01:30:02 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix: --fp8_vl to work

Browse Source
This commit is contained in:
Kohya S
2025-09-18 23:46:18 +09:00
parent f6b4bdc83f
commit f834b2e0d4
2 changed files with 3 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
library/hunyuan_image_text_encoder.py
View File

@@ -15,7 +15,7 @@ from transformers.models.t5.modeling_t5 import T5Stack
from accelerate import init_empty_weights
from library.safetensors_utils import load_safetensors
from library.utils import setup_logging
from library.utils import setup_logging
setup_logging()
import logging
@@ -542,7 +542,6 @@ def get_qwen_prompt_embeds_from_tokens(
attention_mask = attention_mask.to(device=device)
if dtype.itemsize == 1: # fp8
# TODO dtype should be vlm.dtype?
with torch.no_grad(), torch.autocast(device_type=device.type, dtype=torch.bfloat16, enabled=True):
encoder_hidden_states = vlm(input_ids=input_ids, attention_mask=attention_mask, output_hidden_states=True)
else:
@@ -564,7 +563,7 @@ def get_qwen_prompt_embeds_from_tokens(
prompt_embeds = hidden_states[:, drop_idx:, :]
encoder_attention_mask = attention_mask[:, drop_idx:]
prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
prompt_embeds = prompt_embeds.to(device=device)
return prompt_embeds, encoder_attention_mask