vrr/kvcache-ai-ktransformers
2
0
Fork 0
mirror of https://github.com/kvcache-ai/ktransformers.git synced 2025-09-10 06:14:58 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

[Patch] lload DeepSeek-R1-0528

Browse source
This commit is contained in:
qiyuxinlin 2025-05-31 14:19:20 +00:00
parent ac48a58cca
commit a6b3243a56
2 changed files with 18 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

6
ktransformers/util/custom_loader.py
View file

@ -446,7 +446,11 @@ class GGUFLoader(ModelLoader):
blocks_begin = i * blocks_per_iter
blocks_end = min(blocks_begin + blocks_per_iter, num_blocks)
if "cuda" in device.lower():
cur_values = GGML_DEQUANTIZE_GPU[ggml_name](data[blocks_begin*block_size : blocks_end*block_size], device, target_dtype)
try:
cur_values = GGML_DEQUANTIZE_GPU[ggml_name](data[blocks_begin*block_size : blocks_end*block_size], device, target_dtype)
except:
cur_values = GGML_DEQUANTIZE[ggml_name](data[blocks_begin*block_size : blocks_end*block_size])
cur_values = torch.from_numpy(cur_values.copy()).to(device)
else:
cur_values = GGML_DEQUANTIZE[ggml_name](data[blocks_begin*block_size : blocks_end*block_size])
cur_values = torch.from_numpy(cur_values.copy())