[GH-ISSUE #1151] [Bug] 2张2080TI 用KLinearTorch 的形式运行，聊天报错 #799

New issue

Open

opened 2026-10-29 14:44:38 +00:00 by bsakaguchi · 2 comments

bsakaguchi commented

2026-10-29 14:44:38 +00:00

Owner

Originally created by @GEK1 on GitHub (Apr 16, 2025).
Original GitHub issue: https://github.com/kvcache-ai/ktransformers/issues/1151

Originally assigned to: @Azure-Tang on GitHub.

检查清单

1. 我已经搜索过相关问题，但未能获得预期的帮助
2. 该问题在最新版本中尚未修复
3. 请注意，如果您提交的BUG相关 issue 缺少对应环境信息和最小可复现示例，我们将难以复现和定位问题，降低获得反馈的可能性
4. 如果您提出的不是bug而是问题，请在讨论区发起讨论 https://github.com/kvcache-ai/ktransformers/discussions。否则该 issue 将被关闭
5. 为方便社区交流，我将使用中文/英文或附上中文/英文翻译（如使用其他语言）。未附带翻译的非中文/英语内容可能会被关闭

问题描述

2张2080TI 用KLinearTorch 的形式运行，能跑起来，但是一聊天就报错

复现步骤

模型：unsloth/DeepSeek-R1-UD-Q2_K_XL
运行命令：
ktransformers --gguf_path /opt/ktransformers/data/DeepSeek-R1-UD-Q2_K_XL/ --model_path deepseek-ai/DeepSeek-R1 --model_name unsloth/DeepSeek-R1-UD-Q2_K_XL --optimize_config_path /opt/ktransformers/ktransformers/ktransformers/optimize/optimize_rules/DeepSeek-V3-Chat-multi-gpu_GEK.yaml --max_new_tokens 8192 --cpu_infer 64 --port 10012 --batch_size 5 --web True

环境信息

2080ti 22G *2
ubuntu 2204
Intel(R) Xeon(R) CPU E5-2686 v4 @ 2.30GHz *2

Originally created by @GEK1 on GitHub (Apr 16, 2025). Original GitHub issue: https://github.com/kvcache-ai/ktransformers/issues/1151 Originally assigned to: @Azure-Tang on GitHub. ### 检查清单 - [x] 1. 我已经搜索过相关问题，但未能获得预期的帮助 - [ ] 2. 该问题在最新版本中尚未修复 - [ ] 3. 请注意，如果您提交的BUG相关 issue 缺少对应环境信息和最小可复现示例，我们将难以复现和定位问题，降低获得反馈的可能性 - [ ] 4. 如果您提出的不是bug而是问题，请在讨论区发起讨论 https://github.com/kvcache-ai/ktransformers/discussions。否则该 issue 将被关闭 - [ ] 5. 为方便社区交流，我将使用中文/英文或附上中文/英文翻译（如使用其他语言）。未附带翻译的非中文/英语内容可能会被关闭 ### 问题描述 2张2080TI 用KLinearTorch 的形式运行，能跑起来，但是一聊天就报错 ### 复现步骤模型：unsloth/DeepSeek-R1-UD-Q2_K_XL 运行命令： ktransformers --gguf_path /opt/ktransformers/data/DeepSeek-R1-UD-Q2_K_XL/ --model_path deepseek-ai/DeepSeek-R1 --model_name unsloth/DeepSeek-R1-UD-Q2_K_XL --optimize_config_path /opt/ktransformers/ktransformers/ktransformers/optimize/optimize_rules/DeepSeek-V3-Chat-multi-gpu_GEK.yaml --max_new_tokens 8192 --cpu_infer 64 --port 10012 --batch_size 5 --web True ![Image](https://github.com/user-attachments/assets/f0c4116c-c46b-4e91-954f-57970afe79a8) ### 环境信息 2080ti 22G *2 ubuntu 2204 Intel(R) Xeon(R) CPU E5-2686 v4 @ 2.30GHz *2

bsakaguchi commented

2026-10-29 14:44:39 +00:00

Author

Owner

@zx1896 commented on GitHub (Apr 16, 2025):

我刚刚也报了这个错误：TypeError: KLinearTorch.forward() takes 2 positional arguments but 3 were given
然后将y = self.generate_linear.forward(x, bsz_tensor)改为y = self.generate_linear.forward(x)就好使了

@zx1896 commented on GitHub (Apr 16, 2025): 我刚刚也报了这个错误：TypeError: KLinearTorch.forward() takes 2 positional arguments but 3 were given 然后将y = self.generate_linear.forward(x, bsz_tensor)改为y = self.generate_linear.forward(x)就好使了

bsakaguchi commented

2026-10-29 14:44:39 +00:00

Author

Owner

@GEK1 commented on GitHub (Apr 16, 2025):

我刚刚也报了这个错误：TypeError: KLinearTorch.forward() takes 2 positional arguments but 3 were given 然后将y = self.generate_linear.forward(x, bsz_tensor)改为y = self.generate_linear.forward(x)就好使了

我也是这样子改了，改完又报新的错了

@GEK1 commented on GitHub (Apr 16, 2025): > 我刚刚也报了这个错误：TypeError: KLinearTorch.forward() takes 2 positional arguments but 3 were given 然后将y = self.generate_linear.forward(x, bsz_tensor)改为y = self.generate_linear.forward(x)就好使了我也是这样子改了，改完又报新的错了 ![Image](https://github.com/user-attachments/assets/162c4b7e-4ac8-45b4-b325-9896914c7faf)

bsakaguchi referenced this issue

2026-10-29 14:52:50 +00:00

[PR #799] [MERGED] Restore CPU offloading capability #1484