Ecosyste.ms: Timeline
Browse the timeline of events for every public repo on GitHub. Data updated hourly from GH Archive.
sixsixcoder created a repository: sixsixcoder/GithubToLark
https://github.com/sixsixcoder/GithubToLark - Receive information from GitHub and push it to Lark
sixsixcoder pushed 1 commit to glm-4 sixsixcoder/vllm
- [Bugfix] Ignore ray reinit error when current platform is ROCm or XPU (#10375) Signed-off-by: Hollow Man <hollowman@... 47826ca
sixsixcoder created a branch on sixsixcoder/vllm
glm-4 - A high-throughput and memory-efficient inference and serving engine for LLMs
sixsixcoder pushed 495 commits to main sixsixcoder/vllm
- [Misc] Collect model support info in a single process per model (#9233) e808156
- [Misc][LoRA] Support loading LoRA weights for target_modules in reg format (#9275) 36ea790
- [Bugfix] Fix priority in multiprocessing engine (#9277) df3dcdf
- [Model] Support Mamba (#6484) 7342a7d
- [Kernel] adding fused moe kernel config for L40S TP4 (#9245) f710090
- [Model] Add GLM-4v support and meet vllm==0.6.2 (#9242) 6cf1167
- [Doc] Remove outdated comment to avoid misunderstanding (#9287) 1a18238
- [Doc] Compatibility matrix for mutual exclusive features (#8512) Signed-off-by: Wallas Santos <[email protected]> 8baf85e
- [Bugfix][CI/Build] Fix docker build where CUDA archs < 7.0 are being detected (#9254) de9fb4b
- [Bugfix] Sets `is_first_step_output` for TPUModelRunner (#9202) c6cf929
- [bugfix] fix f-string for error (#9295) Signed-off-by: Prashant Gupta <[email protected]> d11b46f
- [BugFix] Fix tool call finish reason in streaming case (#9209) Signed-off-by: Max de Bayser <[email protected]> ec10cb8
- [SpecDec] Remove Batch Expansion (2/3) (#9298) 89feb4c
- [Bugfix] Fix bug of xformer prefill for encoder-decoder (#9026) 00298e0
- [Misc][Installation] Improve source installation script and doc (#9309) Co-authored-by: youkaichao <[email protected]> 2b184dd
- [Bugfix]Fix MiniCPM's LoRA bug (#9286) 250e26a
- [CI] Fix merge conflict (#9317) f519902
- [Bugfix] Bandaid fix for speculative decoding tests (#9327) 16b24e7
- [Model] Molmo vLLM Integration (#9016) Co-authored-by: sanghol <[email protected]> Co-authored-by: Roger Wang <13... dfe43a2
- [Hardware][intel GPU] add async output process for xpu (#8897) 4141608
- and 475 more ...
sixsixcoder created a comment on an issue on THUDM/GLM-4-Voice
你这是哪个requirement.txt，voice依赖里面没有pyinin啊
sixsixcoder created a comment on an issue on THUDM/GLM-4-Voice
理论上m系列芯片是可行的，但是GLM-4-voice好像`model_server.py`可以运行在cpu上，但是`web_demo.py`好像只能运行在GPU上
sixsixcoder created a comment on an issue on THUDM/GLM-4-Voice
python 3.12 好像确实有点问题，我用的python 3.10可以运行
sixsixcoder created a comment on an issue on THUDM/GLM-4-Voice
根据报错信息，您需要安装pip install -U bitsandbytes
sixsixcoder created a comment on an issue on THUDM/GLM-4
Please refer to this issue：https://github.com/THUDM/GLM-4/issues/578
sixsixcoder created a comment on an issue on THUDM/GLM-4-Voice
感谢您帮忙排查问题，看起来是H20显卡不支持bf16类型加载的问题，请有同样问题的开发者移步 https://github.com/vllm-project/vllm/issues/4392 尝试解决该问题
sixsixcoder created a comment on an issue on THUDM/GLM-4-Voice
from: https://github.com/THUDM/GLM-4-Voice/issues/91
sixsixcoder closed an issue on THUDM/GLM-4
glm4v_server.py 可以支持视频分析吗？
### Feature request / 功能建议 glm4v_server.py 可以支持视频分析吗？ ### Motivation / 动机 glm4v_server.py 可以增加支持视频分析吗？ 类似 image_url 一样，支持 video_url ### Your contribution / 您的贡献 看dify 已经可以对接到智谱AI的API了 **...
sixsixcoder closed an issue on THUDM/GLM-4
使用llama-factory进行lora微调后，推理正常，但是tools调用异常
### System Info / 系統信息 操作系统=ubuntu20.04 显卡=4 x v100 model=glm-9b-chat python=3.11.8 llama-factory=0.8.3 transformers=4.43.3 vllm=0.5.3.post1 ### Who can help? / 谁可以帮助到您？ _No response_ #...
sixsixcoder created a comment on an issue on THUDM/GLM-4
> @sixsixcoder 调用工具不报错了。通过llamafactory微调glm4后乱码是什么原因导致的啊，随便一个训练数据都是这个问题。 (glm4) [root@adsl-172-10-0-187 basic_demo]# python openai_api_request.py ChatCompletion(id='chatcmpl-ZPaE1pu1lJo7XBetF5gIRHY...
sixsixcoder created a comment on an issue on THUDM/GLM-4-Voice
> > > Floating point exception (core dumped) > > > > > > 可以给出更详细的报错信息吗 > > 这是控制台从启动到结束的所有信息 (glm) root@autodl-container-9ef943a6c4-5e23a6d1:~/autodl-tmp/GLM-4-Voice# python model_server.py ...
sixsixcoder created a comment on an issue on THUDM/GLM-4
GLM-4v不支持视频输入分析，我不清楚dify中是如何实现的，你可以使用[THUDM/CogVLM2](https://github.com/THUDM/CogVLM2)模型进行视频分析
sixsixcoder created a comment on an issue on THUDM/GLM-4-Voice
> Floating point exception (core dumped) 可以给出更详细的报错信息吗
sixsixcoder created a comment on an issue on THUDM/GLM-4-Voice
理论上可以，这个模型的原理是输出文本和音频token，然后编码音频token成语音，你可以尝试hack[web_demo.py](https://github.com/THUDM/GLM-4-Voice/blob/main/web_demo.py)文件，截取文本进行输出。
sixsixcoder created a comment on an issue on THUDM/GLM-4-Voice
启动模型服务时，将device设置成cpu即可 ``` python model_server.py --host localhost --model-path THUDM/glm-4-voice-9b --port 10000 --dtype bfloat16 --device cpu ```
sixsixcoder created a comment on an issue on THUDM/GLM-4-Voice
这是我的开发环境可以完美运行，仅供参考，多余的依赖可以忽略，直接按照[requirements.txt](https://github.com/THUDM/GLM-4-Voice/blob/main/requirements.txt)安装依赖应该是没问题的 ``` absl-py 2.1.0 accelerate 1.0...
sixsixcoder created a comment on an issue on THUDM/GLM-4-Voice
你的torchaudio版本是多少，最好能看一下你的pip list还有操作步骤
sixsixcoder closed an issue on THUDM/GLM-4
web推理demo错误Exception in ASGI application
运行：basic_demo# python trans_web_vision_demo.py 报错如下：  pydantic.errors.PydanticSchemaGenerationError: Un...
sixsixcoder created a comment on an issue on THUDM/GLM-4-Voice
我的软硬件环境 ``` GPU A800-SXM4-80GB cuda 12.1 torch 2.4.0 torchaudio 2.4.0 transformers 4.45.2 python 3.10 显存 80G 精度 BF16 max_new_tokens = 256 ``` 我的测试结果，我迭代了3次，计算了平均首token时延和平均解码时延 ``` Av...
sixsixcoder created a comment on an issue on THUDM/GLM-4-Voice
You can follow the detailed technical report released later
sixsixcoder created a comment on an issue on THUDM/GLM-4-Voice
尝试安装 ```Shell pip install gradio==4.44.1 ```
sixsixcoder created a comment on an issue on THUDM/GLM-4
这不太好直接对比，精调后的模型在某些场景下能达到比原始模型更加好的效果。GLM-4 github仓库只是提供最小的使用示例demo，提供一个中英文输入的理解能力基座，更多的用法、更多的场景还需要开发者进一步微调或使用框架实现。