THUDM/CogVLM issues and pull requests

#528 - 微调效果

Issue - State: closed - Opened by GuGuGuGun 30 days ago - 1 comment

#527 - sat 模型中的 lm_head 和 transformer.word_embeddings 有什么区别

Issue - State: open - Opened by GondorFu about 2 months ago

#526 - Demo is dead, Streamlit link is not accessible

Issue - State: open - Opened by AayushSameerShah about 2 months ago

#525 - 升级到 vlm2后，低像素图片识别效果明显变低，甚至不如老的 vm

Issue - State: open - Opened by schild 2 months ago

#524 - KeyError: 'cogvlm-chat'

Issue - State: open - Opened by nulllt 2 months ago

#523 - CogAgent fine-tune python script

Issue - State: open - Opened by SamithVa 2 months ago

#522 - How to set Prompt in fine tuning

Issue - State: open - Opened by piperino11 2 months ago

#521 - 本地服务器部署，对操作系统有什么建议和推荐吗

Issue - State: open - Opened by schild 3 months ago

#520 - 微调的时候发现没有test_config_bf16.json

Issue - State: closed - Opened by Akhim-yun 3 months ago

#519 - cogvlm_grounding 并行调用和串行结果存在一定差异

Issue - State: open - Opened by qinzhenyi1314 3 months ago

#518 - Why is it that running inference on two cards is slower than running on a single card?

Issue - State: open - Opened by Leeon-K 3 months ago

#517 - Better set a specific version of transformers lib

Issue - State: open - Opened by cocoshe 3 months ago - 1 comment

#516 - AttributeError: 'str' object has no attribute 'shape'

Issue - State: open - Opened by spol25 3 months ago - 1 comment

#515 - When running the code of the module finetune_demo on the windosw11 system, an error will be reported

Issue - State: open - Opened by gyjlll 3 months ago

#514 - 無法複現飛書上的示例

Issue - State: open - Opened by d223302 3 months ago

#513 - Reproduce results on visual7w.

Issue - State: open - Opened by sleepyshep 4 months ago - 2 comments

#512 - run evaluate_cogagent_demo.py error from sat import mpu, get_args, get_tokenizer error

Issue - State: open - Opened by gyjlll 4 months ago

#511 - 拉了一个多模态大模型技术交流群，大家可以加入进来进行技术交流

Issue - State: open - Opened by feihuamantian 4 months ago

#510 - How to evaluate the performance?

Issue - State: open - Opened by Originlightwkp 4 months ago - 1 comment

#509 - a trouble in finetune chatglm: ZeroDivisionError: integer division or modulo by zero

Issue - State: open - Opened by Originlightwkp 4 months ago - 1 comment

#508 - What's the difference between two CogVLM in Table 2 and Table 2

Issue - State: open - Opened by viyjy 4 months ago

#507 - Did you fine-tune VE and MLP Adaptor at both pre-training stage and SFT stage?

Issue - State: open - Opened by viyjy 4 months ago

#506 - 类型问题

Issue - State: open - Opened by Shawnzheng011019 4 months ago - 1 comment

#505 - 预训练阶段段loss下降后升高了，模型训崩了

Issue - State: open - Opened by liuheng0111 4 months ago - 2 comments

#504 - A bug while running cli_demo_hf.py

Issue - State: closed - Opened by JesseZZZZZ 4 months ago - 2 comments

#503 - 请问怎么用transformer来进行推理？

Issue - State: open - Opened by Zeng555 4 months ago - 2 comments

#502 - vllm适配

Issue - State: open - Opened by weiminw 4 months ago - 2 comments

#501 - 加载报错

Issue - State: open - Opened by itbithubman 5 months ago

#500 - openai_api.py调用的模型类型

Issue - State: open - Opened by wolfcx5 5 months ago - 1 comment

#499 - openai_api.py

Issue - State: closed - Opened by wolfcx5 5 months ago

#498 - 请问为什么使用cogvlm2的demo的时候无法使用grounding能力？

Issue - State: open - Opened by Honey-Mydear 5 months ago

#496 - Confidence score

Issue - State: open - Opened by joel-simp 5 months ago - 1 comment

#495 - 批量推理

Issue - State: open - Opened by zcqzcqzcq88 5 months ago

#494 - Error of visual.py file / load model from local

Issue - State: closed - Opened by joel-simp 5 months ago

#493 - raise IndexError('piece id is out of range.') IndexError: piece id is out of range.

Issue - State: closed - Opened by stay-leave 5 months ago - 2 comments

#492 - 请问能否实现描述图片某个位置然后返回相应坐标的功能

Issue - State: closed - Opened by lplplpoooouuu 5 months ago - 7 comments

#491 - 请问你们的代码使用了流行的agent框架吗，比如Langchain,如果没有使用，请问一下为什么不使用Langchain呢？目前做agent开发需要使用什么框架呢

Issue - State: closed - Opened by TheUnbearableLightnessofBeing 5 months ago - 1 comment

#490 - 如何多机多卡微调模型？

Issue - State: closed - Opened by zcqzcqzcq88 5 months ago

#489 - Hi，请问后续会给出CogAgent在mind2web数据集上评估的代码吗？

Issue - State: closed - Opened by micro-cosmos 6 months ago - 1 comment

#488 - torch.cuda.OutOfMemoryError: CUDA out of memory.

Issue - State: closed - Opened by mwaikul 6 months ago - 1 comment

#487 - 更新demo地址

Pull Request - State: closed - Opened by zRzRzRzRzRzRzR 6 months ago

#486 - Cogvlm2

Pull Request - State: closed - Opened by 1049451037 6 months ago

#485 - Cogagent demo can not be accessed

Issue - State: closed - Opened by Honey-Mydear 6 months ago - 1 comment

#484 - 原来带grounding功能的是哪个web demo地址?

Issue - State: closed - Opened by JackieXuu 6 months ago - 2 comments

#483 - 我该使用什么格式的输入来用模型进行visual grounding 任务？

Issue - State: open - Opened by lxiaohaung 6 months ago - 1 comment

#482 - 加载cogvlm-chat-hf模型报错 Error while deserializing header: MetadataIncompleteBuffer

Issue - State: open - Opened by xxWeiDG 6 months ago - 1 comment

#481 - Llava

Pull Request - State: closed - Opened by OliverSavolainen 6 months ago

#480 - Add link to CogVLM2

Pull Request - State: closed - Opened by Sleepychord 6 months ago

#479 - 两张3090微调CogVLM的可能性？

Issue - State: closed - Opened by KDD2018 6 months ago - 1 comment

#478 - 如何构建CogAgent的微调数据集？

Issue - State: closed - Opened by Li-mingshen 6 months ago - 1 comment

#477 - 运行微调脚本报错缺少相关参数

Issue - State: closed - Opened by 1212wuhu 6 months ago - 2 comments

#476 - 关于模型视觉定位原理

Issue - State: closed - Opened by zcqzcqzcq88 6 months ago

#475 - CogVLM源代码是否支持多轮对话训练

Issue - State: closed - Opened by elesun2018 6 months ago - 5 comments

#474 - CogAgent 视觉预训练模型 EVA2-CLIP-L

Issue - State: open - Opened by hzhiyuan 6 months ago - 1 comment

#473 - Chat with PDF documentation instead of images

Issue - State: open - Opened by moncefarajdal 6 months ago

#472 - 我想用同样的promt，在每次都清除上下文的情况下得到3种答案，为什么结果都是一样的

Issue - State: open - Opened by tygogogo 6 months ago - 2 comments

#471 - Could we replace the vicuna-7b directly with stronger llm?

Issue - State: closed - Opened by leoozy 6 months ago - 1 comment

#470 - Deploy CogVLM using Docker

Issue - State: closed - Opened by moncefarajdal 6 months ago - 1 comment

#469 - Code of finetuning the cogagent on Mind2Web ?

Issue - State: closed - Opened by leoozy 6 months ago - 2 comments

#468 - Using CogVLM as an API

Issue - State: closed - Opened by moncefarajdal 6 months ago - 1 comment

#467 - Running Gradio app locally results in inappropriate error: "NETWORK ERROR DUE TO HIGH TRAFFIC. PLEASE REGENERATE OR REFRESH THIS PAGE."

Issue - State: closed - Opened by aus10powell 7 months ago - 1 comment

#466 - [CogVLM-chat-v1.1] LM weights are different with vicuna-7b-v1.5

Issue - State: open - Opened by minostauros 7 months ago - 4 comments

#464 - Chat using one image and three prompt

Issue - State: open - Opened by nlylmz 7 months ago - 1 comment

#463 - Deploy

Issue - State: closed - Opened by liangdebugger 7 months ago - 2 comments

#461 - 关于模型量化

Issue - State: closed - Opened by elesun2018 7 months ago - 14 comments

#459 - 用cogagent-chat-hf直接做预测结果中出现数字列表（basic-demo）

Issue - State: closed - Opened by WXLJZ 7 months ago - 4 comments

#456 - 使用 fp16 训练，merge lora 之后的模型推理结果异常

Issue - State: open - Opened by GondorFu 7 months ago - 5 comments

#450 - 8张A800(80G)微调Cogagent依然报错： CUDA out of memory

Issue - State: open - Opened by GuoXu-booo 7 months ago - 6 comments

#447 - 参数说明

Issue - State: open - Opened by chensongcan 7 months ago

#445 - 关于多轮对话

Issue - State: open - Opened by Jackluisus 7 months ago - 9 comments

#441 - Some Clarification

Issue - State: closed - Opened by deadpipe 7 months ago - 1 comment

#440 - Can I run CogVLM using actual openai API

Issue - State: open - Opened by PhilipAmadasun 7 months ago - 3 comments

#438 - could it finetune in qlora or quant 4/8?

Issue - State: closed - Opened by wyclike 7 months ago - 2 comments

#434 - Release of cogagent dataset

Issue - State: closed - Opened by lightmatmul 7 months ago - 2 comments

#433 - web demo无法中文对话，但可以英文对话

Issue - State: closed - Opened by ydygxu123 7 months ago - 1 comment

#432 - Running CogVLM and CogAgent on MPS

Issue - State: closed - Opened by Allisterlim 8 months ago - 5 comments

#430 - 断点继续训练如何设置

Issue - State: open - Opened by elesun2018 8 months ago - 15 comments

#428 - CogVLM源代码是否支持多轮对话训练

Issue - State: closed - Opened by elesun2018 8 months ago - 9 comments

#419 - ValueError: model_parallel_size is inconsistent with prior configuration.We currently do not support changing model_parallel_size.

Issue - State: open - Opened by Hakan-Khenda 8 months ago - 1 comment

#418 - Web demo无法对话。

Issue - State: closed - Opened by serend1p1ty 8 months ago - 3 comments

#416 - how to deny the auto created output.png

Issue - State: closed - Opened by HarryXuancy 8 months ago - 2 comments

#407 - 请问一个多模态大模型代替多个yolo检测小模型场景下多轮对话数据格式如何设计？

Issue - State: closed - Opened by elesun2018 8 months ago - 9 comments

#397 - Inquiry on CogVLM's capacity for comprehension

Issue - State: closed - Opened by PhilipAmadasun 8 months ago - 1 comment

#395 - add Dockerfile and deploy.py

Pull Request - State: closed - Opened by abrichr 8 months ago - 1 comment

#390 - Replacing the 224-image_size EVA model in CogAgent model with the 490-image_size one

Issue - State: closed - Opened by jeffchan0830 9 months ago

#388 - 如何使用CogVLM finetune训练自有数据集具有视觉定位能力的多模态大模型？

Issue - State: closed - Opened by elesun2018 9 months ago - 4 comments

#386 - Could you please provide a simple script to use your multimodel for extracting image features and text features like openclip with encode_image() or encode_text()?

Issue - State: closed - Opened by AlexandraDobrescu 9 months ago

#383 - 官方代码的hf版本的cogvlm模型推理报错Out of Memory

Issue - State: closed - Opened by zhaohongyin 9 months ago - 4 comments

#382 - Huggingface版本是否提供lora微调的支持？

Issue - State: closed - Opened by xuanyaoming 9 months ago - 4 comments

#377 - How to add lora matrix to CogAgent model via Peft?

Issue - State: closed - Opened by TankNee 9 months ago - 1 comment

#342 - Will the pretraining-data Common Crawl Screenshot 400K and the 140 million REC and REG question-answer pairs be released?

Issue - State: closed - Opened by XuRui314 10 months ago - 3 comments

#341 - web demo时error message list index out of range

Issue - State: closed - Opened by elesun2018 10 months ago - 4 comments

#338 - cogagent 24GB gpu memory is not enough

Issue - State: closed - Opened by 3303341078 10 months ago - 2 comments

#333 - Cogvlm-How to batch inference？

Issue - State: closed - Opened by aa7784171 10 months ago - 5 comments

#269 - 微调CogAgent-chat模型，MP_SIZE设为8，不采用lora微调，保存checkpoint时出错

Issue - State: closed - Opened by VittorioDong 10 months ago - 2 comments

#256 - RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:6 and cuda:5!

Issue - State: closed - Opened by Luoyang144 11 months ago - 9 comments

GitHub / THUDM/CogVLM issues and pull requests