Ecosyste.ms: Timeline
Browse the timeline of events for every public repo on GitHub. Data updated hourly from GH Archive.
zRzRzRzRzRzRzR closed an issue on THUDM/CogVideo
How can we use LoRAs with current pipeline?
I am using this pipeline `pipe = CogVideoXPipeline.from_pretrained("THUDM/CogVideoX-5b", torch_dtype=torch.bfloat16).to("cuda")` Can we use LoRAs? Example LoRA : https://civitai.com/model...
zRzRzRzRzRzRzR created a comment on an issue on THUDM/CogAgent
在readme中已经显著提到了最低运行要求 <img width="832" alt="image" src="https://github.com/user-attachments/assets/04e28048-1e82-4b24-b491-19d5fd05539a" />
zRzRzRzRzRzRzR closed an issue on THUDM/CogAgent
运行CogAgent最低的显卡配置要求是什么？
### System Info / 系統信息 希望能添加运行该项目最低的配置要求说明 ### Who can help? / 谁可以帮助到您？ _No response_ ### Information / 问题信息 - [ ] The official example scripts / 官方的示例脚本 - [ ] My own modified scripts / 我自己修改的...
zRzRzRzRzRzRzR created a comment on an issue on THUDM/CogAgent
We use NVIDIA's `bitsandbytes` library for simple quantization, and there is a line in the CLI DEMO. ``` quantization_config=BitsAndBytesConfig(load_in_8bit=True), ``` However, this method will...
zRzRzRzRzRzRzR created a comment on an issue on THUDM/CogAgent
你可以在我们的技术文档中查看我们拼接prompt的所有细节，我们测试时候的使用的就是这样的组合，现在推理的超参数就是内部测试的时候使用。 关于训练技术，可以查看技术报告。
zRzRzRzRzRzRzR created a comment on an issue on THUDM/GLM-4
你用的是 glm-4-9b-chat 还是 glm-4-9b-chat-hf，后者vLLM不支持，会乱答
zRzRzRzRzRzRzR created a comment on an issue on THUDM/GLM-4
尝试进行更多轮训练，是否能实现相似效果，大概让loss降低到0.1，推理的时候保持贪婪采样
zRzRzRzRzRzRzR closed an issue on THUDM/GLM-4
大模型不知道今天日期，提示词告诉它也不行
### System Info / 系統信息 glm-4-9b-chat Ubuntu22 4090 ### Who can help? / 谁可以帮助到您？ 我用tool call方式问昨天发生的事情，大模型调用总是把2022年8月23日日期作为参数传给函数。 哪怕增加一个今天日期的function或者之间提示词告诉大模型今天日期也不行： 提示词如下： 你是一个xx领...
zRzRzRzRzRzRzR created a comment on an issue on THUDM/CogVideo
You should check our cli_demo, ``` pipe.load_lora_weights(lora_path, weight_name="pytorch_lora_weights.safetensors", adapter_name="test_1") ``` Here is a place to mount diffusers. It might help...
zRzRzRzRzRzRzR created a comment on an issue on THUDM/CogVideo
In the short term, we currently have no relevant plans, thank you for your understanding. If there are updates, we will release them in the update announcement.
zRzRzRzRzRzRzR closed an issue on THUDM/CogVideo
Any plan for "Identity-Preserving T2V Generation"
### Feature request / 功能建议 Identity-Preserving T2V Generation ### Motivation / 动机 many application ### Your contribution / 您的贡献 https://arxiv.org/pdf/2411.17440 Identity-Preserving Text-to-V...
zRzRzRzRzRzRzR created a comment on an issue on THUDM/CogVideo
Have you used load_in_4bit? This parameter is supported by our model, and there is content about quantization configuration in transformers. Can you manually check if a quantized configuration was ...
zRzRzRzRzRzRzR pushed 3 commits to main THUDM/GLM-Edge
zRzRzRzRzRzRzR closed a pull request on THUDM/GLM-Edge
Update convert_v.py
To align with the modelling update
zRzRzRzRzRzRzR pushed 1 commit to CogVideoX_dev THUDM/CogVideo
- put lora back(sat), unavailable running b080c6a
zRzRzRzRzRzRzR closed an issue on THUDM/CogAgent
不支持纯cpu 模式么？
### System Info / 系統信息 Linux Ubuntu20.04 ### Who can help? / 谁可以帮助到您？ _No response_ ### Information / 问题信息 - [X] The official example scripts / 官方的示例脚本 - [ ] My own modified scripts / 我自己修改的脚...
zRzRzRzRzRzRzR pushed 7 commits to CogVideoX_dev THUDM/CogVideo
- chore: update .gitignore - Add new ignore patterns for dataset and model directories - Update rules for development ... 3c930f4
- chore: remove unused configuration files after refactoring Delete accelerate configs, deepspeed config and host file... f6768d9
- feat: add video frame extraction tool Add utility script to extract first frames from videos, helping users convert ... b33d8db
- feat(datasets): implement video dataset modules - Add dataset implementations for text-to-video and image-to-video -... 34e3b35
- feat(models): add scaffolding aea3cf9
- add lora sat back 9af98ab
- Merge branch 'CogVideoX_dev' of github.com:THUDM/CogVideo into CogVideoX_dev fd5c885
zRzRzRzRzRzRzR pushed 6 commits to CogVideoX_dev THUDM/CogVideo
- chore: update .gitignore - Add new ignore patterns for dataset and model directories - Update rules for development ... 3c930f4
- chore: remove unused configuration files after refactoring Delete accelerate configs, deepspeed config and host file... f6768d9
- feat: add video frame extraction tool Add utility script to extract first frames from videos, helping users convert ... b33d8db
- feat(datasets): implement video dataset modules - Add dataset implementations for text-to-video and image-to-video -... 34e3b35
- feat(models): add scaffolding aea3cf9
- add lora sat back 9af98ab
zRzRzRzRzRzRzR pushed 27 commits to main THUDM/CogVideo
- fix bugs of image-to-video without image-condition 21693ca
- [fix]fix typo in train_cogvideox_image_to_video_lora.py 48ac9c1
- [bugfix] fix specific resolution setting Different models use different resolutions, for example, for the CogVideoX1... 2508c83
- [docs] improve help messages in argument parser Fix and clarify help documentation in parser.add_argument() to bette... ba85627
- chore: update .gitignore - Add new ignore patterns for dataset and model directories - Update rules for development ... 5cb9303
- chore: remove unused configuration files after refactoring Delete accelerate configs, deepspeed config and host file... 7b28224
- feat: add video frame extraction tool Add utility script to extract first frames from videos, helping users convert ... e3f6def
- feat(datasets): implement video dataset modules - Add dataset implementations for text-to-video and image-to-video -... 918ebb5
- feat(models): add scaffolding 85e00a1
- Add utils 78f655a
- Add constants.py a505f2e
- feat: add base trainer implementation and training script - Add Trainer base class with core training loop functiona... 60f6a3d
- Export file_utils.py 6971364
- feat(trainer): add validation functionality to Trainer class Add validation capabilities to the Trainer class includ... fa4659f
- Add type conversion and validation checks 2a6cca0
- feat: add latent caching for video encodings - Add caching mechanism to store VAE-encoded video latents to disk - Ca... 6eae5c2
- refactor: simplify dataset implementation and add latent precomputation - Replace bucket-based dataset with simpler ... 45d4045
- feat: add schemas module for configuration and state management Add Pydantic models to handle: - CLI arguments and c... 91d79fd
- feat: implement CogVideoX trainers for I2V and T2V tasks Add and refactor trainers for CogVideoX model variants: - I... a001842
- Change logger name to trainer 04a60e7
- and 7 more ...
zRzRzRzRzRzRzR pushed 135 commits to main zRzRzRzRzRzRzR/transformers
- Corrected typo in agent system prompts (#35143) 1452dc2
- Option to set 'non_blocking' for to(device) in BatchEncoding and BatchFeature (#34883) * Option to set 'non_blocking... de8a0b7
- Fix typo in EETQ Tests (#35160) fix 7238387
- Cleanup: continue the init refactor (#35167) Round 2 8e806a3
- Super tiny fix logging message (#35132) Update integration_utils.py 4bc39de
- Fixed typo of 'avilable' in prompts.py (#35145) fa8763c
- [CI] Fix bnb quantization tests with accelerate>=1.2.0 (#35172) 34f4080
- Fix `num_items_in_batch` not being an integer (#35115) In method `Trainer#get_batch_samples`, the return values shou... dada0fd
- Assisted decoding multi-gpu (#35116) * fix * move a few lines up 0938b57
- Fix file path for shard_num 1 with mllama converter (#35053) "#35049 fix path for num_shard 1" 80f2b16
- Support BatchNorm in Hubert pos_conv_emb as in fairseq (#34389) * Support BatchNorm in Hubert pos_conv_emb as in fai... 6acb4e4
- Remove unnecessary masked_fill in deberta models (#35182) 5fba3f9
- Fix DBRX LayerNorm init method (#35177) fix dbrx layernorm init 3e2769a
- Fixing GGUF support for StableLm (#35060) fix Co-authored-by: Marc Sun <[email protected]> e5c45a6
- [i18n-ar] Translated file : `docs/source/ar/community.md` into Arabic (#33027) * Add docs/source/ar/community.md to ... 425af6c
- Multiple typo fixes in NLP, Audio docs (#35181) Fixed multiple typos in Tutorials, NLP, and Audio sections 52d1354
- Only import torch.distributed if it is available (#35133) 217c47e
- [i18n-<languageCode>] Translating Benchmarks.md to Chinese (#35137) * add "Translating Benchmarks.md to Chinese " ... 91b8ab1
- [docs] Fix FlashAttention link (#35171) fix link 5290f6a
- Update data collator docstrings to accurately reference Nvidia tensor core compute capability version (#35188) updat... e850892
- and 115 more ...