openchat
openchat copied to clipboard
imoneoi
OpenChat: Advancing Open-source Language Models with Imperfect Data
I am using the following method for openchat response generation ``` model_open_chat = AutoModelForCausalLM.from_pretrained("openchat/openchat_3.5", device_map="auto", temperature=0.1,torch_dtype=torch.bfloat16) tokenizer_open_chat = AutoTokenizer.from_pretrained("openchat/openchat_3.5") prompt = "" model_input = tokenizer_open_chat(prompt, return_tensors="pt").to("cuda") response = tokenizer_open_chat.decode(model_open_chat.generate(**model_input,temperature=0.1, max_length=5000)[0],...
Sometimes it is strange that it answers in Chinese while query is English. I check that there is no chinese characters in query.
As the essay said, the fine-tuning proccess consists two parts: c-SFT and c-RLFT. However, it seems like the train process in code is only c-SFT. I don't know how to...
I installed ochat into new venv with `pip3 install ochat`. Then I run the server with `python -m ochat.serving.openai_api_server --model openchat/openchat-3.6-8b-20240522 --model-type openchat_3.6` However, when trying curl `curl http://localhost:18888/v1/chat/completions -H...
ChatGPT helped remove all the swear words. I've encountered difficulties trying to set this up on Ubuntu, MacOS, and Windows. I've noticed some inconsistencies in the instructions, and it seems...
Openchat.team was broken for months, last update it started working after I deleted cache and lost all dialogues, but now it is down yet again and been down for a...
Seems like Qwen/Qwen2-7B-Instruct outperforms llama 3. Any plans to make your version of this model?
Thank you for your work! I find a bug in README, in the section of Training. You mentioned that the data needs to be pre-tokenizing, but the params are not...
