kungfu-eric

I'm running the example here: https://pandas-datareader.readthedocs.io/en/latest/remote_data.html#yahoo-finance-data and it fails: `df = web.DataReader('GE', 'yahoo', start='2019-09-10', end='2019-10-09')` `RemoteDataError: Unable to read URL: https://finance.yahoo.com/quote/GE/history?period1=1568088000&period2=1570679999&interval=1d&frequency=1d&filter=history Response Text: b'\n \n \n \n Yahoo\n \n \n...

The last few lines need to be fixed. This example doesn't repro with the current pip install langchain. Particularly, the RetreivalQA has some validation step: ``` --------------------------------------------------------------------------- ValidationError Traceback (most...

The RetrievalQA import doesn't resolve. Checking the source repo shows there is no chains folder: https://github.com/langchain-ai/langchain/tree/ed49cca1919c71fa51256346a51a5316196173c7/libs/community/langchain_community ``` --------------------------------------------------------------------------- ModuleNotFoundError Traceback (most recent call last) Cell In[1], line 9 7 from...

EDIT: yea I'm calling BS. I had a go and numerous reason why subclassing and monkeypatching huggingface implementation doesnt work. Have to rip out the model arch and manually displace...

> I've been round the houses with this as above - eventually got it working with a `stopSequence` of `[""]` - tells the engine to stop asking for more responses...

Observing this too https://github.com/ollama/ollama/issues/4212. Horrible perf regression

Confirmed that downgrading to 0.1.31 resolves this issue for me as per @zhqfdn suggestion It must be the new GPU detection added post 0.1.31. Should it revert to the cudart...

This should be good because it's not clear without inferencing a long sequence whether the sequence will fit in memory or not. The model and some default overhead (2k?) might...

Removing bfloat and fp32 flag and adding fp16 flag should get further

Encountering the same problem. It seems to be some issue with flash attention however the sample in the flash attention repo runs fine. I notice it clears one stage if...