@microsoft-github-policy-service agree

Thank you!

@alxspiker I found a couple of problems with my implementation and are fixing them right now!

Anyway to support mmap? Seems like its not.

@alxspiker I have fixed all errors on my side but couldn't reproduce your error, but I added nmap to the settings!

@alxspiker At the moment you have to use my fork of llama-cpp-python to use guidance.
You will find the fork here:
https://github.com/Maximilian-Winter/llama-cpp-python

Related PR in llama-cpp-python: abetlen/llama-cpp-python#252

It would be awesome to use guidance with llama.cpp! I'm excited :)

@Maximilian-Winter this is great, thanks! It will probably be Monday before I can review it properly. Are the there any basic units tests we can add for this? (with small LMs that don't slow down the test process too much) ...might not be possible with LLaMA, but even a file with test that only run locally would be good so we can make sure this stays working :)

(I also just approved the unit tests to run for this)

@slundberg I have added a test file in the tests/llms folder called "test_llamacpp.py".
I used the test_transformers file as a template.

After many attempts I could not get the role chat to work

I've use this code:

import re
import guidance

# define the model we will use

settings = guidance.llms.LlamaCppSettings()
settings.n_gpu_layers = 14
settings.n_threads = 16
settings.n_ctx = 1024
settings.use_mlock = True
settings.model = "path/to/model"
# Create a LlamaCpp instance and pass the settings to it.
llama = guidance.llms.LlamaCpp(settings=settings)
guidance.llm = llama
def parse_best(prosandcons, options):
    best = int(re.findall(r'Best=(\d+)', prosandcons)[0])
    return options[best]

create_plan = guidance('''
{{#system~}}
You are a helpful assistant.
{{~/system}}

{{! generate five potential ways to accomplish a goal }}
{{#block hidden=True}}
{{#user~}}
I want to {{goal}}.
{{~! generate potential options ~}}
Can you please generate one option for how to accomplish this?
Please make the option very short, at most one line.
{{~/user}}

{{#assistant~}}
{{gen 'options' n=5 temperature=1.0 max_tokens=500}}
{{~/assistant}}
{{/block}}

{{! generate pros and cons for each option and select the best option }}
{{#block hidden=True}}
{{#user~}}
I want to {{goal}}.

Can you please comment on the pros and cons of each of the following options, and then pick the best option?
---{{#each options}}
Option {{@index}}: {{this}}{{/each}}
---
Please discuss each option very briefly (one line for pros, one for cons), and end by saying Best=X, where X is the best option.
{{~/user}}

{{#assistant~}}
{{gen 'prosandcons' temperature=0.0 max_tokens=500}}
{{~/assistant}}
{{/block}}

{{! generate a plan to accomplish the chosen option }}
{{#user~}}
I want to {{goal}}.
{{~! Create a plan }}
Here is my plan:
{{parse_best prosandcons options}}
Please elaborate on this plan, and tell me how to best accomplish it.
{{~/user}}

{{#assistant~}}
{{gen 'plan' max_tokens=500}}
{{~/assistant}}''')
out = create_plan(
    goal='read more books',
    parse_best=parse_best # a custom python function we call in the program
)

@slundberg I have implemented proper role_end again, also implemented streaming support.

@slundberg I think the best way would be to test just locally. The smallest model right now is a 7B parameter model which is already 3.8gb of memory.

Just a note here, I was still getting some tokenization issues and realized it is going to be hard to maintain so much code that is similar between transformers and llamacpp, so I am going to try and push a proposal to share more code tonight.

I pushed a proposal in the form of LlamaCpp2, along with lots of updates to Transformers that are related because we will want to depend on them. I think we need to inherit from the Transformers LLM class because otherwise we duplicate lots of code that is tricky and should only live in one place :)

LlamaCpp2 does not work fully yet, but I am pushing to to see what you think @Maximilian-Winter.

thanks again for all the hard work pushing on this :)

@slundberg Will take a look later today

@slundberg Looks good to me! But I think the tokenizer of llama.cpp is bugged because it refuse to give me the eos or bos token.
It is always empty when I try to decode it from the id!

@slundberg Looks good to me! But I think the tokenizer of llama.cpp is bugged because it refuse to give me the eos or bos token.
It is always empty when I try to decode it from the id!

Yeah, I just think we can just return </s> directly for now. I just pushed a few more fixes. Can I hand this back over to you to wrap up? There is some difference with the way the logprobs are returned that is not quite matching how transformers returns it yet, but otherwise I think we are close!

I also noticed that logit bias processor inside llama-cpp-python seems to save the bias values after the local logits variable is already set:
https://github.com/abetlen/llama-cpp-python/blob/232880cbc677db1998afa240c25e58090f399072/llama_cpp/llama.py#L373-L383

@slundberg I will try to make it work later!

Please merge this microsoft or at least help support it. This would be HUGE for guidance

Is there progress with this? Oh I see there is already a message. Yes, there are a few of us spamming refresh on this...

Sorry, was very busy with other stuff at work! Will look into this!

Any progress?

Any updates on this?

Hmm. Looks like there's a conflict with 47b1cd4. Trying to use a519012, a merge which brings the former commit into the PR...

>>> import guidance
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/nix/store/0j6zxl1kvqkvhcw0i2chdqxx22xsi5sf-python3-3.10.11-env/lib/python3.10/site-packages/guidance/__init__.py", line 7, in <module>
    from ._program import Program
  File "/nix/store/0j6zxl1kvqkvhcw0i2chdqxx22xsi5sf-python3-3.10.11-env/lib/python3.10/site-packages/guidance/_program.py", line 17, in <module>
    from .llms import _openai
  File "/nix/store/0j6zxl1kvqkvhcw0i2chdqxx22xsi5sf-python3-3.10.11-env/lib/python3.10/site-packages/guidance/llms/__init__.py", line 7, in <module>
    from ._llama_cpp import LlamaCpp
  File "/nix/store/0j6zxl1kvqkvhcw0i2chdqxx22xsi5sf-python3-3.10.11-env/lib/python3.10/site-packages/guidance/llms/_llama_cpp.py", line 17, in <module>
    class LlamaCpp(LLM):
  File "/nix/store/0j6zxl1kvqkvhcw0i2chdqxx22xsi5sf-python3-3.10.11-env/lib/python3.10/site-packages/guidance/llms/_llama_cpp.py", line 21, in LlamaCpp
    cache = LLM._open_cache("_llama_cpp.diskcache")
AttributeError: type object 'LLM' has no attribute '_open_cache'

yeah this needs to be updated. I've tried forking Max's and git pulling Microsoft's main branch, but there have been a lot of changes since June so lots of things need tweaking.

Any updates on this?

Any plans on merging this at some point?

guidance's templating is miles more friendly to use than lmql.

But... guidance, are you still alive?

any update on this ???

@Maximilian-Winter thank you so much for all your hard work on this! Due to some external circumstances over the summer I couldn't come back to push it over the finish line until this fall (with v0.1). This PR strongly informed the design decisions we made for Llama.cpp support in v0.1 though so it was very useful.

I am closing this now since we now have full llama.cpp support in v0.1 :)