The super-json-mode from varunshenoy

Not able to run the Mistral 7B example from README

Hi Varun, Great work on the repo so far.

I am trying to integrate the super json module to generate neatly formatted jsons for RAG evaluation using custom LLMs

The Mistral 7B example mentioned in README is not working in my machine. Can you please have a look? Also, import json was missing.

Code


from transformers import AutoTokenizer, AutoModelForCausalLM
from superjsonmode.integrations.transformers import StructuredOutputForModel
from pydantic import BaseModel
import json

device = "cuda"
model = AutoModelForCausalLM.from_pretrained("mistralai/Mistral-7B-Instruct-v0.2").to(device)
tokenizer = AutoTokenizer.from_pretrained("mistralai/Mistral-7B-Instruct-v0.2")

# Create a structured output object
structured_model = StructuredOutputForModel(model, tokenizer)

passage = """..."""

class QuarterlyReport(BaseModel):
    company: str
    stock_ticker: str
    date: str
    reported_revenue: str
    dividend: str

prompt_template = """[INST]{prompt}

Based on this excerpt, extract the correct value for "{key}". Keep it succinct. It should have a type of `{type}`.[/INST]

{key}: """

output = structured_model.generate(passage,
                                   extraction_prompt_template=prompt_template,
                                   schema=QuarterlyReport,
                                   batch_size=6)

print(json.dumps(output, indent=2))

# {
#   "company": "NVIDIA",
#   "stock_ticker": "NVDA",
#   "date": "2023-10",
#   "reported_revenue": "18.12 billion dollars",
#   "dividend": "0.04"
# }

Runtime Log

Loading checkpoint shards: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:01<00:00,  2.38it/s]
Traceback (most recent call last):
  File "/home/deval/Documents/Work/Deval/ES/LlamaSearch/src/temp1.py", line 29, in <module>
    output = structured_model.generate(passage,
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/deval/Document/Work/miniconda3/envs/es/lib/python3.11/site-packages/superjsonmode/integrations/transformers.py", line 45, in generate
    embeds = self.tokenizer(prompts, return_tensors="pt", padding=True).to(
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/deval/Document/Work/miniconda3/envs/es/lib/python3.11/site-packages/transformers/tokenization_utils_base.py", line 2803, in __call__
    encodings = self._call_one(text=text, text_pair=text_pair, **all_kwargs)
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/deval/Document/Work/miniconda3/envs/es/lib/python3.11/site-packages/transformers/tokenization_utils_base.py", line 2889, in _call_one
    return self.batch_encode_plus(
           ^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/deval/Document/Work/miniconda3/envs/es/lib/python3.11/site-packages/transformers/tokenization_utils_base.py", line 3071, in batch_encode_plus
    padding_strategy, truncation_strategy, max_length, kwargs = self._get_padding_truncation_strategies(
                                                                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/deval/Document/Work/miniconda3/envs/es/lib/python3.11/site-packages/transformers/tokenization_utils_base.py", line 2708, in _get_padding_truncation_strategies
    raise ValueError(
ValueError: Asking to pad but the tokenizer does not have a padding token. Please select a token to use as `pad_token` `(tokenizer.pad_token = tokenizer.eos_token e.g.)` or add a new pad token via `tokenizer.add_special_tokens({'pad_token': '[PAD]'})`.

SGLang Integration

Nice project!

I believe this project can greatly benefit from https://github.com/sgl-project/sglang. You can try to use SGLang as a backend for local models.

The fast JSON decoding feature can help you force additional constraints and probably help the nested JSON schemas. You can find the example here.
The RadixAttention feature can help you reuse the KV cache for the shared prefix. You can find one example on parallel decoding here.

token usage

maybe this is a dumb question, but if you decompose the schema into K queries to extract values in parallel, don't you have to repeat the text input K times? Or do you have a way around this?

We have an application where users supply the JSON schema for use directly. I saw that you convert the Pydantic model to JSON internally. Any thoughts on supporting JSON schema directly instead of JSON models or suggestions for converting json schema strings into the pydantic models?

Very cool work by the way. Thanks for this effort!

Ollama Integration using single processing at a time

Dear @varunshenoy,

I modified code to incorporate Ollama into the SuperJSON. However, It is unable to do batch processing. If you think it would be considerable placeholder until Ollama adds batch processing I can make a PR.

Best,
Naman

varunshenoy / super-json-mode Goto Github PK

super-json-mode's People

Contributors

Stargazers

Watchers

Forkers

super-json-mode's Issues

Not able to run the Mistral 7B example from README

Code

Runtime Log

SGLang Integration

token usage

Directly Provide JSON Schema

Ollama Integration using single processing at a time

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent