!!! note This example was adapted from Morgan Giraud on our Discord. You can find their twitter here. Thank you Morgan!
Outlines provides experimental support for context-free grammars (CFGs) for text generation. Future versions will provide more comprehensive support for structured outputs.
SQL is a context-free language, meaning that the structure of the query is independent of the content.
| """ | |
| pip install outlines torch==2.4.0 transformers accelerate typing-extensions pillow pdf2image rich requests | |
| may need to install tkinter: https://stackoverflow.com/questions/25905540/importerror-no-module-named-tkinter | |
| sudo apt-get install poppler-utils | |
| """ | |
| from enum import Enum | |
| from io import BytesIO |
| import outlines | |
| sql_grammar = """ | |
| start: set_expr -> final | |
| set_expr: query_expr | |
| | set_expr "UNION"i ["DISTINCT"i] set_expr -> union_distinct | |
| | set_expr "UNION"i "ALL"i set_expr -> union_all | |
| | set_expr "INTERSECT"i ["DISTINCT"i] set_expr -> intersect_distinct | |
| | set_expr "EXCEPT"i ["DISTINCT"i] set_expr -> except_distinct |
| import outlines | |
| from transformers import BitsAndBytesConfig | |
| # Load the model | |
| model = outlines.models.transformers( | |
| "microsoft/Phi-3-mini-4k-instruct", | |
| model_kwargs={ | |
| 'quantization_config':BitsAndBytesConfig( | |
| # Load the model in 4-bit mode | |
| load_in_4bit=True, |
| #!/usr/bin/env bash | |
| program=" | |
| import Markdown | |
| map(ARGS) do f | |
| c = if isfile(f) | |
| read(f, String) | |
| else | |
| f | |
| end | |
| display(Markdown.parse(c)) |
| # Set up environment | |
| import Pkg; Pkg.activate(".") | |
| # Imports | |
| using Distributions | |
| using Plots | |
| using Polynomials | |
| using Optim | |
| using ForwardDiff |
| Formula: | |
| returns ~ BITX + lag_returns + log_active + lag_active + log_avg_size + log_med_size + log_mode_size + native_transactions + :(fe(yearmonth)) | |
| Fixed Effect Model | |
| ======================================================================================= | |
| Number of obs: 1066 Degrees of freedom: 44 | |
| R2: 0.157 R2 Adjusted: 0.122 | |
| F Statistic: 12.4004 p-value: 0.000 | |
| R2 within: 0.106 Iterations: 1 | |
| Converged: true |
| using Turing, MCMCChains | |
| using AbstractMCMC | |
| using JLD2, FileIO | |
| import Random: GLOBAL_RNG | |
| # Create a model. | |
| @model model(y) = begin | |
| μ ~ Normal(0, 1) | |
| s ~ InverseGamma(2,3) |
| using Turing | |
| using CSVFiles | |
| using DataFrames | |
| using Dates | |
| using StatsPlots | |
| # The TV syntax allows sampling to be type-stable. | |
| @model garch(r, ::Type{TV}=Vector{Float64}) where {TV} = begin | |
| T = length(r) |
| using MCMCChains | |
| function sim(nsamples) | |
| cnt = 0 | |
| for i in 1:1000 | |
| chn = Chains(randn(nsamples,1,1)) | |
| ess = describe(chn)[1].df[:ess][1] | |
| ess > nsamples ? cnt+=1 : nothing | |
| end | |
| return cnt |