MultiPL-T-StarCoderBase_1b/README.md

license, library_name, tags, datasets, metrics, model-index

license	library_name	tags	datasets	metrics	model-index
bigscience-openrail-m	transformers	code gpt_bigcode	nuprl/MultiPL-T	code_eval	name results MultiPLCoder-1b-OCaml task dataset metrics type text-generation name type MultiPL-HumanEval (Lua) nuprl/MultiPL-E type value name verified pass@1 0.173 pass@1 true type value name verified pass@1 0.113 pass@1 true type value name verified pass@1 0.097 pass@1 true

MultiPLCoder-1b

1 billion parameter version of MultiPLCoder, a set of StarCoder-based models finetuned on the MultiPL-T dataset. These models are state-of-the-art at low-resource languages, such as: Lua, Racket, and OCaml.

Language Revision Index

This is the revision index for the best-performing models for their respective langauge.

Langauge	Revision ID	Epoch
Lua	7e96d931547e342ad0661cdd91236fe4ccf52545	3
Racket	2cdc541bee1db4da80c0b43384b0d6a0cacca5b2	5
OCaml	e8a24f9e2149cbda8c3cca264a53c2b361b7a031	6

Usage

To utilize one of the models in this repository, you must first select a commit revision for that model from the table above. For example, to use the Lua model:

from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("nuprl/MultiPLCoder-1b")
lua_revision="7e96d931547e342ad0661cdd91236fe4ccf52545"
model = AutoModelForCausalLM.from_pretrained("nuprl/MultiPLCoder-1b", revision=lua_revision)

Note that the model's default configuration does not enable caching, therefore you must specify to use the cache on generation.

toks = tokenizer.encode("-- Hello World", return_tensors="pt")
out = model.generate(toks, use_cache=True,  do_sample=True, temperature=0.2, top_p=0.95, max_length=50)
print(tokenizer.decode(out[0], skip_special_tokens=True))

-- Hello World!
-- :param name: The name of the person to say hello to
-- :return: A greeting
local function say_hello(name)
  return "Hello ".. name
end