File size: 2,255 Bytes
93cadd1 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 |
---
license: apache-2.0
datasets:
- wikitext
- ptb_text_only
language:
- en
metrics:
- perplexity
pipeline_tag: text-generation
model-index:
- name: distilgpt2
results:
- task:
type: text-generation
dataset:
name: penn_treebank
type: ptb_text_only
metrics:
- name: perlexity@BASELINE
type: dmx-perlexity
value: 63.45857238769531
- name: perlexity@FALLBACK
type: dmx-perlexity
value: 64.36720275878906
- task:
type: text-generation
dataset:
name: wikitext2
type: wikitext-2-raw-v1
metrics:
- name: perlexity@BASELINE
type: dmx-perlexity
value: 46.05925369262695
- name: perlexity@FALLBACK
type: dmx-perlexity
value: 46.570838928222656
---
This is a quantized version of [DistilGPT2](https://huggingface.co/distilbert/distilgpt2). We provide the following two quantization configurations:
BASELINE: Everything in original format, equivalent to original model.
FALLBACK: Quantized Linear and Conv1D layers to BFP16. Added approximation functions for Layer Norm, GELU and Softmax.
### Usage Example
Prerequisites:
- Install dmx-mltools: "pip install dmx-mltools"
- clone this repo. "cd" to the cloned repo.
```python
>>> import os
>>> import torch
>>> from mltools import dmx
>>> from transformers import pipeline,AutoModelForCausalLM
>>> import evaluate
>>> from datasets import load_dataset
# Get model
>>> my_hf_token = os.environ.get("Dmatrix_HF_Token")
>>> device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
>>> pipe = pipeline(
>>> "text-generation",
>>> model="d-matrix/distilgpt2",
>>> device=device,
>>> use_auth_token=my_hf_token,
>>> )
>>> pipe.model = dmx.Model(pipe.model,monkey_patched=False,hf=True,input_names=["input_ids", "labels"])
# Configure quantization formats
>>> pipe.model.transform('FALLBACK.yaml')
# Evaluate
>>> perplexity = evaluate.load("d-matrix/dmx_perplexity", module_type="metric")
>>> input_texts = load_dataset("ptb_text_only", "penn_treebank", split="test")["sentence"]
>>> pipe.model.eval()
>>> results = perplexity.compute(model=pipe.model.body,references=input_texts)
>>> print(results)
{'loss': 4.164604187011719, 'perplexity': 64.36720275878906}
``` |