d-matrix
commited on
Commit
•
a3da97a
1
Parent(s):
1f48029
Update README.md
Browse files
README.md
CHANGED
@@ -36,14 +36,14 @@ model-index:
|
|
36 |
type: dmx-perlexity
|
37 |
value: 46.570838928222656
|
38 |
---
|
39 |
-
This is a d-Matrix functional reference of the GPT2 model family, of the following revisions
|
40 |
- [`distilgpt2`](https://huggingface.co/distilbert/distilgpt2)
|
41 |
- [`gpt2`](https://huggingface.co/openai-community/gpt2)
|
42 |
- [`gpt2-medium`](https://huggingface.co/openai-community/gpt2-medium)
|
43 |
- [`gpt2-large`](https://huggingface.co/openai-community/gpt2-large)
|
44 |
-
- [`gpt2-xl`](https://huggingface.co/openai-community/gpt2-xl)
|
45 |
|
46 |
-
The reference provides the following functional configurations
|
47 |
Configuration | Explanation
|
48 |
:-- | :--
|
49 |
**`BASELINE`** | a reference functionally equivalent to the original model
|
@@ -66,7 +66,8 @@ Prerequisites:
|
|
66 |
>>> pipe = pipeline(
|
67 |
>>> "text-generation",
|
68 |
>>> model="d-matrix/gpt2",
|
69 |
-
>>> revision="gpt2-xl",
|
|
|
70 |
>>> use_auth_token=os.environ.get("HUGGING_FACE_HUB_TOKEN"),
|
71 |
>>> trust_remote_code=True,
|
72 |
>>> # device_map="auto", # enabling model parallel on multi-GPU nodes
|
@@ -75,8 +76,6 @@ Prerequisites:
|
|
75 |
>>> pipe.model, monkey_patched=False, hf=True, input_names=["input_ids", "labels"]
|
76 |
>>> )
|
77 |
|
78 |
-
>>> pipe.model.transform("/path/to/BASIC.yaml")
|
79 |
-
|
80 |
>>> perplexity = evaluate.load("d-matrix/dmx_perplexity", module_type="metric")
|
81 |
>>> input_texts = load_dataset("ptb_text_only", "penn_treebank", split="test")["sentence"]
|
82 |
>>> results = perplexity.compute(model=pipe.model.body, references=input_texts)
|
|
|
36 |
type: dmx-perlexity
|
37 |
value: 46.570838928222656
|
38 |
---
|
39 |
+
This is a d-Matrix functional reference of the GPT2 model family, of the following *revisions*:
|
40 |
- [`distilgpt2`](https://huggingface.co/distilbert/distilgpt2)
|
41 |
- [`gpt2`](https://huggingface.co/openai-community/gpt2)
|
42 |
- [`gpt2-medium`](https://huggingface.co/openai-community/gpt2-medium)
|
43 |
- [`gpt2-large`](https://huggingface.co/openai-community/gpt2-large)
|
44 |
+
- [`gpt2-xl`](https://huggingface.co/openai-community/gpt2-xl)
|
45 |
|
46 |
+
The reference provides the following functional *configurations*:
|
47 |
Configuration | Explanation
|
48 |
:-- | :--
|
49 |
**`BASELINE`** | a reference functionally equivalent to the original model
|
|
|
66 |
>>> pipe = pipeline(
|
67 |
>>> "text-generation",
|
68 |
>>> model="d-matrix/gpt2",
|
69 |
+
>>> revision="gpt2-xl",
|
70 |
+
>>> dmx_config="BASELINE",
|
71 |
>>> use_auth_token=os.environ.get("HUGGING_FACE_HUB_TOKEN"),
|
72 |
>>> trust_remote_code=True,
|
73 |
>>> # device_map="auto", # enabling model parallel on multi-GPU nodes
|
|
|
76 |
>>> pipe.model, monkey_patched=False, hf=True, input_names=["input_ids", "labels"]
|
77 |
>>> )
|
78 |
|
|
|
|
|
79 |
>>> perplexity = evaluate.load("d-matrix/dmx_perplexity", module_type="metric")
|
80 |
>>> input_texts = load_dataset("ptb_text_only", "penn_treebank", split="test")["sentence"]
|
81 |
>>> results = perplexity.compute(model=pipe.model.body, references=input_texts)
|