Initial Query

Create a JSON metadata for these models: Mistral-7B-v0.1 RedPajama-INCITE-7B-Base RedPajama-INCITE-Base-3B-v1 falcon40b falcon7b gpt2-xl llama-65b llama-7b neo-1.3 neo-2.7 neo-6 open_llama_3b_v2 open_llama_7b_v2 opt-1.3b opt-6.7b pythia-1.4 pythia-1.4-dedup pythia-6.9 pythia-6.9-dedup With the format: {"Mistral-7B-v0.1": {"size": 7, dataset: "", "family": "Mistral"}, ...} where family is one of base = [ 'pythia', 'llama', 'Mistral', 'gpt2', 'opt', 'RedPajama', 'neo', 'open_llama', 'falcon' ] gpt2-xl is 1.5b parameters.

LLM Generation

Query

Create a JSON metadata for these models: Mistral-7B-v0.1 RedPajama-INCITE-7B-Base RedPajama-INCITE-Base-3B-v1 falcon40b falcon7b gpt2-xl llama-65b llama-7b neo-1.3 neo-2.7 neo-6 open_llama_3b_v2 open_llama_7b_v2 opt-1.3b opt-6.7b pythia-1.4 pythia-1.4-dedup pythia-6.9 pythia-6.9-dedup With the format: {"Mistral-7B-v0.1": {"size": 7, dataset: "", "family": "Mistral"}, ...} where family is one of base = [ 'pythia', 'llama', 'Mistral', 'gpt2', 'opt', 'RedPajama', 'neo', 'open_llama', 'falcon' ] gpt2-xl is 1.5b parameters.

Output

Here's a JSON metadata for the given models: 

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
{
    "Mistral-7B-v0.1": {"size": 7, "dataset": "", "family": "Mistral"},
    "RedPajama-INCITE-7B-Base": {"size": 7, "dataset": "", "family": "RedPajama"},
    "RedPajama-INCITE-Base-3B-v1": {"size": 3, "dataset": "", "family": "RedPajama"},
    "falcon40b": {"size": 40, "dataset": "", "family": "falcon"},
    "falcon7b": {"size": 7, "dataset": "", "family": "falcon"},
    "gpt2-xl": {"size": 1.5, "dataset": "", "family": "gpt2"},
    "llama-65b": {"size": 65, "dataset": "", "family": "llama"},
    "llama-7b": {"size": 7, "dataset": "", "family": "llama"},
    "neo-1.3": {"size": 1.3, "dataset": "", "family": "neo"},
    "neo-2.7": {"size": 2.7, "dataset": "", "family": "neo"},
    "neo-6": {"size": 6, "dataset": "", "family": "neo"},
    "open_llama_3b_v2": {"size": 3, "dataset": "", "family": "open_llama"},
    "open_llama_7b_v2": {"size": 7, "dataset": "", "family": "open_llama"},
    "opt-1.3b": {"size": 1.3, "dataset": "", "family": "opt"},
    "opt-6.7b": {"size": 6.7, "dataset": "", "family": "opt"},
    "pythia-1.4": {"size": 1.4, "dataset": "", "family": "pythia"},
    "pythia-1.4-dedup": {"size": 1.4, "dataset": "", "family": "pythia"},
    "pythia-6.9": {"size": 6.9, "dataset": "", "family": "pythia"},
    "pythia-6.9-dedup": {"size": 6.9, "dataset": "", "family": "pythia"}
}
Note: Since the dataset information is not provided, it is left empty in the JSON metadata. Also, the size is assumed to be in billion parameters, unless specified otherwise (e.g., gpt2-xl is 1.5 billion parameters). If you need to generate this JSON programmatically, you can use the following Python code: 
 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
base = [
    'pythia',
    'llama',
    'Mistral',
    'gpt2',
    'opt',
    'RedPajama',
    'neo',
    'open_llama',
    'falcon'
]

models = [
    "Mistral-7B-v0.1",
    "RedPajama-INCITE-7B-Base",
    "RedPajama-INCITE-Base-3B-v1",
    "falcon40b",
    "falcon7b",
    "gpt2-xl",
    "llama-65b",
    "llama-7b",
    "neo-1.3",
    "neo-2.7",
    "neo-6",
    "open_llama_3b_v2",
    "open_llama_7b_v2",
    "opt-1.3b",
    "opt-6.7b",
    "pythia-1.4",
    "pythia-1.4-dedup",
    "pythia-6.9",
    "pythia-6.9-dedup"
]

metadata = {}
for model in models:
    size = None
    family = None
    for b in base:
        if b in model:
            family = b
            break
    if 'xl' in model:
        size = 1.5
    elif 'b' in model:
        size = float(model.split('b')[-2])
    elif 'B' in model:
        size = float(model.split('B')[-2])
    metadata[model] = {"size": size, "dataset": "", "family": family}

import json
print(json.dumps(metadata, indent=4))

Extract Json

I extracted the following JSON from that output: 

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
{
    "Mistral-7B-v0.1": {"size": 7, "dataset": "", "family": "Mistral"},
    "RedPajama-INCITE-7B-Base": {"size": 7, "dataset": "", "family": "RedPajama"},
    "RedPajama-INCITE-Base-3B-v1": {"size": 3, "dataset": "", "family": "RedPajama"},
    "falcon40b": {"size": 40, "dataset": "", "family": "falcon"},
    "falcon7b": {"size": 7, "dataset": "", "family": "falcon"},
    "gpt2-xl": {"size": 1.5, "dataset": "", "family": "gpt2"},
    "llama-65b": {"size": 65, "dataset": "", "family": "llama"},
    "llama-7b": {"size": 7, "dataset": "", "family": "llama"},
    "neo-1.3": {"size": 1.3, "dataset": "", "family": "neo"},
    "neo-2.7": {"size": 2.7, "dataset": "", "family": "neo"},
    "neo-6": {"size": 6, "dataset": "", "family": "neo"},
    "open_llama_3b_v2": {"size": 3, "dataset": "", "family": "open_llama"},
    "open_llama_7b_v2": {"size": 7, "dataset": "", "family": "open_llama"},
    "opt-1.3b": {"size": 1.3, "dataset": "", "family": "opt"},
    "opt-6.7b": {"size": 6.7, "dataset": "", "family": "opt"},
    "pythia-1.4": {"size": 1.4, "dataset": "", "family": "pythia"},
    "pythia-1.4-dedup": {"size": 1.4, "dataset": "", "family": "pythia"},
    "pythia-6.9": {"size": 6.9, "dataset": "", "family": "pythia"},
    "pythia-6.9-dedup": {"size": 6.9, "dataset": "", "family": "pythia"}
}

JSON Subset Evaluator

Testing if the previous output matches the JSON: { "Mistral-7B-v0.1": { "size": 7, "dataset": "", "family": "Mistral" }, "RedPajama-INCITE-7B-Base": { "size": 7, "dataset": "", "family": "RedPajama" }, "RedPajama-INCITE-Base-3B-v1": { "size": 3, "dataset": "", "family": "RedPajama" }, "falcon40b": { "size": 40, "dataset": "", "family": "falcon" }, "falcon7b": { "size": 7, "dataset": "", "family": "falcon" }, "gpt2-xl": { "size": 1.5, "dataset": "", "family": "gpt2" }, "llama-65b": { "size": 65, "dataset": "", "family": "llama" }, "llama-7b": { "size": 7, "dataset": "", "family": "llama" }, "neo-1.3": { "size": 1.3, "dataset": "", "family": "neo" }, "neo-2.7": { "size": 2.7, "dataset": "", "family": "neo" }, "neo-6": { "size": 6, "dataset": "", "family": "neo" }, "open_llama_3b_v2": { "size": 3, "dataset": "", "family": "open_llama" }, "open_llama_7b_v2": { "size": 7, "dataset": "", "family": "open_llama" }, "opt-1.3b": { "size": 1.3, "dataset": "", "family": "opt" }, "opt-6.7b": { "size": 6.7, "dataset": "", "family": "opt" }, "pythia-1.4": { "size": 1.4, "dataset": "", "family": "pythia" }, "pythia-1.4-dedup": { "size": 1.4, "dataset": "", "family": "pythia" }, "pythia-6.9": { "size": 6.9, "dataset": "", "family": "pythia" }, "pythia-6.9-dedup": { "size": 6.9, "dataset": "", "family": "pythia" } }: True