mgoin commited on
Commit
9a5732d
1 Parent(s): ae92e50

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +105 -0
README.md ADDED
@@ -0,0 +1,105 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: other
3
+ license_name: nvidia-open-model-license
4
+ license_link: >-
5
+ https://developer.download.nvidia.com/licenses/nvidia-open-model-license-agreement-june-2024.pdf
6
+ inference: false
7
+ fine-tuning: false
8
+ tags:
9
+ - vllm
10
+ base_model: nvidia/Nemotron-4-340B-Instruct
11
+ ---
12
+
13
+ ## Nemotron-4-340B-Instruct-hf
14
+
15
+ Converted checkpoint of [nvidia/Nemotron-4-340B-Instruct](https://huggingface.co/nvidia/Nemotron-4-340B-Instruct). Specifically it was produced from the [v1.0 .nemo checkpoint on NGC](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/nemo/models/nemotron-4-340b-instruct/files).
16
+
17
+ You can deploy this model with `vllm>=0.5.4` ([PR#6611](https://github.com/vllm-project/vllm/pull/6611)):
18
+ ```
19
+ vllm serve mgoin/Nemotron-4-340B-Instruct-hf --tensor-parallel-size 16
20
+ ```
21
+
22
+ ### Evaluations
23
+
24
+ All the below evaluations were run with the [FP8 checkpoint](https://huggingface.co/mgoin/Nemotron-4-340B-Instruct-hf-FP8) using `lm-eval==0.4.3` on 8xA100 GPUs.
25
+
26
+ ```
27
+ lm_eval --model vllm --model_args pretrained=/home/mgoin/code/Nemotron-4-340B-Instruct-hf-FP8,tensor_parallel_size=8,distributed_executor_backend="ray",gpu_memory_utilization=0.6,enforce_eager=True --tasks mmlu --num_fewshot 0 --batch_size 4
28
+ vllm (pretrained=/home/mgoin/code/Nemotron-4-340B-Instruct-hf-FP8,tensor_parallel_size=8,distributed_executor_backend=ray,gpu_memory_utilization=0.6,enforce_eager=True), gen_kwargs: (None), limit: None, num_fewshot: 0, batch_size: 4
29
+ | Tasks |Version|Filter|n-shot|Metric| |Value | |Stderr|
30
+ |---------------------------------------|-------|------|-----:|------|---|-----:|---|-----:|
31
+ |mmlu |N/A |none | 0|acc |↑ |0.7905|± |0.0033|
32
+ | - abstract_algebra | 0|none | 0|acc |↑ |0.5100|± |0.0502|
33
+ | - anatomy | 0|none | 0|acc |↑ |0.8074|± |0.0341|
34
+ | - astronomy | 0|none | 0|acc |↑ |0.8816|± |0.0263|
35
+ | - business_ethics | 0|none | 0|acc |↑ |0.7300|± |0.0446|
36
+ | - clinical_knowledge | 0|none | 0|acc |↑ |0.8453|± |0.0223|
37
+ | - college_biology | 0|none | 0|acc |↑ |0.9236|± |0.0222|
38
+ | - college_chemistry | 0|none | 0|acc |↑ |0.5300|± |0.0502|
39
+ | - college_computer_science | 0|none | 0|acc |↑ |0.7400|± |0.0441|
40
+ | - college_mathematics | 0|none | 0|acc |↑ |0.4500|± |0.0500|
41
+ | - college_medicine | 0|none | 0|acc |↑ |0.7630|± |0.0324|
42
+ | - college_physics | 0|none | 0|acc |↑ |0.5490|± |0.0495|
43
+ | - computer_security | 0|none | 0|acc |↑ |0.8300|± |0.0378|
44
+ | - conceptual_physics | 0|none | 0|acc |↑ |0.8043|± |0.0259|
45
+ | - econometrics | 0|none | 0|acc |↑ |0.7105|± |0.0427|
46
+ | - electrical_engineering | 0|none | 0|acc |↑ |0.7448|± |0.0363|
47
+ | - elementary_mathematics | 0|none | 0|acc |↑ |0.6667|± |0.0243|
48
+ | - formal_logic | 0|none | 0|acc |↑ |0.5952|± |0.0439|
49
+ | - global_facts | 0|none | 0|acc |↑ |0.4900|± |0.0502|
50
+ | - high_school_biology | 0|none | 0|acc |↑ |0.9097|± |0.0163|
51
+ | - high_school_chemistry | 0|none | 0|acc |↑ |0.7143|± |0.0318|
52
+ | - high_school_computer_science | 0|none | 0|acc |↑ |0.9100|± |0.0288|
53
+ | - high_school_european_history | 0|none | 0|acc |↑ |0.8788|± |0.0255|
54
+ | - high_school_geography | 0|none | 0|acc |↑ |0.9242|± |0.0189|
55
+ | - high_school_government_and_politics| 0|none | 0|acc |↑ |0.9845|± |0.0089|
56
+ | - high_school_macroeconomics | 0|none | 0|acc |↑ |0.8333|± |0.0189|
57
+ | - high_school_mathematics | 0|none | 0|acc |↑ |0.4630|± |0.0304|
58
+ | - high_school_microeconomics | 0|none | 0|acc |↑ |0.8824|± |0.0209|
59
+ | - high_school_physics | 0|none | 0|acc |↑ |0.6159|± |0.0397|
60
+ | - high_school_psychology | 0|none | 0|acc |↑ |0.9394|± |0.0102|
61
+ | - high_school_statistics | 0|none | 0|acc |↑ |0.7639|± |0.0290|
62
+ | - high_school_us_history | 0|none | 0|acc |↑ |0.9412|± |0.0165|
63
+ | - high_school_world_history | 0|none | 0|acc |↑ |0.9409|± |0.0153|
64
+ | - human_aging | 0|none | 0|acc |↑ |0.8072|± |0.0265|
65
+ | - human_sexuality | 0|none | 0|acc |↑ |0.8855|± |0.0279|
66
+ | - humanities |N/A |none | 0|acc |↑ |0.7594|± |0.0060|
67
+ | - international_law | 0|none | 0|acc |↑ |0.9091|± |0.0262|
68
+ | - jurisprudence | 0|none | 0|acc |↑ |0.8704|± |0.0325|
69
+ | - logical_fallacies | 0|none | 0|acc |↑ |0.8528|± |0.0278|
70
+ | - machine_learning | 0|none | 0|acc |↑ |0.6786|± |0.0443|
71
+ | - management | 0|none | 0|acc |↑ |0.8641|± |0.0339|
72
+ | - marketing | 0|none | 0|acc |↑ |0.9359|± |0.0160|
73
+ | - medical_genetics | 0|none | 0|acc |↑ |0.8400|± |0.0368|
74
+ | - miscellaneous | 0|none | 0|acc |↑ |0.9221|± |0.0096|
75
+ | - moral_disputes | 0|none | 0|acc |↑ |0.8382|± |0.0198|
76
+ | - moral_scenarios | 0|none | 0|acc |↑ |0.6168|± |0.0163|
77
+ | - nutrition | 0|none | 0|acc |↑ |0.8791|± |0.0187|
78
+ | - other |N/A |none | 0|acc |↑ |0.8214|± |0.0065|
79
+ | - philosophy | 0|none | 0|acc |↑ |0.8521|± |0.0202|
80
+ | - prehistory | 0|none | 0|acc |↑ |0.8796|± |0.0181|
81
+ | - professional_accounting | 0|none | 0|acc |↑ |0.6383|± |0.0287|
82
+ | - professional_law | 0|none | 0|acc |↑ |0.6838|± |0.0119|
83
+ | - professional_medicine | 0|none | 0|acc |↑ |0.8824|± |0.0196|
84
+ | - professional_psychology | 0|none | 0|acc |↑ |0.8611|± |0.0140|
85
+ | - public_relations | 0|none | 0|acc |↑ |0.8000|± |0.0383|
86
+ | - security_studies | 0|none | 0|acc |↑ |0.8204|± |0.0246|
87
+ | - social_sciences |N/A |none | 0|acc |↑ |0.8811|± |0.0057|
88
+ | - sociology | 0|none | 0|acc |↑ |0.9055|± |0.0207|
89
+ | - stem |N/A |none | 0|acc |↑ |0.7180|± |0.0076|
90
+ | - us_foreign_policy | 0|none | 0|acc |↑ |0.9600|± |0.0197|
91
+ | - virology | 0|none | 0|acc |↑ |0.5482|± |0.0387|
92
+ | - world_religions | 0|none | 0|acc |↑ |0.9006|± |0.0229|
93
+
94
+ | Groups |Version|Filter|n-shot|Metric| |Value | |Stderr|
95
+ |------------------|-------|------|-----:|------|---|-----:|---|-----:|
96
+ |mmlu |N/A |none | 0|acc |↑ |0.7905|± |0.0033|
97
+ | - humanities |N/A |none | 0|acc |↑ |0.7594|± |0.0060|
98
+ | - other |N/A |none | 0|acc |↑ |0.8214|± |0.0065|
99
+ | - social_sciences|N/A |none | 0|acc |↑ |0.8811|± |0.0057|
100
+ | - stem |N/A |none | 0|acc |↑ |0.7180|± |0.0076|
101
+ ```
102
+
103
+ The [original paper](https://d1qx31qr3h6wln.cloudfront.net/publications/Nemotron_4_340B_8T_0.pdf) evals for reference:
104
+
105
+ ![image/png](https://cdn-uploads.huggingface.co/production/uploads/60466e4b4f40b01b66151416/R98q-Uh-3n-crwGcSLkuS.png)