NeMo
File size: 12,871 Bytes
91fa21e
 
 
91adaf3
 
91fa21e
be08e80
 
8d2ac2d
be08e80
 
 
8d2ac2d
 
 
be08e80
 
4e7a5c2
8d2ac2d
4e7a5c2
8d2ac2d
4e7a5c2
 
 
 
71a1b51
 
 
be08e80
4e7a5c2
be08e80
4e7a5c2
 
 
 
 
 
 
 
8d2ac2d
be08e80
 
 
8d2ac2d
be08e80
8d2ac2d
be08e80
 
 
 
8d2ac2d
 
 
be08e80
 
 
 
8d2ac2d
be08e80
 
 
4e7a5c2
 
be08e80
8d2ac2d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
be08e80
 
8d2ac2d
 
 
 
 
 
 
be08e80
593d8f5
4e7a5c2
 
 
be08e80
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4e7a5c2
 
8d2ac2d
ed5d0f6
 
593d8f5
8d2ac2d
be08e80
8d2ac2d
4e7a5c2
be08e80
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4e7a5c2
8d2ac2d
be08e80
 
 
 
4e7a5c2
be08e80
4e7a5c2
be08e80
 
 
 
 
 
 
 
 
 
 
 
 
4e7a5c2
be08e80
 
 
 
 
4e7a5c2
be08e80
 
8d2ac2d
be08e80
8d2ac2d
be08e80
 
 
8d2ac2d
be08e80
 
 
 
 
4e7a5c2
 
 
 
 
 
be08e80
4e7a5c2
be08e80
 
 
4e7a5c2
be08e80
 
 
 
 
8d2ac2d
3872adc
8d2ac2d
 
be08e80
 
 
 
 
 
 
8d2ac2d
be08e80
 
 
 
 
 
 
8d2ac2d
be08e80
 
 
 
4e7a5c2
be08e80
 
8d2ac2d
be08e80
 
 
 
 
 
 
 
8d2ac2d
be08e80
 
4e7a5c2
 
 
 
 
8d2ac2d
4e7a5c2
 
 
be08e80
 
 
 
 
8d2ac2d
be08e80
 
 
 
 
 
4e7a5c2
8d2ac2d
4e7a5c2
be08e80
 
 
 
 
 
 
8d2ac2d
be08e80
 
 
4e7a5c2
 
 
71a1b51
 
 
8d2ac2d
 
 
 
4e7a5c2
 
 
be08e80
8d2ac2d
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
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
---
license: other
license_name: nvidia-open-model-license
license_link: >-
  https://developer.download.nvidia.com/licenses/nvidia-open-model-license-agreement-june-2024.pdf
---
## Nemotron-4-340B-Instruct

[![Model architecture](https://img.shields.io/badge/Model%20Arch-Transformer%20Decoder-green)](#model-architecture)[![Model size](https://img.shields.io/badge/Params-340B-green)](#model-architecture)[![Language](https://img.shields.io/badge/Language-Multilingual-green)](#datasets)

### Model Overview

Nemotron-4-340B-Instruct is a large language model (LLM) that can be used as part of a synthetic data generation pipeline to create training data that helps researchers and developers build their own LLMs. It is a fine-tuned version of the Nemotron-4-340B-Base model, optimized for English-based single and multi-turn chat use-cases. It supports a context length of 4,096 tokens. 

The base model was pre-trained on a corpus of 9 trillion tokens consisting of a diverse assortment of English based texts, 50+ natural languages, and 40+ coding languages. Subsequently the Nemotron-4-340B-Instruct model went through additional alignment steps including:

- Supervised Fine-tuning (SFT)
- Direct Preference Optimization (DPO)
- Reward-aware Preference Optimization (RPO) ([Additional in-house alignment technique](https://research.nvidia.com/publication/2024-06_nemotron-4-340b)) 

Throughout the alignment process, we relied on only approximately 20K human-annotated data while our data generation pipeline synthesized over 98% of the data used for supervised fine-tuning and preference fine-tuning (DPO & RPO). We provide comprehensive details about our synthetic data generation pipeline in the [technical report](https://research.nvidia.com/publication/2024-06_nemotron-4-340b).

This results in a model that is aligned for human chat preferences, improvements in mathematical reasoning, coding and instruction-following, and is capable of generating high quality synthetic data for a variety of use cases.

Under the NVIDIA Open Model License, NVIDIA confirms: 
- Models are commercially usable. 
- You are free to create and distribute Derivative Models. 
- NVIDIA does not claim ownership to any outputs generated using the Models or Derivative Models.

### License: 

[NVIDIA Open Model License](https://developer.download.nvidia.com/licenses/nvidia-open-model-license-agreement-june-2024.pdf)

### Intended use

Nemotron-4-340B-Instruct is a chat model intended for use for the English language. 

Nemotron-4-340B-Instruct is designed for Synthetic Data Generation to enable developers and enterprises for building and customizing their own large language models and LLM applications. 

The instruct model itself can be further customized using the [NeMo Framework](https://docs.nvidia.com/nemo-framework/index.html) suite of customization tools including Parameter-Efficient Fine-Tuning (P-tuning, Adapters, LoRA, and more), and Model Alignment (SFT, SteerLM, RLHF, and more) using [NeMo-Aligner](https://github.com/NVIDIA/NeMo-Aligner).

**Model Developer:** NVIDIA

**Model Dates:** Nemotron-4-340B-Instruct was trained between December 2023 and May 2024.

**Data Freshness:** The pretraining data has a cutoff of June 2023.

### Required Hardware

BF16 Inference:
- 8x H200 (1x H200 node)
- 16x H100 (2x H100 nodes)
- 16x A100 80GB (2x A100 80GB nodes)


### Model Architecture:

Nemotron-4-340B-Instruct is standard decoder-only Transformer, trained with a sequence length of 4096 tokens, uses Grouped-Query Attention (GQA), and Rotary Position Embeddings (RoPE).

**Architecture Type:** Transformer Decoder (auto-regressive language model)

**Network Architecture:**
Nemotron-4

### Prompt Format

Note: For Nemotron-4-340B-Instruct we recommend keeping the system prompt empty.

#### Single Turn

```text
<extra_id_0>System

<extra_id_1>User
{prompt}
<extra_id_1>Assistant
```

#### Multi-Turn or Few-shot

```text
<extra_id_0>System

<extra_id_1>User
{prompt 1}
<extra_id_1>Assistant
{response 1}
<extra_id_1>User
{prompt 2}
<extra_id_1>Assistant
{response 2}
...
<extra_id_1>User
{prompt N}
<extra_id_1>Assistant
```

An example of a formattable prompt template is available in the following section.

### Usage

Deployment and inference with Nemotron-4-340B-Instruct can be done in three steps using NeMo Framework:

Create a Python script to interact with the deployed model.
Create a Bash script to start the inference server
Schedule a Slurm job to distribute the model across 4 nodes and associate them with the inference server.

1. Define the Python script ``call_server.py``

```python
import json
import requests

headers = {"Content-Type": "application/json"}

def text_generation(data, ip='localhost', port=None):
    resp = requests.put(f'http://{ip}:{port}/generate', data=json.dumps(data), headers=headers)
    return resp.json()


def get_generation(prompt, greedy, add_BOS, token_to_gen, min_tokens, temp, top_p, top_k, repetition, batch=False):
    data = {
        "sentences": [prompt] if not batch else prompt,
        "tokens_to_generate": int(token_to_gen),
        "temperature": temp,
        "add_BOS": add_BOS,
        "top_k": top_k,
        "top_p": top_p,
        "greedy": greedy,
        "all_probs": False,
        "repetition_penalty": repetition,
        "min_tokens_to_generate": int(min_tokens),
        "end_strings": ["<|endoftext|>", "<extra_id_1>", "\x11", "<extra_id_1>User"],
    }
    sentences = text_generation(data, port=1424)['sentences']
    return sentences[0] if not batch else sentences

PROMPT_TEMPLATE = """<extra_id_0>System

<extra_id_1>User
{prompt}
<extra_id_1>Assistant
"""

question = "Write a poem on NVIDIA in the style of Shakespeare"
prompt = PROMPT_TEMPLATE.format(prompt=question)
print(prompt)

response = get_generation(prompt, greedy=True, add_BOS=False, token_to_gen=1024, min_tokens=1, temp=1.0, top_p=1.0, top_k=0, repetition=1.0, batch=False)
response = response[len(prompt):]
if response.endswith("<extra_id_1>"):
    response = response[:-len("<extra_id_1>")]
print(response)
```

2. Given this Python script, create a Bash script which spins up the inference server within the [NeMo container](https://catalog.ngc.nvidia.com/orgs/nvidia/containers/nemo) (docker pull nvcr.io/nvidia/nemo:24.01.framework) and calls the Python script ``call_server.py``. The Bash script ``nemo_inference.sh`` is as follows,

```
NEMO_FILE=$1
WEB_PORT=1424

depends_on () {
    HOST=$1
    PORT=$2
    STATUS=$(curl -X PUT http://$HOST:$PORT >/dev/null 2>/dev/null; echo $?)
    while [ $STATUS -ne 0 ]
    do
         echo "waiting for server ($HOST:$PORT) to be up"
         sleep 10
         STATUS=$(curl -X PUT http://$HOST:$PORT >/dev/null 2>/dev/null; echo $?)
    done
    echo "server ($HOST:$PORT) is up running"
}



/usr/bin/python3 /opt/NeMo/examples/nlp/language_modeling/megatron_gpt_eval.py \
        gpt_model_file=$NEMO_FILE \
        pipeline_model_parallel_split_rank=0 \
        server=True tensor_model_parallel_size=8 \
        trainer.precision=bf16 pipeline_model_parallel_size=2 \
        trainer.devices=8 \
        trainer.num_nodes=2 \
        web_server=False \
        port=${WEB_PORT} &
    SERVER_PID=$!

    readonly local_rank="${LOCAL_RANK:=${SLURM_LOCALID:=${OMPI_COMM_WORLD_LOCAL_RANK:-}}}"
    if [ $SLURM_NODEID -eq 0 ] && [ $local_rank -eq 0 ]; then
        depends_on "0.0.0.0" ${WEB_PORT}

        echo "start get json"
        sleep 5

        echo "SLURM_NODEID: $SLURM_NODEID"
        echo "local_rank: $local_rank"
        /usr/bin/python3 /scripts/call_server.py
        echo "clean up dameons: $$"
        kill -9 $SERVER_PID
        pkill python
    fi
    wait
```


3, Launch ``nemo_inference.sh`` with a Slurm script defined like below, which starts a 4-node job for model inference.

```
#!/bin/bash
#SBATCH -A SLURM-ACCOUNT
#SBATCH -p SLURM-PARITION
#SBATCH -N 2 # number of nodes
#SBATCH -J generation      
#SBATCH --ntasks-per-node=8   
#SBATCH --gpus-per-node=8
set -x

RESULTS=<PATH_TO_YOUR_SCRIPTS_FOLDER>
OUTFILE="${RESULTS}/slurm-%j-%n.out"
ERRFILE="${RESULTS}/error-%j-%n.out"
MODEL=<PATH_TO>/Nemotron-4-340B-Instruct

MOUNTS="--container-mounts=<PATH_TO_YOUR_SCRIPTS_FOLDER>:/scripts,MODEL:/model"
read -r -d '' cmd <<EOF
bash /scripts/nemo_inference.sh /model
EOF

srun -o $OUTFILE -e $ERRFILE --container-image="$CONTAINER" $MOUNTS bash -c "${cmd}"
```

### Evaluation Results

#### MT-Bench (GPT-4-Turbo)

Evaluated using MT-Bench judging by GPT-4-0125-Preview as described in Appendix H in the [HelpSteer2 Dataset Paper](https://arxiv.org/abs/2406.08673)

| total | writing | roleplay | extraction | stem | humanities | reasoning | math | coding | turn 1 | turn 2 |
| :----- | :------- | :-------- | :---------- | :---- | :---------- | :--------- | :---- | ------ | :------ | :------ | 
| 8.22 | 8.70 | 8.70  | 9.20 | 8.75 | 8.95 | 6.40 | 8.40 | 6.70 | 8.61 | 7.84 | 

#### IFEval

Evaluated using the Instruction Following Eval (IFEval) introduced in [Instruction-Following Evaluation for Large Language Models](https://arxiv.org/pdf/2311.07911).

| Prompt-Strict Acc | Instruction-Strict Acc |
| :----------------------- | :---------------------------- |
| 79.9 | 86.1 |

#### MMLU

Evaluated using the Multi-task Language Understanding benchmarks as introduced in [Measuring Massive Multitask Language Understanding](https://arxiv.org/pdf/2009.03300)

|MMLU 0-shot |
| :----------------- |
| 78.7  | 

#### GSM8K

Evaluated using the Grade School Math 8K (GSM8K) benchmark as introduced in [Training Verifiers to Solve Math Word Problems](https://arxiv.org/pdf/2110.14168v2).

| GSM8K 0-shot |
| :----------------- | 
| 92.3 | 

#### HumanEval

Evaluated using the HumanEval benchmark as introduced in [Evaluating Large Language Models Trained on Code](https://arxiv.org/pdf/2107.03374).


| HumanEval 0-shot |
| :----- |
| 73.2 |

#### MBPP

Evaluated using the MBPP Dataset as introduced in the [Program Synthesis with Large Language Models](https://arxiv.org/abs/2108.07732) paper.

| MBPP 0-shot|
| :----------------- | 
| 75.4 | 


#### Arena Hard

Evaluated using the [Arena-Hard Pipeline](https://lmsys.org/blog/2024-04-19-arena-hard/) from the LMSys Org.

| Arena Hard |
| :----------------- | 
| 54.2 | 

#### AlpacaEval 2.0 LC

Evaluated using the AlpacaEval 2.0 LC (Length Controlled) as introduced in the paper: [Length-Controlled AlpacaEval: A Simple Way to Debias Automatic Evaluators](https://arxiv.org/abs/2404.04475)

| AlpacaEval 2.0 LC|
| :----------------- | 
| 41.5 | 


#### TFEval

Evaluated using the CantTalkAboutThis Dataset as introduced in the [CantTalkAboutThis: Aligning Language Models to Stay on Topic in Dialogues](https://arxiv.org/abs/2404.03820) paper.

| Distractor F1 | On-topic F1 |
| :----------------------- | :---------------------------- |
| 81.7  | 97.7 |


### Adversarial Testing and Red Teaming Efforts 

The Nemotron-4 340B-Instruct model underwent extensive safety evaluation including adversarial testing via three distinct methods: 
- [Garak](https://docs.garak.ai/garak), is an automated LLM vulnerability scanner that probes for common weaknesses, including prompt injection and data leakage. 
- [AEGIS](https://arxiv.org/pdf/2404.05993), is a content safety evaluation dataset and LLM based content safety classifier model, that adheres to a broad taxonomy of 13 categories of critical risks in human-LLM interactions.
- Human Content Red Teaming leveraging human interaction and evaluation of the models' responses.

### Limitations

The model was trained on data that contains toxic language, unsafe content, and societal biases originally crawled from the internet. Therefore, the model may amplify those biases and return toxic responses especially when prompted with toxic prompts. The model may generate answers that may be inaccurate, omit key information, or include irrelevant or redundant text producing socially unacceptable or undesirable text, even if the prompt itself does not include anything explicitly offensive.


### Ethical Considerations

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications.  When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.  For more detailed information on ethical considerations for this model, please see the Model Card++ Explainability, Bias, Safety & Security, and Privacy Subcards [here](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/nemo/models/nemotron-4-340b-instruct).  Please report security vulnerabilities or NVIDIA AI Concerns [here](https://www.nvidia.com/en-us/support/submit-security-vulnerability/).