launcher.md 16.6 KB
Newer Older
Merve Noyan's avatar
Merve Noyan committed
1
# Text-generation-launcher arguments
Mishig's avatar
Mishig committed
2

3
4
<!-- WRAP CODE BLOCKS -->

5
```shell
Merve Noyan's avatar
Merve Noyan committed
6
7
8
9
10
Text Generation Launcher

Usage: text-generation-launcher [OPTIONS]

Options:
11
12
13
```
## MODEL_ID
```shell
Merve Noyan's avatar
Merve Noyan committed
14
15
16
17
18
19
      --model-id <MODEL_ID>
          The name of the model to load. Can be a MODEL_ID as listed on <https://hf.co/models> like `gpt2` or `OpenAssistant/oasst-sft-1-pythia-12b`. Or it can be a local directory containing the necessary files as saved by `save_pretrained(...)` methods of transformers
          
          [env: MODEL_ID=]
          [default: bigscience/bloom-560m]

20
21
22
```
## REVISION
```shell
Merve Noyan's avatar
Merve Noyan committed
23
24
25
26
27
      --revision <REVISION>
          The actual revision of the model if you're referring to a model on the hub. You can use a specific commit id or a branch like `refs/pr/2`
          
          [env: REVISION=]

28
29
30
```
## VALIDATION_WORKERS
```shell
Merve Noyan's avatar
Merve Noyan committed
31
32
33
34
35
36
      --validation-workers <VALIDATION_WORKERS>
          The number of tokenizer workers used for payload validation and truncation inside the router
          
          [env: VALIDATION_WORKERS=]
          [default: 2]

37
38
39
```
## SHARDED
```shell
Merve Noyan's avatar
Merve Noyan committed
40
41
42
43
44
45
      --sharded <SHARDED>
          Whether to shard the model across multiple GPUs By default text-generation-inference will use all available GPUs to run the model. Setting it to `false` deactivates `num_shard`
          
          [env: SHARDED=]
          [possible values: true, false]

46
47
48
```
## NUM_SHARD
```shell
Merve Noyan's avatar
Merve Noyan committed
49
50
51
52
53
      --num-shard <NUM_SHARD>
          The number of shards to use if you don't want to use all GPUs on a given machine. You can use `CUDA_VISIBLE_DEVICES=0,1 text-generation-launcher... --num_shard 2` and `CUDA_VISIBLE_DEVICES=2,3 text-generation-launcher... --num_shard 2` to launch 2 copies with 2 shard each on a given machine with 4 GPUs for instance
          
          [env: NUM_SHARD=]

54
55
56
```
## QUANTIZE
```shell
Merve Noyan's avatar
Merve Noyan committed
57
      --quantize <QUANTIZE>
58
          Whether you want the model to be quantized
Merve Noyan's avatar
Merve Noyan committed
59
60
          
          [env: QUANTIZE=]
61
62

          Possible values:
63
64
          - awq:              4 bit quantization. Requires a specific AWQ quantized model: <https://hf.co/models?search=awq>. Should replace GPTQ models wherever possible because of the better latency
          - eetq:             8 bit quantization, doesn't require specific model. Should be a drop-in replacement to bitsandbytes with much better performance. Kernels are from <https://github.com/NetEase-FuXi/EETQ.git>
65
          - exl2:             Variable bit quantization. Requires a specific EXL2 quantized model: <https://hf.co/models?search=exl2>. Requires exllama2 kernels and does not support tensor parallelism (num_shard > 1)
66
          - gptq:             4 bit quantization. Requires a specific GTPQ quantized model: <https://hf.co/models?search=gptq>. text-generation-inference will use exllama (faster) kernels wherever possible, and use triton kernel (wider support) when it's not. AWQ has faster kernels
67
          - marlin:           4 bit quantization. Requires a specific Marlin quantized model: <https://hf.co/models?search=marlin>
68
69
70
          - bitsandbytes:     Bitsandbytes 8bit. Can be applied on any model, will cut the memory requirement in half, but it is known that the model will be much slower to run than the native f16
          - bitsandbytes-nf4: Bitsandbytes 4bit. Can be applied on any model, will cut the memory requirement by 4x, but it is known that the model will be much slower to run than the native f16
          - bitsandbytes-fp4: Bitsandbytes 4bit. nf4 should be preferred in most cases but maybe this one has better perplexity performance for you model
Nicolas Patry's avatar
Nicolas Patry committed
71
          - fp8:              [FP8](https://developer.nvidia.com/blog/nvidia-arm-and-intel-publish-fp8-specification-for-standardization-as-an-interchange-format-for-ai/) (e4m3) works on H100 and above This dtype has native ops should be the fastest if available. This is currently not the fastest because of local unpacking + padding to satisfy matrix multiplication limitations
Merve Noyan's avatar
Merve Noyan committed
72

Nicolas Patry's avatar
Nicolas Patry committed
73
74
75
76
77
78
79
80
```
## SPECULATE
```shell
      --speculate <SPECULATE>
          The number of input_ids to speculate on If using a medusa model, the heads will be picked up automatically Other wise, it will use n-gram speculation which is relatively free in terms of compute, but the speedup heavily depends on the task
          
          [env: SPECULATE=]

81
82
83
```
## DTYPE
```shell
Merve Noyan's avatar
Merve Noyan committed
84
85
86
87
88
89
      --dtype <DTYPE>
          The dtype to be forced upon the model. This option cannot be used with `--quantize`
          
          [env: DTYPE=]
          [possible values: float16, bfloat16]

90
91
92
```
## TRUST_REMOTE_CODE
```shell
Merve Noyan's avatar
Merve Noyan committed
93
94
95
96
97
      --trust-remote-code
          Whether you want to execute hub modelling code. Explicitly passing a `revision` is encouraged when loading a model with custom code to ensure no malicious code has been contributed in a newer revision
          
          [env: TRUST_REMOTE_CODE=]

98
99
100
```
## MAX_CONCURRENT_REQUESTS
```shell
Merve Noyan's avatar
Merve Noyan committed
101
102
103
104
105
106
      --max-concurrent-requests <MAX_CONCURRENT_REQUESTS>
          The maximum amount of concurrent requests for this particular deployment. Having a low limit will refuse clients requests instead of having them wait for too long and is usually good to handle backpressure correctly
          
          [env: MAX_CONCURRENT_REQUESTS=]
          [default: 128]

107
108
109
```
## MAX_BEST_OF
```shell
Merve Noyan's avatar
Merve Noyan committed
110
111
112
113
114
115
      --max-best-of <MAX_BEST_OF>
          This is the maximum allowed value for clients to set `best_of`. Best of makes `n` generations at the same time, and return the best in terms of overall log probability over the entire generated sequence
          
          [env: MAX_BEST_OF=]
          [default: 2]

116
117
118
```
## MAX_STOP_SEQUENCES
```shell
Merve Noyan's avatar
Merve Noyan committed
119
120
121
122
123
124
      --max-stop-sequences <MAX_STOP_SEQUENCES>
          This is the maximum allowed value for clients to set `stop_sequences`. Stop sequences are used to allow the model to stop on more than just the EOS token, and enable more complex "prompting" where users can preprompt the model in a specific way and define their "own" stop token aligned with their prompt
          
          [env: MAX_STOP_SEQUENCES=]
          [default: 4]

125
126
127
```
## MAX_TOP_N_TOKENS
```shell
Merve Noyan's avatar
Merve Noyan committed
128
      --max-top-n-tokens <MAX_TOP_N_TOKENS>
129
          This is the maximum allowed value for clients to set `top_n_tokens`. `top_n_tokens` is used to return information about the the `n` most likely tokens at each generation step, instead of just the sampled token. This information can be used for downstream tasks like for classification or ranking
Merve Noyan's avatar
Merve Noyan committed
130
131
132
133
          
          [env: MAX_TOP_N_TOKENS=]
          [default: 5]

134
135
136
137
138
139
140
141
```
## MAX_INPUT_TOKENS
```shell
      --max-input-tokens <MAX_INPUT_TOKENS>
          This is the maximum allowed input length (expressed in number of tokens) for users. The larger this value, the longer prompt users can send which can impact the overall memory required to handle the load. Please note that some models have a finite range of sequence they can handle. Default to min(max_position_embeddings - 1, 4095)
          
          [env: MAX_INPUT_TOKENS=]

142
143
144
```
## MAX_INPUT_LENGTH
```shell
Merve Noyan's avatar
Merve Noyan committed
145
      --max-input-length <MAX_INPUT_LENGTH>
146
          Legacy version of [`Args::max_input_tokens`]
Merve Noyan's avatar
Merve Noyan committed
147
148
149
          
          [env: MAX_INPUT_LENGTH=]

150
151
152
```
## MAX_TOTAL_TOKENS
```shell
Merve Noyan's avatar
Merve Noyan committed
153
      --max-total-tokens <MAX_TOTAL_TOKENS>
154
          This is the most important value to set as it defines the "memory budget" of running clients requests. Clients will send input sequences and ask to generate `max_new_tokens` on top. with a value of `1512` users can send either a prompt of `1000` and ask for `512` new tokens, or send a prompt of `1` and ask for `1511` max_new_tokens. The larger this value, the larger amount each request will be in your RAM and the less effective batching can be. Default to min(max_position_embeddings, 4096)
Merve Noyan's avatar
Merve Noyan committed
155
156
157
          
          [env: MAX_TOTAL_TOKENS=]

158
159
160
```
## WAITING_SERVED_RATIO
```shell
Merve Noyan's avatar
Merve Noyan committed
161
162
163
164
165
166
      --waiting-served-ratio <WAITING_SERVED_RATIO>
          This represents the ratio of waiting queries vs running queries where you want to start considering pausing the running queries to include the waiting ones into the same batch. `waiting_served_ratio=1.2` Means when 12 queries are waiting and there's only 10 queries left in the current batch we check if we can fit those 12 waiting queries into the batching strategy, and if yes, then batching happens delaying the 10 running queries by a `prefill` run.
          
          This setting is only applied if there is room in the batch as defined by `max_batch_total_tokens`.
          
          [env: WAITING_SERVED_RATIO=]
167
          [default: 0.3]
Merve Noyan's avatar
Merve Noyan committed
168

169
170
171
```
## MAX_BATCH_PREFILL_TOKENS
```shell
Merve Noyan's avatar
Merve Noyan committed
172
      --max-batch-prefill-tokens <MAX_BATCH_PREFILL_TOKENS>
173
          Limits the number of tokens for the prefill operation. Since this operation take the most memory and is compute bound, it is interesting to limit the number of requests that can be sent. Default to `max_input_tokens + 50` to give a bit of room
Merve Noyan's avatar
Merve Noyan committed
174
175
176
          
          [env: MAX_BATCH_PREFILL_TOKENS=]

177
178
179
```
## MAX_BATCH_TOTAL_TOKENS
```shell
Merve Noyan's avatar
Merve Noyan committed
180
181
182
183
184
185
186
187
188
189
190
191
192
      --max-batch-total-tokens <MAX_BATCH_TOTAL_TOKENS>
          **IMPORTANT** This is one critical control to allow maximum usage of the available hardware.
          
          This represents the total amount of potential tokens within a batch. When using padding (not recommended) this would be equivalent of `batch_size` * `max_total_tokens`.
          
          However in the non-padded (flash attention) version this can be much finer.
          
          For `max_batch_total_tokens=1000`, you could fit `10` queries of `total_tokens=100` or a single query of `1000` tokens.
          
          Overall this number should be the largest possible amount that fits the remaining memory (after the model is loaded). Since the actual memory overhead depends on other parameters like if you're using quantization, flash attention or the model implementation, text-generation-inference cannot infer this number automatically.
          
          [env: MAX_BATCH_TOTAL_TOKENS=]

193
194
195
```
## MAX_WAITING_TOKENS
```shell
Merve Noyan's avatar
Merve Noyan committed
196
197
198
199
200
201
202
203
204
205
206
207
      --max-waiting-tokens <MAX_WAITING_TOKENS>
          This setting defines how many tokens can be passed before forcing the waiting queries to be put on the batch (if the size of the batch allows for it). New queries require 1 `prefill` forward, which is different from `decode` and therefore you need to pause the running batch in order to run `prefill` to create the correct values for the waiting queries to be able to join the batch.
          
          With a value too small, queries will always "steal" the compute to run `prefill` and running queries will be delayed by a lot.
          
          With a value too big, waiting queries could wait for a very long time before being allowed a slot in the running batch. If your server is busy that means that requests that could run in ~2s on an empty server could end up running in ~20s because the query had to wait for 18s.
          
          This number is expressed in number of tokens to make it a bit more "model" agnostic, but what should really matter is the overall latency for end users.
          
          [env: MAX_WAITING_TOKENS=]
          [default: 20]

208
209
210
211
212
213
214
215
```
## MAX_BATCH_SIZE
```shell
      --max-batch-size <MAX_BATCH_SIZE>
          Enforce a maximum number of requests per batch Specific flag for hardware targets that do not support unpadded inference
          
          [env: MAX_BATCH_SIZE=]

216
```
217
## CUDA_GRAPHS
218
```shell
219
      --cuda-graphs <CUDA_GRAPHS>
220
          Specify the batch sizes to compute cuda graphs for. Use "0" to disable. Default = "1,2,4,8,16,32"
221
          
222
          [env: CUDA_GRAPHS=]
223

224
225
226
```
## HOSTNAME
```shell
Merve Noyan's avatar
Merve Noyan committed
227
228
229
230
231
232
      --hostname <HOSTNAME>
          The IP address to listen on
          
          [env: HOSTNAME=]
          [default: 0.0.0.0]

233
234
235
```
## PORT
```shell
Merve Noyan's avatar
Merve Noyan committed
236
237
238
239
240
241
  -p, --port <PORT>
          The port to listen on
          
          [env: PORT=]
          [default: 3000]

242
243
244
```
## SHARD_UDS_PATH
```shell
Merve Noyan's avatar
Merve Noyan committed
245
246
247
248
249
250
      --shard-uds-path <SHARD_UDS_PATH>
          The name of the socket for gRPC communication between the webserver and the shards
          
          [env: SHARD_UDS_PATH=]
          [default: /tmp/text-generation-server]

251
252
253
```
## MASTER_ADDR
```shell
Merve Noyan's avatar
Merve Noyan committed
254
255
256
257
258
259
      --master-addr <MASTER_ADDR>
          The address the master shard will listen on. (setting used by torch distributed)
          
          [env: MASTER_ADDR=]
          [default: localhost]

260
261
262
```
## MASTER_PORT
```shell
Merve Noyan's avatar
Merve Noyan committed
263
264
265
266
267
268
      --master-port <MASTER_PORT>
          The address the master port will listen on. (setting used by torch distributed)
          
          [env: MASTER_PORT=]
          [default: 29500]

269
270
271
```
## HUGGINGFACE_HUB_CACHE
```shell
Merve Noyan's avatar
Merve Noyan committed
272
273
274
275
276
      --huggingface-hub-cache <HUGGINGFACE_HUB_CACHE>
          The location of the huggingface hub cache. Used to override the location if you want to provide a mounted disk for instance
          
          [env: HUGGINGFACE_HUB_CACHE=]

277
278
279
```
## WEIGHTS_CACHE_OVERRIDE
```shell
Merve Noyan's avatar
Merve Noyan committed
280
281
282
283
284
      --weights-cache-override <WEIGHTS_CACHE_OVERRIDE>
          The location of the huggingface hub cache. Used to override the location if you want to provide a mounted disk for instance
          
          [env: WEIGHTS_CACHE_OVERRIDE=]

285
286
287
```
## DISABLE_CUSTOM_KERNELS
```shell
Merve Noyan's avatar
Merve Noyan committed
288
289
290
291
292
      --disable-custom-kernels
          For some models (like bloom), text-generation-inference implemented custom cuda kernels to speed up inference. Those kernels were only tested on A100. Use this flag to disable them if you're running on different hardware and encounter issues
          
          [env: DISABLE_CUSTOM_KERNELS=]

293
294
295
```
## CUDA_MEMORY_FRACTION
```shell
Merve Noyan's avatar
Merve Noyan committed
296
297
298
299
300
301
      --cuda-memory-fraction <CUDA_MEMORY_FRACTION>
          Limit the CUDA available memory. The allowed value equals the total visible memory multiplied by cuda-memory-fraction
          
          [env: CUDA_MEMORY_FRACTION=]
          [default: 1.0]

302
303
304
```
## ROPE_SCALING
```shell
Merve Noyan's avatar
Merve Noyan committed
305
306
307
308
309
310
311
312
313
314
315
316
      --rope-scaling <ROPE_SCALING>
          Rope scaling will only be used for RoPE models and allow rescaling the position rotary to accomodate for larger prompts.
          
          Goes together with `rope_factor`.
          
          `--rope-factor 2.0` gives linear scaling with a factor of 2.0 `--rope-scaling dynamic` gives dynamic scaling with a factor of 1.0 `--rope-scaling linear` gives linear scaling with a factor of 1.0 (Nothing will be changed basically)
          
          `--rope-scaling linear --rope-factor` fully describes the scaling you want
          
          [env: ROPE_SCALING=]
          [possible values: linear, dynamic]

317
318
319
```
## ROPE_FACTOR
```shell
Merve Noyan's avatar
Merve Noyan committed
320
321
322
323
324
      --rope-factor <ROPE_FACTOR>
          Rope scaling will only be used for RoPE models See `rope_scaling`
          
          [env: ROPE_FACTOR=]

325
326
327
```
## JSON_OUTPUT
```shell
Merve Noyan's avatar
Merve Noyan committed
328
329
330
331
332
      --json-output
          Outputs the logs in JSON format (useful for telemetry)
          
          [env: JSON_OUTPUT=]

333
334
335
```
## OTLP_ENDPOINT
```shell
Merve Noyan's avatar
Merve Noyan committed
336
337
338
      --otlp-endpoint <OTLP_ENDPOINT>
          [env: OTLP_ENDPOINT=]

339
340
341
342
343
344
345
```
## OTLP_SERVICE_NAME
```shell
      --otlp-service-name <OTLP_SERVICE_NAME>
          [env: OTLP_SERVICE_NAME=]
          [default: text-generation-inference.router]

346
347
348
```
## CORS_ALLOW_ORIGIN
```shell
Merve Noyan's avatar
Merve Noyan committed
349
350
351
      --cors-allow-origin <CORS_ALLOW_ORIGIN>
          [env: CORS_ALLOW_ORIGIN=]

352
353
354
```
## WATERMARK_GAMMA
```shell
Merve Noyan's avatar
Merve Noyan committed
355
356
357
      --watermark-gamma <WATERMARK_GAMMA>
          [env: WATERMARK_GAMMA=]

358
359
360
```
## WATERMARK_DELTA
```shell
Merve Noyan's avatar
Merve Noyan committed
361
362
363
      --watermark-delta <WATERMARK_DELTA>
          [env: WATERMARK_DELTA=]

364
365
366
```
## NGROK
```shell
Merve Noyan's avatar
Merve Noyan committed
367
368
369
370
371
      --ngrok
          Enable ngrok tunneling
          
          [env: NGROK=]

372
373
374
```
## NGROK_AUTHTOKEN
```shell
Merve Noyan's avatar
Merve Noyan committed
375
376
377
378
379
      --ngrok-authtoken <NGROK_AUTHTOKEN>
          ngrok authentication token
          
          [env: NGROK_AUTHTOKEN=]

380
381
382
```
## NGROK_EDGE
```shell
Merve Noyan's avatar
Merve Noyan committed
383
384
385
386
387
      --ngrok-edge <NGROK_EDGE>
          ngrok edge
          
          [env: NGROK_EDGE=]

388
389
390
391
392
393
394
395
```
## TOKENIZER_CONFIG_PATH
```shell
      --tokenizer-config-path <TOKENIZER_CONFIG_PATH>
          The path to the tokenizer config file. This path is used to load the tokenizer configuration which may include a `chat_template`. If not provided, the default config will be used from the model hub
          
          [env: TOKENIZER_CONFIG_PATH=]

drbh's avatar
drbh committed
396
397
398
399
400
401
402
403
```
## DISABLE_GRAMMAR_SUPPORT
```shell
      --disable-grammar-support
          Disable outlines grammar constrained generation. This is a feature that allows you to generate text that follows a specific grammar
          
          [env: DISABLE_GRAMMAR_SUPPORT=]

404
405
406
```
## ENV
```shell
Merve Noyan's avatar
Merve Noyan committed
407
408
409
  -e, --env
          Display a lot of information about your runtime environment

410
411
412
413
414
415
416
417
418
```
## MAX_CLIENT_BATCH_SIZE
```shell
      --max-client-batch-size <MAX_CLIENT_BATCH_SIZE>
          Control the maximum number of inputs that a client can send in a single request
          
          [env: MAX_CLIENT_BATCH_SIZE=]
          [default: 4]

419
420
421
```
## HELP
```shell
Merve Noyan's avatar
Merve Noyan committed
422
423
424
  -h, --help
          Print help (see a summary with '-h')

425
426
427
```
## VERSION
```shell
Merve Noyan's avatar
Merve Noyan committed
428
429
430
  -V, --version
          Print version

431
```