root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-11-13 14:29:52 +00:00
Code Issues Actions 11 Packages Projects Releases Wiki Activity

readme : remove --memory-f32 references (#9925)

Browse Source
This commit is contained in:
Georgi Gerganov 2024-10-17 23:43:05 +03:00
parent 9f45fc1e99
commit 17bb928080
No known key found for this signature in database
GPG Key ID: 449E073F9DC10735
2 changed files with 3 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
examples/main/README.md
View File

@ -297,10 +297,6 @@ These options help improve the performance and memory usage of the LLaMA models.
These flags attempt optimizations that help on some systems with non-uniform memory access. This currently consists of one of the above strategies, and disabling prefetch and readahead for mmap. The latter causes mapped pages to be faulted in on first access instead of all at once, and in combination with pinning threads to NUMA nodes, more of the pages end up on the NUMA node where they are used. Note that if the model is already in the system page cache, for example because of a previous run without this option, this will have little effect unless you drop the page cache first. This can be done by rebooting the system or on Linux by writing '3' to '/proc/sys/vm/drop_caches' as root.
### Memory Float 32
- `--memory-f32`: Use 32-bit floats instead of 16-bit floats for memory key+value. This doubles the context memory requirement and cached prompt file size but does not appear to increase generation quality in a measurable way. Not recommended.
### Batch Size
- `-b N, --batch-size N`: Set the batch size for prompt processing (default: `2048`). This large batch size benefits users who have BLAS installed and enabled it during the build. If you don't have BLAS enabled ("BLAS=0"), you can use a smaller number, such as 8, to see the prompt progress as it's evaluated in some situations.

6
scripts/run-with-preset.py
View File

@ -15,7 +15,7 @@ CLI_ARGS_LLAMA_CLI_PERPLEXITY = [
"export", "file", "frequency-penalty", "grammar", "grammar-file", "hellaswag",
"hellaswag-tasks", "ignore-eos", "in-prefix", "in-prefix-bos", "in-suffix",
"interactive", "interactive-first", "keep", "logdir", "logit-bias", "lora", "lora-base",
"low-vram", "main-gpu", "memory-f32", "mirostat", "mirostat-ent", "mirostat-lr", "mlock",
"low-vram", "main-gpu", "mirostat", "mirostat-ent", "mirostat-lr", "mlock",
"model", "multiline-input", "n-gpu-layers", "n-predict", "no-mmap", "no-mul-mat-q",
"np-penalize-nl", "numa", "ppl-output-type", "ppl-stride", "presence-penalty", "prompt",
"prompt-cache", "prompt-cache-all", "prompt-cache-ro", "repeat-last-n",
@ -25,12 +25,12 @@ CLI_ARGS_LLAMA_CLI_PERPLEXITY = [
]
CLI_ARGS_LLAMA_BENCH = [
"batch-size", "memory-f32", "low-vram", "model", "mul-mat-q", "n-gen", "n-gpu-layers",
"batch-size", "low-vram", "model", "mul-mat-q", "n-gen", "n-gpu-layers",
"n-prompt", "output", "repetitions", "tensor-split", "threads", "verbose"
]
CLI_ARGS_LLAMA_SERVER = [
"alias", "batch-size", "ctx-size", "embedding", "host", "memory-f32", "lora", "lora-base",
"alias", "batch-size", "ctx-size", "embedding", "host", "lora", "lora-base",
"low-vram", "main-gpu", "mlock", "model", "n-gpu-layers", "n-probs", "no-mmap", "no-mul-mat-q",
"numa", "path", "port", "rope-freq-base", "timeout", "rope-freq-scale", "tensor-split",
"threads", "verbose"