Nicholai/jan
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: correct flash_attn and main_gpu flag checks in llamacpp extension

Browse Source 
Previously the condition for `flash_attn` was always truthy, causing
unnecessary or incorrect `--flash-attn` arguments to be added. The
`main_gpu` check also used a loose inequality which could match values
that were not intended. The updated logic uses strict comparison and
correctly handles the empty string case, ensuring the command line
arguments are generated only when appropriate.
This commit is contained in:
Akarshan 2025-10-30 19:49:55 +05:30
parent 1f4977c1d1
commit ea231676bf
No known key found for this signature in database
GPG Key ID: D75C9634A870665F
1 changed files with 8 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
extensions/llamacpp-extension/src/index.ts
View File

@ -1646,10 +1646,15 @@ export default class llamacpp_extension extends AIEngine {
if (cfg.device.length > 0) args.push('--device', cfg.device) if (cfg.device.length > 0) args.push('--device', cfg.device)
if (cfg.split_mode.length > 0 && cfg.split_mode != 'layer') if (cfg.split_mode.length > 0 && cfg.split_mode != 'layer')
args.push('--split-mode', cfg.split_mode) args.push('--split-mode', cfg.split_mode)
if (cfg.main_gpu !== undefined && cfg.main_gpu != 0) if (cfg.main_gpu !== undefined && cfg.main_gpu !== 0)
args.push('--main-gpu', String(cfg.main_gpu)) args.push('--main-gpu', String(cfg.main_gpu))
// Note: Older llama.cpp versions are no longer supported // Note: Older llama.cpp versions are no longer supported
if (cfg.flash_attn !== undefined || !cfg.flash_attn || cfg.flash_attn !== '') args.push('--flash-attn', String(cfg.flash_attn)) //default: auto = ON when supported if (
cfg.flash_attn !== undefined ||
!cfg.flash_attn ||
cfg.flash_attn !== ''
)
args.push('--flash-attn', String(cfg.flash_attn)) //default: auto = ON when supported
// Boolean flags // Boolean flags
if (cfg.ctx_shift) args.push('--context-shift') if (cfg.ctx_shift) args.push('--context-shift')
@ -1666,7 +1671,7 @@ export default class llamacpp_extension extends AIEngine {
if (cfg.cache_type_k && cfg.cache_type_k != 'f16') if (cfg.cache_type_k && cfg.cache_type_k != 'f16')
args.push('--cache-type-k', cfg.cache_type_k) args.push('--cache-type-k', cfg.cache_type_k)
if ( if (
cfg.flash_attn && cfg.flash_attn !== 'on' &&
cfg.cache_type_v != 'f16' && cfg.cache_type_v != 'f16' &&
cfg.cache_type_v != 'f32' cfg.cache_type_v != 'f32'
) { ) {