make sure GPTQ will not turn on triton when triton is not available

chichun-charlie-liu · chichun-charlie-liu · commit ce13e2c5f32a · 2025-03-20T10:15:23.000-04:00
Signed-off-by: cliu-us &lt;cliu@us.ibm.com&gt;
diff --git a/fms_mo/run_quant.py b/fms_mo/run_quant.py
@@ -92,6 +92,7 @@ def quantize(
                 "auto_gptq module not found. For more instructions on installing the appropriate "
                 "package, see https://github.com/AutoGPTQ/AutoGPTQ?tab=readme-ov-file#installation"
             )
+        gptq_args.use_triton = gptq_args.use_triton and available_packages["triton"]
         run_gptq(model_args, data_args, opt_args, gptq_args)
     elif opt_args.quant_method == "fp8":
         if not available_packages["llmcompressor"]:

Original file line number	Diff line number	Diff line change
`@@ -92,6 +92,7 @@ def quantize(`
`92`	`92`	`"auto_gptq module not found. For more instructions on installing the appropriate "`
`93`	`93`	`"package, see https://github.com/AutoGPTQ/AutoGPTQ?tab=readme-ov-file#installation"`
`94`	`94`	`)`
	`95`	`+ gptq_args.use_triton = gptq_args.use_triton and available_packages["triton"]`
`95`	`96`	`run_gptq(model_args, data_args, opt_args, gptq_args)`
`96`	`97`	`elif opt_args.quant_method == "fp8":`
`97`	`98`	`if not available_packages["llmcompressor"]:`