add quant support for qbits

CSY-ModelCloud · Jul 5, 2024 · e838217 · e838217
1 parent 61191d5
commit e838217
Show file tree

Hide file tree

Showing 2 changed files with 4 additions and 0 deletions.
diff --git a/gptqmodel/quantization/config.py b/gptqmodel/quantization/config.py
@@ -40,6 +40,7 @@ class FORMAT:
     GPTQ_V2 = "gptq_v2"
     MARLIN = "marlin"
     BITBLAS = "bitblas"
+    QBITS = "qbits"
 
 
 # quant methods
@@ -53,6 +54,7 @@ class QUANT_METHOD:
         FORMAT.GPTQ_V2,
         FORMAT.MARLIN,
         FORMAT.BITBLAS,
+        FORMAT.QBITS,
     },
 }
 

diff --git a/gptqmodel/utils/importer.py b/gptqmodel/utils/importer.py
@@ -16,13 +16,15 @@
     Backend.EXLLAMA: ExllamaQuantLinear,
     Backend.TRITON: TritonV2QuantLinear,
     Backend.BITBLAS: BitBLASQuantLinear,
+    Backend.QBITS: QBitsQuantLinear,
 })
 
 format_dict = {
     FORMAT.GPTQ: [Backend.EXLLAMA_V2, Backend.EXLLAMA],
     FORMAT.GPTQ_V2: [Backend.EXLLAMA_V2, Backend.EXLLAMA],
     FORMAT.MARLIN: [Backend.MARLIN],
     FORMAT.BITBLAS: [Backend.BITBLAS],
+    FORMAT.QBITS: [Backend.QBITS],
 }
 
 logger = getLogger(__name__)