fix(l4): fix fp8 logic on l4 (huggingface#2277)

* fix(l4): fix fp8 logic on l4

* also quant weights with single scale

* use marlin even on 89

server/text_generation_server/layers/fp8.py

@@ -32,8 +32,8 @@ def get_fp8_linear() -> torch.nn.Module:
     """
 
     if SYSTEM == "cuda":
-        major, minor = torch.cuda.get_device_capability()
-        if major == 8 and minor < 9:
+        major, _ = torch.cuda.get_device_capability()
+        if major == 8:
             from text_generation_server.layers.marlin import GPTQMarlinFP8Linear
 
             return GPTQMarlinFP8Linear
@@ -42,8 +42,10 @@ def get_fp8_linear() -> torch.nn.Module:
     return Fp8Linear
 
 
-def fp8_quantize(weight, scale_upper_bound=None, qdtype=torch.float8_e4m3fn):
-    if FBGEMM_DYN_AVAILABLE:
+def fp8_quantize(
+    weight, scale_upper_bound=None, qdtype=torch.float8_e4m3fn, scalar=False
+):
+    if FBGEMM_DYN_AVAILABLE and not scalar:
         qweight, scale = torch.ops.fbgemm.quantize_fp8_per_row(
             weight, bs=None, scale_ub=scale_upper_bound, output_dtype=qdtype
         )
@@ -186,6 +188,9 @@ def __init__(
         dtype,
     ) -> None:
         super().__init__()
+        if FBGEMM_MM_AVAILABLE:
+            log_once(logger.info, "Using FBGEMM fp8 optimized kernels")
+
         self.dtype = dtype
         self.qweight = qweight
         self.scale = scale
@@ -201,7 +206,7 @@ def __init__(
 
     @classmethod
     def from_unquant(cls, weight, bias, dtype):
-        qweight, scale = fp8_quantize(weight)
+        qweight, scale = fp8_quantize(weight, scalar=not FBGEMM_MM_AVAILABLE)
         return cls(
             qweight=qweight, scale=scale, scale_upper_bound=None, bias=bias, dtype=dtype
         )
@@ -232,7 +237,7 @@ def forward(self, input: torch.Tensor) -> torch.Tensor:
             )
             return y.to(self.dtype)
 
-        qinput, scale = fp8_quantize(input)
+        qinput, scale = fp8_quantize(input, scalar=True)
         output, _ = torch._scaled_mm(
             qinput,
             self.qweight.t(),