Add support for exl2-quantized models

docs/source/basic_tutorials/launcher.md

@@ -62,6 +62,7 @@ Options:
           Possible values:
           - awq:              4 bit quantization. Requires a specific AWQ quantized model: <https://hf.co/models?search=awq>. Should replace GPTQ models wherever possible because of the better latency
           - eetq:             8 bit quantization, doesn't require specific model. Should be a drop-in replacement to bitsandbytes with much better performance. Kernels are from <https://github.com/NetEase-FuXi/EETQ.git>
+          - exl2:             Variable bit quantization. Requires a specific EXL2 quantized model: <https://hf.co/models?search=exl2>. Requires exllama2 kernels and does not support tensor parallelism (num_shard > 1)
           - gptq:             4 bit quantization. Requires a specific GTPQ quantized model: <https://hf.co/models?search=gptq>. text-generation-inference will use exllama (faster) kernels wherever possible, and use triton kernel (wider support) when it's not. AWQ has faster kernels
           - bitsandbytes:     Bitsandbytes 8bit. Can be applied on any model, will cut the memory requirement in half, but it is known that the model will be much slower to run than the native f16
           - bitsandbytes-nf4: Bitsandbytes 4bit. Can be applied on any model, will cut the memory requirement by 4x, but it is known that the model will be much slower to run than the native f16

docs/source/conceptual/guidance.md

@@ -2,7 +2,6 @@
 
 ## What is Guidance?
 
-
 Guidance is a feature that allows users to constrain the generation of a large language model with a specified grammar. This feature is particularly useful when you want to generate text that follows a specific structure or uses a specific set of words or produce output in a specific format. A prominent example is JSON grammar, where the model is forced to output valid JSON.
 
 ## How is it used?

integration-tests/conftest.py

@@ -38,6 +38,7 @@
 
 class ResponseComparator(JSONSnapshotExtension):
     rtol = 0.2
+    ignore_logprob = False
 
     def serialize(
         self,
@@ -95,7 +96,10 @@ def eq_token(token: Token, other: Token) -> bool:
             return (
                 token.id == other.id
                 and token.text == other.text
-                and math.isclose(token.logprob, other.logprob, rel_tol=self.rtol)
+                and (
+                    self.ignore_logprob
+                    or math.isclose(token.logprob, other.logprob, rel_tol=self.rtol)
+                )
                 and token.special == other.special
             )
 
@@ -105,8 +109,11 @@ def eq_prefill_token(prefill_token: InputToken, other: InputToken) -> bool:
                     prefill_token.id == other.id
                     and prefill_token.text == other.text
                     and (
-                        math.isclose(
-                            prefill_token.logprob, other.logprob, rel_tol=self.rtol
+                        self.ignore_logprob
+                        or math.isclose(
+                            prefill_token.logprob,
+                            other.logprob,
+                            rel_tol=self.rtol,
                         )
                         if prefill_token.logprob is not None
                         else prefill_token.logprob == other.logprob
@@ -223,6 +230,10 @@ class GenerousResponseComparator(ResponseComparator):
     rtol = 0.75
 
 
+class IgnoreLogProbResponseComparator(ResponseComparator):
+    ignore_logprob = True
+
+
 class LauncherHandle:
     def __init__(self, port: int):
         self.client = AsyncClient(f"http://localhost:{port}")
@@ -274,6 +285,11 @@ def generous_response_snapshot(snapshot):
     return snapshot.use_extension(GenerousResponseComparator)
 
 
+@pytest.fixture
+def ignore_logprob_response_snapshot(snapshot):
+    return snapshot.use_extension(IgnoreLogProbResponseComparator)
+
+
 @pytest.fixture(scope="module")
 def event_loop():
     loop = asyncio.get_event_loop()

integration-tests/models/__snapshots__/test_flash_llama_exl2/test_flash_llama_exl2.json

@@ -0,0 +1,84 @@
+{
+  "details": {
+    "best_of_sequences": null,
+    "finish_reason": "length",
+    "generated_tokens": 10,
+    "prefill": [
+      {
+        "id": 2323,
+        "logprob": null,
+        "text": "Test"
+      },
+      {
+        "id": 1715,
+        "logprob": -11.4375,
+        "text": " request"
+      }
+    ],
+    "seed": null,
+    "tokens": [
+      {
+        "id": 25,
+        "logprob": -2.9316406,
+        "special": false,
+        "text": ":"
+      },
+      {
+        "id": 330,
+        "logprob": -3.5136719,
+        "special": false,
+        "text": " \""
+      },
+      {
+        "id": 489,
+        "logprob": -0.7783203,
+        "special": false,
+        "text": " +"
+      },
+      {
+        "id": 1715,
+        "logprob": -1.2314453,
+        "special": false,
+        "text": " request"
+      },
+      {
+        "id": 489,
+        "logprob": -2.0019531,
+        "special": false,
+        "text": " +"
+      },
+      {
+        "id": 2990,
+        "logprob": -1.5009766,
+        "special": false,
+        "text": " \"\\"
+      },
+      {
+        "id": 77,
+        "logprob": -0.057434082,
+        "special": false,
+        "text": "n"
+      },
+      {
+        "id": 702,
+        "logprob": -1.4912109,
+        "special": false,
+        "text": "\"\n"
+      },
+      {
+        "id": 262,
+        "logprob": -1.2636719,
+        "special": false,
+        "text": "   "
+      },
+      {
+        "id": 557,
+        "logprob": -2.4042969,
+        "special": false,
+        "text": " }\n\n"
+      }
+    ],
+    "top_tokens": null
+  },
+  "generated_text": ": \" + request + \"\\n\"\n    }\n\n"
+}

integration-tests/models/__snapshots__/test_flash_llama_exl2/test_flash_llama_exl2_all_params.json

@@ -0,0 +1,84 @@
+{
+  "details": {
+    "best_of_sequences": null,
+    "finish_reason": "length",
+    "generated_tokens": 10,
+    "prefill": [
+      {
+        "id": 2323,
+        "logprob": null,
+        "text": "Test"
+      },
+      {
+        "id": 1715,
+        "logprob": -11.453125,
+        "text": " request"
+      }
+    ],
+    "seed": 0,
+    "tokens": [
+      {
+        "id": 13,
+        "logprob": -1.9980469,
+        "special": false,
+        "text": "."
+      },
+      {
+        "id": 578,
+        "logprob": -0.15795898,
+        "special": false,
+        "text": " The"
+      },
+      {
+        "id": 3622,
+        "logprob": -1.0458984,
+        "special": false,
+        "text": " server"
+      },
+      {
+        "id": 31680,
+        "logprob": -1.3623047,
+        "special": false,
+        "text": " responds"
+      },
+      {
+        "id": 449,
+        "logprob": 0.0,
+        "special": false,
+        "text": " with"
+      },
+      {
+        "id": 264,
+        "logprob": 0.0,
+        "special": false,
+        "text": " a"
+      },
+      {
+        "id": 330,
+        "logprob": -0.5678711,
+        "special": false,
+        "text": " \""
+      },
+      {
+        "id": 1049,
+        "logprob": -0.12322998,
+        "special": false,
+        "text": "200"
+      },
+      {
+        "id": 10619,
+        "logprob": 0.0,
+        "special": false,
+        "text": " OK"
+      },
+      {
+        "id": 1,
+        "logprob": 0.0,
+        "special": false,
+        "text": "\""
+      }
+    ],
+    "top_tokens": null
+  },
+  "generated_text": "Test request. The server responds with a \"200 OK\""
+}