Make awq install optional + integration tests values.

huggingface · Sep 25, 2023 · 02d4f62 · 02d4f62
1 parent a8f870a
commit 02d4f62
Show file tree

Hide file tree

Showing 6 changed files with 556 additions and 12 deletions.
diff --git a/Dockerfile b/Dockerfile
@@ -175,6 +175,8 @@ COPY --from=flash-att-v2-builder /usr/src/flash-attention-v2/build/lib.linux-x86
 COPY --from=custom-kernels-builder /usr/src/build/lib.linux-x86_64-cpython-39 /opt/conda/lib/python3.9/site-packages
 # Copy build artifacts from exllama kernels builder
 COPY --from=exllama-kernels-builder /usr/src/build/lib.linux-x86_64-cpython-39 /opt/conda/lib/python3.9/site-packages
+# Copy build artifacts from exllama kernels builder
+COPY --from=awq-kernels-builder /usr/src/build/lib.linux-x86_64-cpython-39 /opt/conda/lib/python3.9/site-packages
 
 # Copy builds artifacts from vllm builder
 COPY --from=vllm-builder /usr/src/vllm/build/lib.linux-x86_64-cpython-39 /opt/conda/lib/python3.9/site-packages

diff --git a/integration-tests/models/__snapshots__/test_flash_awq/test_flash_llama_awq.json b/integration-tests/models/__snapshots__/test_flash_awq/test_flash_llama_awq.json
@@ -0,0 +1,89 @@
+{
+  "details": {
+    "best_of_sequences": null,
+    "finish_reason": "length",
+    "generated_tokens": 10,
+    "prefill": [
+      {
+        "id": 1,
+        "logprob": null,
+        "text": "<s>"
+      },
+      {
+        "id": 4321,
+        "logprob": -8.515625,
+        "text": "Test"
+      },
+      {
+        "id": 2009,
+        "logprob": -15.4140625,
+        "text": "request"
+      }
+    ],
+    "seed": null,
+    "tokens": [
+      {
+        "id": 29896,
+        "logprob": -2.0292969,
+        "special": false,
+        "text": "1"
+      },
+      {
+        "id": 13,
+        "logprob": -2.2597656,
+        "special": false,
+        "text": "\n"
+      },
+      {
+        "id": 30166,
+        "logprob": -3.8671875,
+        "special": false,
+        "text": ""
+      },
+      {
+        "id": 30166,
+        "logprob": -1.0488281,
+        "special": false,
+        "text": ""
+      },
+      {
+        "id": 30166,
+        "logprob": -0.24523926,
+        "special": false,
+        "text": ""
+      },
+      {
+        "id": 30166,
+        "logprob": -0.07897949,
+        "special": false,
+        "text": ""
+      },
+      {
+        "id": 30166,
+        "logprob": -0.023513794,
+        "special": false,
+        "text": ""
+      },
+      {
+        "id": 30166,
+        "logprob": -0.011444092,
+        "special": false,
+        "text": ""
+      },
+      {
+        "id": 30166,
+        "logprob": -0.008430481,
+        "special": false,
+        "text": ""
+      },
+      {
+        "id": 30166,
+        "logprob": -0.007648468,
+        "special": false,
+        "text": ""
+      }
+    ],
+    "top_tokens": null
+  },
+  "generated_text": "1\n"
+}
diff --git a/integration-tests/models/__snapshots__/test_flash_awq/test_flash_llama_awq_all_params.json b/integration-tests/models/__snapshots__/test_flash_awq/test_flash_llama_awq_all_params.json
@@ -0,0 +1,89 @@
+{
+  "details": {
+    "best_of_sequences": null,
+    "finish_reason": "length",
+    "generated_tokens": 10,
+    "prefill": [
+      {
+        "id": 1,
+        "logprob": null,
+        "text": "<s>"
+      },
+      {
+        "id": 4321,
+        "logprob": -8.515625,
+        "text": "Test"
+      },
+      {
+        "id": 2009,
+        "logprob": -15.4140625,
+        "text": "request"
+      }
+    ],
+    "seed": 0,
+    "tokens": [
+      {
+        "id": 29896,
+        "logprob": 0.0,
+        "special": false,
+        "text": "1"
+      },
+      {
+        "id": 13,
+        "logprob": -0.6254883,
+        "special": false,
+        "text": "\n"
+      },
+      {
+        "id": 30166,
+        "logprob": 0.0,
+        "special": false,
+        "text": ""
+      },
+      {
+        "id": 29918,
+        "logprob": -0.20141602,
+        "special": false,
+        "text": "_"
+      },
+      {
+        "id": 29906,
+        "logprob": -0.6254883,
+        "special": false,
+        "text": "2"
+      },
+      {
+        "id": 29871,
+        "logprob": 0.0,
+        "special": false,
+        "text": " "
+      },
+      {
+        "id": 30166,
+        "logprob": 0.0,
+        "special": false,
+        "text": ""
+      },
+      {
+        "id": 30166,
+        "logprob": 0.0,
+        "special": false,
+        "text": ""
+      },
+      {
+        "id": 30166,
+        "logprob": 0.0,
+        "special": false,
+        "text": ""
+      },
+      {
+        "id": 30166,
+        "logprob": 0.0,
+        "special": false,
+        "text": ""
+      }
+    ],
+    "top_tokens": null
+  },
+  "generated_text": "Test request1\n_2 "
+}