fix: prefer seed 1 in all cases

integration-tests/models/__snapshots__/test_bloom_560m/test_bloom_560m.json

@@ -60,69 +60,70 @@
         "text": " d'abord"
       }
     ],
-    "seed": 0,
+    "seed": 1,
     "tokens": [
       {
-        "id": 578,
+        "id": 2997,
-        "logprob": -1.6591797,
+        "logprob": -4.4101562,
         "special": false,
-        "text": " le"
+        "text": " vous"
       },
       {
-        "id": 5608,
+        "id": 71256,
-        "logprob": -2.4492188,
+        "logprob": -5.3828125,
         "special": false,
-        "text": " faire"
+        "text": " retrouver"
       },
       {
-        "id": 159570,
+        "id": 693,
-        "logprob": -6.6835938,
+        "logprob": -2.1308594,
         "special": false,
-        "text": " réch"
+        "text": " à"
       },
       {
-        "id": 810,
+        "id": 366,
-        "logprob": 0.0,
+        "logprob": -1.5234375,
         "special": false,
-        "text": "au"
+        "text": " la"
       },
       {
-        "id": 12736,
+        "id": 221398,
-        "logprob": 0.0,
+        "logprob": -2.671875,
         "special": false,
-        "text": "ffer"
+        "text": " terrasse"
       },
       {
-        "id": 1742,
+        "id": 1375,
-        "logprob": -2.5175781,
+        "logprob": -4.375,
         "special": false,
-        "text": " au"
+        "text": " pour"
       },
       {
-        "id": 6105,
+        "id": 86887,
-        "logprob": -2.0078125,
+        "logprob": -4.859375,
         "special": false,
-        "text": " bain"
+        "text": " essayer"
       },
       {
-        "id": 88254,
+        "id": 2155,
-        "logprob": -0.12695312,
+        "logprob": -2.7519531,
         "special": false,
-        "text": "-mar"
+        "text": " ce"
       },
       {
-        "id": 641,
+        "id": 5743,
-        "logprob": 0.0,
+        "logprob": -3.1992188,
         "special": false,
-        "text": "ie"
+        "text": " jus"
       },
       {
-        "id": 2940,
+        "id": 1479,
-        "logprob": -3.5175781,
+        "logprob": -3.3203125,
         "special": false,
-        "text": " avec"
+        "text": " qui"
       }
-    ]
+    ],
+    "top_tokens": null
   },
-  "generated_text": " le faire réchauffer au bain-marie avec"
+  "generated_text": " vous retrouver à la terrasse pour essayer ce jus qui"
 }

integration-tests/models/__snapshots__/test_bloom_560m/test_bloom_560m_all_params.json

@@ -30,7 +30,7 @@
         "text": " d'abord"
       }
     ],
-    "seed": 0,
+    "seed": 1,
     "tokens": [
       {
         "id": 408,
@@ -38,61 +38,62 @@
         "special": false,
         "text": " que"
       },
-      {
-        "id": 366,
-        "logprob": -1.2939453,
-        "special": false,
-        "text": " la"
-      },
-      {
-        "id": 8769,
-        "logprob": -0.3708496,
-        "special": false,
-        "text": " personne"
-      },
-      {
-        "id": 1479,
-        "logprob": -2.2871094,
-        "special": false,
-        "text": " qui"
-      },
       {
         "id": 2997,
-        "logprob": -0.8671875,
+        "logprob": -1.2939453,
         "special": false,
         "text": " vous"
       },
       {
-        "id": 35977,
+        "id": 116938,
-        "logprob": -1.5097656,
+        "logprob": -0.8618164,
         "special": false,
-        "text": " suit"
+        "text": " ayez"
-      },
-      {
-        "id": 21558,
-        "logprob": -0.07891846,
-        "special": false,
-        "text": " ait"
       },
       {
         "id": 447,
-        "logprob": -0.12695312,
+        "logprob": 0.0,
         "special": false,
         "text": " un"
       },
       {
-        "id": 78606,
+        "id": 11299,
-        "logprob": -2.21875,
+        "logprob": -0.20141602,
         "special": false,
-        "text": " profil"
+        "text": " compte"
       },
       {
-        "id": 3899,
+        "id": 198236,
-        "logprob": -1.3535156,
+        "logprob": -0.4741211,
         "special": false,
-        "text": " bien"
+        "text": " PayPal"
+      },
+      {
+        "id": 17,
+        "logprob": 0.0,
+        "special": false,
+        "text": "."
+      },
+      {
+        "id": 12424,
+        "logprob": -0.6040039,
+        "special": false,
+        "text": " Pour"
+      },
+      {
+        "id": 11676,
+        "logprob": -0.9741211,
+        "special": false,
+        "text": " cela"
+      },
+      {
+        "id": 915,
+        "logprob": -0.31323242,
+        "special": false,
+        "text": " :"
       }
-    ]
+    ],
+    "top_tokens": null
   },
-  "generated_text": "Pour déguster un ortolan, il faut tout d'abord que la personne qui vous suit ait un profil bien"
+  "generated_text": "Pour déguster un ortolan, il faut tout d'abord que vous ayez un compte PayPal. Pour cela :"
 }

integration-tests/models/test_bloom_560m.py

@@ -20,7 +20,7 @@ async def test_bloom_560m(bloom_560, response_snapshot):
         max_new_tokens=10,
         top_p=0.9,
         decoder_input_details=True,
-        seed=0,
+        seed=1,
     )
 
     assert response.details.generated_tokens == 10
@@ -42,7 +42,7 @@ async def test_bloom_560m_all_params(bloom_560, response_snapshot):
         typical_p=0.9,
         watermark=True,
         decoder_input_details=True,
-        seed=0,
+        seed=1,
     )
 
     assert response.details.generated_tokens == 10

integration-tests/models/test_bloom_560m_sharded.py

@@ -20,7 +20,7 @@ async def test_bloom_560m_sharded(bloom_560m_sharded, response_snapshot):
         max_new_tokens=10,
         top_p=0.9,
         decoder_input_details=True,
-        seed=0,
+        seed=1,
     )
 
     assert response.details.generated_tokens == 10

integration-tests/models/test_flash_awq.py

@@ -45,7 +45,7 @@ async def test_flash_llama_awq_all_params(flash_llama_awq, response_snapshot):
         typical_p=0.9,
         watermark=True,
         decoder_input_details=True,
-        seed=0,
+        seed=1,
     )
 
     assert response.details.generated_tokens == 10

integration-tests/models/test_flash_falcon.py

@@ -42,7 +42,7 @@ async def test_flash_falcon_all_params(flash_falcon, response_snapshot):
         typical_p=0.9,
         watermark=True,
         decoder_input_details=True,
-        seed=0,
+        seed=1,
     )
 
     assert response.details.generated_tokens == 10

integration-tests/models/test_flash_gemma.py

@@ -13,7 +13,7 @@ async def flash_gemma(flash_gemma_handle):
     return flash_gemma_handle.client
 
 
-@pytest.mark.skip
+# @pytest.mark.skip
 @pytest.mark.asyncio
 @pytest.mark.private
 async def test_flash_gemma(flash_gemma, response_snapshot):
@@ -25,7 +25,7 @@ async def test_flash_gemma(flash_gemma, response_snapshot):
     assert response == response_snapshot
 
 
-@pytest.mark.skip
+# @pytest.mark.skip
 @pytest.mark.asyncio
 @pytest.mark.private
 async def test_flash_gemma_all_params(flash_gemma, response_snapshot):
@@ -42,14 +42,14 @@ async def test_flash_gemma_all_params(flash_gemma, response_snapshot):
         typical_p=0.9,
         watermark=True,
         decoder_input_details=True,
-        seed=0,
+        seed=1,
     )
 
     assert response.details.generated_tokens == 10
     assert response == response_snapshot
 
 
-@pytest.mark.skip
+# @pytest.mark.skip
 @pytest.mark.asyncio
 @pytest.mark.private
 async def test_flash_gemma_load(flash_gemma, generate_load, response_snapshot):

integration-tests/models/test_flash_llama.py

@@ -40,7 +40,7 @@ async def test_flash_llama_all_params(flash_llama, response_snapshot):
         typical_p=0.9,
         watermark=True,
         decoder_input_details=True,
-        seed=0,
+        seed=1,
     )
 
     assert response.details.generated_tokens == 5

integration-tests/models/test_flash_llama_gptq.py

@@ -39,7 +39,7 @@ async def test_flash_llama_gptq_all_params(flash_llama_gptq, response_snapshot):
         typical_p=0.9,
         watermark=True,
         decoder_input_details=True,
-        seed=0,
+        seed=1,
     )
 
     assert response.details.generated_tokens == 10

integration-tests/models/test_flash_medusa.py

@@ -38,7 +38,7 @@ async def test_flash_medusa_all_params(flash_medusa, response_snapshot):
         typical_p=0.9,
         watermark=True,
         decoder_input_details=True,
-        seed=0,
+        seed=1,
     )
 
     assert response.details.generated_tokens == 10

integration-tests/models/test_flash_mistral.py

@@ -39,7 +39,7 @@ async def test_flash_mistral_all_params(flash_mistral, response_snapshot):
         typical_p=0.9,
         watermark=True,
         decoder_input_details=True,
-        seed=0,
+        seed=1,
     )
 
     assert response.details.generated_tokens == 10

integration-tests/models/test_flash_phi.py

@@ -39,7 +39,7 @@ async def test_flash_phi_all_params(flash_phi, response_snapshot):
         typical_p=0.9,
         watermark=True,
         decoder_input_details=True,
-        seed=0,
+        seed=1,
     )
 
     assert response.details.generated_tokens == 6

integration-tests/models/test_flash_starcoder.py

@@ -33,7 +33,7 @@ async def test_flash_starcoder_default_params(flash_starcoder, response_snapshot
         temperature=0.2,
         top_p=0.95,
         decoder_input_details=True,
-        seed=0,
+        seed=1,
     )
 
     assert response.details.generated_tokens == 60

integration-tests/models/test_flash_starcoder_gptq.py

@@ -34,7 +34,7 @@ async def test_flash_starcoder_gptq_default_params(
         temperature=0.2,
         top_p=0.95,
         decoder_input_details=True,
-        seed=0,
+        seed=1,
     )
     assert response.details.generated_tokens == 20
     assert response == generous_response_snapshot

integration-tests/models/test_grammar_llama.py

@@ -34,7 +34,7 @@ async def test_flash_llama_grammar_regex(flash_llama_grammar, response_snapshot)
         "Whats Googles DNS",
         max_new_tokens=10,
         decoder_input_details=True,
-        seed=0,
+        seed=1,
         grammar={
             "type": GrammarType.Regex,  # "regex"
             "value": "((25[0-5]|2[0-4]\\d|[01]?\\d\\d?)\\.){3}(25[0-5]|2[0-4]\\d|[01]?\\d\\d?)",
@@ -52,7 +52,7 @@ async def test_flash_llama_grammar_json(flash_llama_grammar, response_snapshot):
         "info: david holtz like trees and has two cats. ",
         max_new_tokens=100,
         decoder_input_details=True,
-        seed=0,
+        seed=1,
         grammar={
             "type": GrammarType.Json,  # "json"
             "value": json.dumps(
@@ -104,7 +104,7 @@ async def test_flash_llama_grammar_load(
         max_new_tokens=10,
         n=4,
         stop_sequences=[".com"],
-        seed=0,
+        seed=1,
         grammar={
             "type": GrammarType.Regex,  # "regex"
             "value": "[\\w-]+@([\\w-]+\\.)+[\\w-]+",  # email regex
@@ -133,7 +133,7 @@ async def test_flash_llama_grammar_single_load_instance(
         "name: david. email:  ",
         max_new_tokens=10,
         stop_sequences=[".com"],
-        seed=0,
+        seed=1,
         grammar={
             "type": GrammarType.Regex,  # "regex"
             "value": "[\\w-]+@([\\w-]+\\.)+[\\w-]+",  # email regex

integration-tests/models/test_mamba.py

@@ -39,7 +39,7 @@ async def test_mamba_all_params(fused_kernel_mamba, response_snapshot):
         typical_p=0.9,
         watermark=True,
         decoder_input_details=True,
-        seed=0,
+        seed=1,
     )
 
     assert response.details.generated_tokens == 10

integration-tests/models/test_mt0_base.py

@@ -20,7 +20,7 @@ async def test_mt0_base(mt0_base, response_snapshot):
         max_new_tokens=10,
         top_p=0.9,
         decoder_input_details=True,
-        seed=0,
+        seed=1,
     )
 
     assert response.details.generated_tokens == 5
@@ -42,7 +42,7 @@ async def test_mt0_base_all_params(mt0_base, response_snapshot):
         typical_p=0.9,
         watermark=True,
         decoder_input_details=True,
-        seed=0,
+        seed=1,
     )
 
     assert response.details.generated_tokens == 9