Update python and rust dependencies (IBM#56)

Including transformers 4.38.2

---------

Signed-off-by: Nick Hill <[email protected]>

Author: njhill

Cargo.lock

@@ -9,9 +9,9 @@ python = ">=3.11"
 
 [tool.poetry.group.dev.dependencies]
 protobuf = "^4.25.3"
-grpcio-tools = "^1.60.1"
-pytest = "^8.0.1"
-pytest-asyncio = "^0.23.3"
+grpcio-tools = "^1.62.1"
+pytest = "^8.0.2"
+pytest-asyncio = "^0.23.5"
 requests = "^2.31.0"
 pyyaml = "^6.0.1"
 

integration_tests/poetry.lock

@@ -6,7 +6,7 @@ authors = ["Olivier Dehaene"]
 description = "Text Generation Launcher"
 
 [dependencies]
-clap = { version = "4.5.1", features = ["derive", "env"] }
+clap = { version = "4.5.2", features = ["derive", "env"] }
 ctrlc = { version = "3.4.2", features = ["termination"] }
 nix = { version = "0.28.0", features = ["process", "signal"] }
 serde_json = "^1.0.114"

integration_tests/pyproject.toml

@@ -16,7 +16,7 @@ path = "src/main.rs"
 [dependencies]
 axum = { version = "0.6.20", features = ["json"] }
 text-generation-client = { path = "client" }
-clap = { version = "^4.5.1", features = ["derive", "env"] }
+clap = { version = "^4.5.2", features = ["derive", "env"] }
 futures = "^0.3.30"
 flume = "^0.11.0"
 metrics = "0.21.1"

launcher/Cargo.toml

@@ -10,8 +10,8 @@ text-generation-server = 'text_generation_server.cli:app'
 [tool.poetry.dependencies]
 python = ">=3.11.0,<3.13"
 protobuf = "^4.25.3"
-grpcio = "^1.60.1"
-grpcio-reflection = "^1.60.1"
+grpcio = "^1.62.1"
+grpcio-reflection = "^1.62.1"
 loguru = "^0.7.2"
 typer = "^0.9.0"
 accelerate = { version = "0.27.2", optional = true }
@@ -21,10 +21,10 @@ safetensors = "^0.4.2"
 sentencepiece = "^0.2.0"
 datasets = { version = "^2.15.0", optional = true }
 texttable = { version = "^1.7.0", optional = true }
-transformers = "4.38.0"
+transformers = "4.38.2"
 optimum = { version = "^1.17.1", extras = ["onnxruntime-gpu"], optional = true }
-onnxruntime = { version = "^1.17.0", optional = true }
-onnxruntime-gpu = { version = "^1.16.3", optional = true }
+onnxruntime = { version = "^1.17.1", optional = true }
+onnxruntime-gpu = { version = "^1.17.1", optional = true }
 onnx = { version = "^1.15.0", optional = true }
 einops = "^0.7.0"
 ibm-fms = { version = "^0.0", optional = true }
@@ -48,7 +48,7 @@ onnx-gpu = ["optimum", "onnxruntime-gpu", "onnx"]
 quantize = ["datasets", "texttable"]
 
 [tool.poetry.group.dev.dependencies]
-grpcio-tools = "^1.60.1"
+grpcio-tools = "^1.62.1"
 pytest = "^8.0.1"
 
 [build-system]

router/Cargo.toml

@@ -608,6 +608,7 @@ def forward(
     ) -> Tuple[torch.Tensor, List[Tuple[torch.Tensor, torch.Tensor]], int]:
         model_inputs = self.model.prepare_inputs_for_generation(
             input_ids, past_key_values,
+            inputs_embeds=inputs_embeds,
             attention_mask=attention_mask,
             position_ids=position_ids,
             use_cache=True,
@@ -619,8 +620,8 @@ def forward(
             # This can be incorrectly overwritten to None in prepare_inputs_for_generation
             model_inputs["position_ids"] = position_ids
 
-        if inputs_embeds is not None:
-            # Add embeddings - if non-None then input_ids should be None
+        if inputs_embeds is not None and "inputs_embeds" not in model_inputs:
+            # Ensure that embeddings were added - if non-None then input_ids should be None
             model_inputs["inputs_embeds"] = inputs_embeds
 
         # Model Forward