[Model] Support Cohere2ForCausalLM (Cohere R7B) (#11203)

docs/source/models/supported_models.rst

@@ -118,9 +118,9 @@ Text Generation (``--task generate``)
     - :code:`THUDM/chatglm2-6b`, :code:`THUDM/chatglm3-6b`, etc.
     - ✅︎
     - ✅︎
-  * - :code:`CohereForCausalLM`
+  * - :code:`CohereForCausalLM`,:code:`Cohere2ForCausalLM`
     - Command-R
-    - :code:`CohereForAI/c4ai-command-r-v01`, etc.
+    - :code:`CohereForAI/c4ai-command-r-v01`, :code:`CohereForAI/c4ai-command-r7b-12-2024`, etc.
     - ✅︎
     - ✅︎
   * - :code:`DbrxForCausalLM`

tests/models/registry.py

@@ -53,6 +53,8 @@ _TEXT_GENERATION_EXAMPLE_MODELS = {
     # ChatGLMModel supports multimodal
     "CohereForCausalLM": _HfExamplesInfo("CohereForAI/c4ai-command-r-v01",
                                          trust_remote_code=True),
+    "Cohere2ForCausalLM": _HfExamplesInfo("CohereForAI/c4ai-command-r7b-12-2024", # noqa: E501
+                                         trust_remote_code=True),
     "DbrxForCausalLM": _HfExamplesInfo("databricks/dbrx-instruct"),
     "DeciLMForCausalLM": _HfExamplesInfo("Deci/DeciLM-7B-instruct",
                                          trust_remote_code=True),

tests/models/test_initialization.py

@@ -1,6 +1,7 @@
 from unittest.mock import patch
 
 import pytest
+import transformers
 from transformers import PretrainedConfig
 
 from vllm import LLM
@@ -11,6 +12,9 @@ from .registry import HF_EXAMPLE_MODELS
 @pytest.mark.parametrize("model_arch", HF_EXAMPLE_MODELS.get_supported_archs())
 def test_can_initialize(model_arch):
     model_info = HF_EXAMPLE_MODELS.get_hf_info(model_arch)
+    if (model_arch == "Cohere2ForCausalLM"
+            and transformers.__version__ < "4.48.0"):
+        pytest.skip(reason="Model introduced in HF >= 4.48.0")
     if not model_info.is_available_online:
         pytest.skip("Model is not available online")
 

vllm/model_executor/models/commandr.py

@@ -48,7 +48,7 @@ from vllm.model_executor.utils import set_weight_attrs
 from vllm.sequence import IntermediateTensors
 
 from .interfaces import SupportsLoRA, SupportsPP
-from .utils import (is_pp_missing_parameter,
+from .utils import (extract_layer_index, is_pp_missing_parameter,
                     make_empty_intermediate_tensors_factory, make_layers,
                     maybe_prefix)
 
@@ -171,12 +171,26 @@ class CohereAttention(nn.Module):
             rope_scaling=self.rope_scaling,
             is_neox_style=False,
         )
+
+        sliding_window = getattr(config, "sliding_window", None)
+        # Model v2 has sliding windows, v1 does not
+        self.v1 = sliding_window is None
+
+        layer_idx = extract_layer_index(prefix)
+        layer_has_sliding_window = (
+            getattr(config, "sliding_window_pattern", False)
+            and (layer_idx + 1) % self.config.sliding_window_pattern != 0)
+
+        self.sliding_window = (sliding_window
+                               if layer_has_sliding_window else None)
+
         self.attn = Attention(self.num_heads,
                               self.head_dim,
                               self.scaling,
                               num_kv_heads=self.num_kv_heads,
                               cache_config=cache_config,
                               quant_config=quant_config,
+                              per_layer_sliding_window=self.sliding_window,
                               prefix=f"{prefix}.attn")
         if self.use_qk_norm:
             self.q_norm = LayerNorm(param_shape=(self.num_heads,
@@ -206,6 +220,7 @@ class CohereAttention(nn.Module):
         q, k, v = qkv.split([self.q_size, self.kv_size, self.kv_size], dim=-1)
         if self.use_qk_norm:
             q, k = self._apply_qk_norm(q, k)
+        if self.v1 or self.sliding_window:
             q, k = self.rotary_emb(positions, q, k)
         attn_output = self.attn(q, k, v, kv_cache, attn_metadata)
         output, _ = self.o_proj(attn_output)

vllm/model_executor/models/registry.py

@@ -41,6 +41,7 @@ _TEXT_GENERATION_MODELS = {
     "BloomForCausalLM": ("bloom", "BloomForCausalLM"),
     # ChatGLMModel supports multimodal
     "CohereForCausalLM": ("commandr", "CohereForCausalLM"),
+    "Cohere2ForCausalLM": ("commandr", "CohereForCausalLM"),
     "DbrxForCausalLM": ("dbrx", "DbrxForCausalLM"),
     "DeciLMForCausalLM": ("decilm", "DeciLMForCausalLM"),
     "DeepseekForCausalLM": ("deepseek", "DeepseekForCausalLM"),