WIP: Complete removal or f16_kv, add offload_kqv field

llama_cpp/llama.py

@@ -751,7 +751,6 @@ def __init__(
         yarn_beta_slow: float = 1.0,
         yarn_orig_ctx: int = 0,
         mul_mat_q: bool = True,
-        f16_kv: bool = True,
         logits_all: bool = False,
         embedding: bool = False,
         # Sampling Params
@@ -817,7 +816,6 @@ def __init__(
             yarn_beta_fast: YaRN low correction dim
             yarn_beta_slow: YaRN high correction dim
             yarn_orig_ctx: YaRN original context size
-            f16_kv: Use fp16 for KV cache, fp32 otherwise
             logits_all: Return logits for all tokens, not just the last token. Must be True for completion to return logprobs.
             embedding: Embedding mode only.
             last_n_tokens_size: Maximum number of tokens to keep in the last_n_tokens deque.
@@ -904,7 +902,6 @@ def __init__(
         )
         self.context_params.yarn_orig_ctx = yarn_orig_ctx if yarn_orig_ctx != 0 else 0
         self.context_params.mul_mat_q = mul_mat_q
-        # self.context_params.f16_kv = f16_kv
         self.context_params.logits_all = logits_all
         self.context_params.embedding = embedding
 
@@ -2155,7 +2152,6 @@ def __getstate__(self):
             yarn_beta_slow=self.context_params.yarn_beta_slow,
             yarn_orig_ctx=self.context_params.yarn_orig_ctx,
             mul_mat_q=self.context_params.mul_mat_q,
-            f16_kv=self.context_params.f16_kv,
             logits_all=self.context_params.logits_all,
             embedding=self.context_params.embedding,
             # Sampling Params
@@ -2198,7 +2194,6 @@ def __setstate__(self, state):
             yarn_beta_slow=state["yarn_beta_slow"],
             yarn_orig_ctx=state["yarn_orig_ctx"],
             mul_mat_q=state["mul_mat_q"],
-            f16_kv=state["f16_kv"],
             logits_all=state["logits_all"],
             embedding=state["embedding"],
             # Sampling Params

llama_cpp/llama_cpp.py

@@ -432,9 +432,9 @@ class llama_context_params(Structure):
         type_k (int): data type for K cache
         type_v (int): data type for V cache
         mul_mat_q (bool): if true, use experimental mul_mat_q kernels (DEPRECATED - always true)
-        f16_kv (bool): use fp16 for KV cache, fp32 otherwise
         logits_all (bool): the llama_eval() call computes all logits, not just the last one (DEPRECATED - set llama_batch.logits instead)
-        embedding (bool): embedding mode only"""
+        embedding (bool): embedding mode only
+        offload_kqv (bool): whether to offload the KQV ops (including the KV cache) to GPU"""
     _fields_ = [
         ("seed", c_uint32),
         ("n_ctx", c_uint32),
@@ -452,9 +452,9 @@ class llama_context_params(Structure):
         ("type_k", c_int),
         ("type_v", c_int),
         ("mul_mat_q", c_bool),
-        ("f16_kv", c_bool),
         ("logits_all", c_bool),
         ("embedding", c_bool),
+        ("offload_kqv", c_bool),
     ]
 
 

llama_cpp/server/app.py

@@ -98,7 +98,6 @@ class Settings(BaseSettings):
     mul_mat_q: bool = Field(
         default=True, description="if true, use experimental mul_mat_q kernels"
     )
-    f16_kv: bool = Field(default=True, description="Whether to use f16 key/value.")
     logits_all: bool = Field(default=True, description="Whether to return logits.")
     embedding: bool = Field(default=True, description="Whether to use embeddings.")
     # Sampling Params
@@ -408,7 +407,6 @@ def create_app(settings: Optional[Settings] = None):
         yarn_beta_slow=settings.yarn_beta_slow,
         yarn_orig_ctx=settings.yarn_orig_ctx,
         mul_mat_q=settings.mul_mat_q,
-        f16_kv=settings.f16_kv,
         logits_all=settings.logits_all,
         embedding=settings.embedding,
         # Sampling Params