llama-cpp-python 0.2.53__tar.gz → 0.2.54__tar.gz

llama_cpp_python-0.2.54/.git/FETCH_HEAD

@@ -0,0 +1 @@
+d5df431278433b580e52222dbf4174f5102585b1		'd5df431278433b580e52222dbf4174f5102585b1' of https://github.com/abetlen/llama-cpp-python

llama_cpp_python-0.2.54/.git/HEAD

@@ -0,0 +1 @@
+d5df431278433b580e52222dbf4174f5102585b1

{llama_cpp_python-0.2.53 → llama_cpp_python-0.2.54}/.git/config

@@ -9,7 +9,7 @@
 [gc]
 	auto = 0
 [http "https://github.com/"]
-	extraheader = AUTHORIZATION: basic eC1hY2Nlc3MtdG9rZW46Z2hzX1FVcUlkd2s5NjZYRlh5RXNidElxQTluejZXYXhOWDNTRUQyaQ==
+	extraheader = AUTHORIZATION: basic eC1hY2Nlc3MtdG9rZW46Z2hzXzdpQXRKODgxbmFTbmVlTHlEM0pqN0pCRE1mazRldzBySUlxMw==
 [submodule "vendor/llama.cpp"]
 	active = true
 	url = https://github.com/ggerganov/llama.cpp.git

llama_cpp_python-0.2.54/.git/logs/HEAD

@@ -0,0 +1 @@
+0000000000000000000000000000000000000000 d5df431278433b580e52222dbf4174f5102585b1 runner <runner@fv-az1240-729.ompflkn1s5uuratdiiqak2yqtf.ex.internal.cloudapp.net> 1709316972 +0000	checkout: moving from master to refs/tags/v0.2.54

llama_cpp_python-0.2.54/.git/modules/vendor/llama.cpp/FETCH_HEAD

@@ -0,0 +1 @@
+c2224f003bf9cf558b1a3c57033563e11a4de9a5		'c2224f003bf9cf558b1a3c57033563e11a4de9a5' of https://github.com/ggerganov/llama.cpp

llama_cpp_python-0.2.54/.git/modules/vendor/llama.cpp/HEAD

@@ -0,0 +1 @@
+c2224f003bf9cf558b1a3c57033563e11a4de9a5

{llama_cpp_python-0.2.53 → llama_cpp_python-0.2.54}/.git/modules/vendor/llama.cpp/config

@@ -16,7 +16,7 @@
 [gc]
 	auto = 0
 [http "https://github.com/"]
-	extraheader = AUTHORIZATION: basic eC1hY2Nlc3MtdG9rZW46Z2hzX1FVcUlkd2s5NjZYRlh5RXNidElxQTluejZXYXhOWDNTRUQyaQ==
+	extraheader = AUTHORIZATION: basic eC1hY2Nlc3MtdG9rZW46Z2hzXzdpQXRKODgxbmFTbmVlTHlEM0pqN0pCRE1mazRldzBySUlxMw==
 [url "https://github.com/"]
 	insteadOf = git@github.com:
 	insteadOf = org-6826477@github.com:

llama_cpp_python-0.2.54/.git/modules/vendor/llama.cpp/logs/HEAD

@@ -0,0 +1,2 @@
+0000000000000000000000000000000000000000 38d16b142624bdd7c41d9955752b7f7b59c5e048 runner <runner@fv-az1240-729.ompflkn1s5uuratdiiqak2yqtf.ex.internal.cloudapp.net> 1709316973 +0000	clone: from https://github.com/ggerganov/llama.cpp.git
+38d16b142624bdd7c41d9955752b7f7b59c5e048 c2224f003bf9cf558b1a3c57033563e11a4de9a5 runner <runner@fv-az1240-729.ompflkn1s5uuratdiiqak2yqtf.ex.internal.cloudapp.net> 1709316974 +0000	checkout: moving from master to c2224f003bf9cf558b1a3c57033563e11a4de9a5

llama_cpp_python-0.2.54/.git/modules/vendor/llama.cpp/logs/refs/heads/master

@@ -0,0 +1 @@
+0000000000000000000000000000000000000000 38d16b142624bdd7c41d9955752b7f7b59c5e048 runner <runner@fv-az1240-729.ompflkn1s5uuratdiiqak2yqtf.ex.internal.cloudapp.net> 1709316973 +0000	clone: from https://github.com/ggerganov/llama.cpp.git

llama_cpp_python-0.2.54/.git/modules/vendor/llama.cpp/logs/refs/remotes/origin/HEAD

@@ -0,0 +1 @@
+0000000000000000000000000000000000000000 38d16b142624bdd7c41d9955752b7f7b59c5e048 runner <runner@fv-az1240-729.ompflkn1s5uuratdiiqak2yqtf.ex.internal.cloudapp.net> 1709316973 +0000	clone: from https://github.com/ggerganov/llama.cpp.git

{llama_cpp_python-0.2.53 → llama_cpp_python-0.2.54}/.git/modules/vendor/llama.cpp/modules/kompute/config

@@ -13,7 +13,7 @@
 [gc]
 	auto = 0
 [http "https://github.com/"]
-	extraheader = AUTHORIZATION: basic eC1hY2Nlc3MtdG9rZW46Z2hzX1FVcUlkd2s5NjZYRlh5RXNidElxQTluejZXYXhOWDNTRUQyaQ==
+	extraheader = AUTHORIZATION: basic eC1hY2Nlc3MtdG9rZW46Z2hzXzdpQXRKODgxbmFTbmVlTHlEM0pqN0pCRE1mazRldzBySUlxMw==
 [url "https://github.com/"]
 	insteadOf = git@github.com:
 	insteadOf = org-6826477@github.com:

llama_cpp_python-0.2.54/.git/modules/vendor/llama.cpp/modules/kompute/logs/HEAD

@@ -0,0 +1,2 @@
+0000000000000000000000000000000000000000 d1e3b0953cf66acc94b2e29693e221427b2c1f3f runner <runner@fv-az1240-729.ompflkn1s5uuratdiiqak2yqtf.ex.internal.cloudapp.net> 1709316974 +0000	clone: from https://github.com/nomic-ai/kompute.git
+d1e3b0953cf66acc94b2e29693e221427b2c1f3f 4565194ed7c32d1d2efa32ceab4d3c6cae006306 runner <runner@fv-az1240-729.ompflkn1s5uuratdiiqak2yqtf.ex.internal.cloudapp.net> 1709316975 +0000	checkout: moving from master to 4565194ed7c32d1d2efa32ceab4d3c6cae006306

llama_cpp_python-0.2.54/.git/modules/vendor/llama.cpp/modules/kompute/logs/refs/heads/master

@@ -0,0 +1 @@
+0000000000000000000000000000000000000000 d1e3b0953cf66acc94b2e29693e221427b2c1f3f runner <runner@fv-az1240-729.ompflkn1s5uuratdiiqak2yqtf.ex.internal.cloudapp.net> 1709316974 +0000	clone: from https://github.com/nomic-ai/kompute.git

llama_cpp_python-0.2.54/.git/modules/vendor/llama.cpp/modules/kompute/logs/refs/remotes/origin/HEAD

@@ -0,0 +1 @@
+0000000000000000000000000000000000000000 d1e3b0953cf66acc94b2e29693e221427b2c1f3f runner <runner@fv-az1240-729.ompflkn1s5uuratdiiqak2yqtf.ex.internal.cloudapp.net> 1709316974 +0000	clone: from https://github.com/nomic-ai/kompute.git

llama_cpp_python-0.2.54/.git/modules/vendor/llama.cpp/packed-refs

@@ -0,0 +1,2 @@
+# pack-refs with: peeled fully-peeled sorted 
+38d16b142624bdd7c41d9955752b7f7b59c5e048 refs/remotes/origin/master

llama_cpp_python-0.2.54/.git/modules/vendor/llama.cpp/refs/heads/master

@@ -0,0 +1 @@
+38d16b142624bdd7c41d9955752b7f7b59c5e048

llama_cpp_python-0.2.54/.git/modules/vendor/llama.cpp/shallow

@@ -0,0 +1,2 @@
+38d16b142624bdd7c41d9955752b7f7b59c5e048
+c2224f003bf9cf558b1a3c57033563e11a4de9a5

llama_cpp_python-0.2.54/.git/refs/tags/v0.2.54

@@ -0,0 +1 @@
+d5df431278433b580e52222dbf4174f5102585b1

llama_cpp_python-0.2.54/.git/shallow

@@ -0,0 +1 @@
+d5df431278433b580e52222dbf4174f5102585b1

{llama_cpp_python-0.2.53 → llama_cpp_python-0.2.54}/CHANGELOG.md

@@ -7,6 +7,11 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.2.54]
+
+- feat: Update llama.cpp to ggerganov/llama.cpp@cb49e0f8c906e5da49e9f6d64a57742a9a241c6a
+- docs: fix typo in README.md embeddings example by @iamlemec in #1232
+
 ## [0.2.53]
 
 - feat: Update llama.cpp to ggerganov/llama.cpp@cb49e0f8c906e5da49e9f6d64a57742a9a241c6a

{llama_cpp_python-0.2.53 → llama_cpp_python-0.2.54}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: llama_cpp_python
-Version: 0.2.53
+Version: 0.2.54
 Summary: Python bindings for the llama.cpp library
 Author-Email: Andrei Betlen <abetlen@gmail.com>
 License: MIT
@@ -329,7 +329,16 @@ By default [`from_pretrained`](https://llama-cpp-python.readthedocs.io/en/latest
 
 The high-level API also provides a simple interface for chat completion.
 
-Note that `chat_format` option must be set for the particular model you are using.
+Chat completion requires that the model know how to format the messages into a single prompt.
+The `Llama` class does this using pre-registered chat formats (ie. `chatml`, `llama-2`, `gemma`, etc) or by providing a custom chat handler object.
+
+The model will will format the messages into a single prompt using the following order of precedence:
+  - Use the `chat_handler` if provided
+  - Use the `chat_format` if provided
+  - Use the `tokenizer.chat_template` from the `gguf` model's metadata (should work for most new models, older models may not have this)
+  - else, fallback to the `llama-2` chat format
+
+Set `verbose=True` to see the selected chat format.
 
 ```python
 >>> from llama_cpp import Llama
@@ -568,7 +577,7 @@ To generate text embeddings use [`create_embedding`](http://localhost:8000/api-r
 ```python
 import llama_cpp
 
-llm = llama_cpp.Llama(model_path="path/to/model.gguf", embeddings=True)
+llm = llama_cpp.Llama(model_path="path/to/model.gguf", embedding=True)
 
 embeddings = llm.create_embedding("Hello, world!")
 

{llama_cpp_python-0.2.53 → llama_cpp_python-0.2.54}/README.md

@@ -286,7 +286,16 @@ By default [`from_pretrained`](https://llama-cpp-python.readthedocs.io/en/latest
 
 The high-level API also provides a simple interface for chat completion.
 
-Note that `chat_format` option must be set for the particular model you are using.
+Chat completion requires that the model know how to format the messages into a single prompt.
+The `Llama` class does this using pre-registered chat formats (ie. `chatml`, `llama-2`, `gemma`, etc) or by providing a custom chat handler object.
+
+The model will will format the messages into a single prompt using the following order of precedence:
+  - Use the `chat_handler` if provided
+  - Use the `chat_format` if provided
+  - Use the `tokenizer.chat_template` from the `gguf` model's metadata (should work for most new models, older models may not have this)
+  - else, fallback to the `llama-2` chat format
+
+Set `verbose=True` to see the selected chat format.
 
 ```python
 >>> from llama_cpp import Llama
@@ -525,7 +534,7 @@ To generate text embeddings use [`create_embedding`](http://localhost:8000/api-r
 ```python
 import llama_cpp
 
-llm = llama_cpp.Llama(model_path="path/to/model.gguf", embeddings=True)
+llm = llama_cpp.Llama(model_path="path/to/model.gguf", embedding=True)
 
 embeddings = llm.create_embedding("Hello, world!")
 

{llama_cpp_python-0.2.53 → llama_cpp_python-0.2.54}/llama_cpp/__init__.py

@@ -1,4 +1,4 @@
 from .llama_cpp import *
 from .llama import *
 
-__version__ = "0.2.53"
+__version__ = "0.2.54"

{llama_cpp_python-0.2.53 → llama_cpp_python-0.2.54}/llama_cpp/_internals.py

@@ -357,21 +357,6 @@ class _LlamaContext:
             penalty_present,
         )
 
-    def sample_classifier_free_guidance(
-        self,
-        candidates: "_LlamaTokenDataArray",
-        guidance_ctx: "_LlamaContext",
-        scale: float,
-    ):
-        assert self.ctx is not None
-        assert guidance_ctx.ctx is not None
-        llama_cpp.llama_sample_classifier_free_guidance(
-            self.ctx,
-            llama_cpp.byref(candidates.candidates),
-            guidance_ctx.ctx,
-            scale,
-        )
-
     def sample_softmax(self, candidates: "_LlamaTokenDataArray"):
         assert self.ctx is not None
         llama_cpp.llama_sample_softmax(
@@ -720,7 +705,7 @@ class _LlamaSamplingContext:
         return ctx_main.model.detokenize(self.prev[-n:]).decode("utf-8")
 
     def sample(
-        self, ctx_main: _LlamaContext, ctx_cfg: Optional[_LlamaContext] = None, idx: int = 0, logits_array: Optional[npt.NDArray[np.single]] = None
+        self, ctx_main: _LlamaContext, idx: int = 0, logits_array: Optional[npt.NDArray[np.single]] = None
     ):
         n_vocab = ctx_main.model.n_vocab()
         id: int = 0
@@ -741,11 +726,6 @@ class _LlamaSamplingContext:
         )  # TODO: Only create this once
         token_data_array.copy_logits(logits_array)
 
-        if ctx_cfg is not None:
-            ctx_main.sample_classifier_free_guidance(
-                token_data_array, ctx_cfg, self.params.cfg_scale
-            )
-
         # apply penalties
         if len(self.prev) > 0:
             nl_token = ctx_main.model.token_nl()

{llama_cpp_python-0.2.53 → llama_cpp_python-0.2.54}/llama_cpp/llama.py

@@ -86,7 +86,6 @@ class Llama:
         yarn_beta_fast: float = 32.0,
         yarn_beta_slow: float = 1.0,
         yarn_orig_ctx: int = 0,
-        mul_mat_q: bool = True,
         logits_all: bool = False,
         embedding: bool = False,
         offload_kqv: bool = True,
@@ -291,7 +290,6 @@ class Llama:
             yarn_beta_slow if yarn_beta_slow != 0.0 else 0
         )
         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.logits_all = (
             logits_all if draft_model is None else True
         )  # Must be set to True for speculative decoding
@@ -412,7 +410,7 @@ class Llama:
                 bos_token = self._model.token_get_text(bos_token_id)
 
                 if self.verbose:
-                    print(f"Using chat template: {template}", file=sys.stderr)
+                    print(f"Using gguf chat template: {template}", file=sys.stderr)
                     print(f"Using chat eos_token: {eos_token}", file=sys.stderr)
                     print(f"Using chat bos_token: {bos_token}", file=sys.stderr)
 
@@ -422,6 +420,8 @@ class Llama:
 
         if self.chat_format is None and self.chat_handler is None:
             self.chat_format = "llama-2"
+            if self.verbose:
+                print(f"Using fallback chat format: {chat_format}", file=sys.stderr)
 
     @property
     def ctx(self) -> llama_cpp.llama_context_p:
@@ -1724,7 +1724,6 @@ class Llama:
             yarn_beta_fast=self.context_params.yarn_beta_fast,
             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,
             logits_all=self.context_params.logits_all,
             embedding=self.context_params.embedding,
             # Sampling Params
@@ -1768,7 +1767,6 @@ class Llama:
             yarn_beta_fast=state["yarn_beta_fast"],
             yarn_beta_slow=state["yarn_beta_slow"],
             yarn_orig_ctx=state["yarn_orig_ctx"],
-            mul_mat_q=state["mul_mat_q"],
             logits_all=state["logits_all"],
             embedding=state["embedding"],
             # Sampling Params

{llama_cpp_python-0.2.53 → llama_cpp_python-0.2.54}/llama_cpp/llama_cpp.py

@@ -111,6 +111,7 @@ if TYPE_CHECKING:
 
 F = TypeVar("F", bound=Callable[..., Any])
 
+
 def ctypes_function_for_shared_library(lib: ctypes.CDLL):
     def ctypes_function(
         name: str, argtypes: List[Any], restype: Any, enabled: bool = True
@@ -558,9 +559,7 @@ class llama_model_params(ctypes.Structure):
 #     enum ggml_type type_k; // data type for K cache
 #     enum ggml_type type_v; // data type for V cache
 
-
 #     // Keep the booleans together to avoid misalignment during copy-by-value.
-#     bool mul_mat_q;   // if true, use experimental mul_mat_q kernels (DEPRECATED - always true)
 #     bool logits_all;  // the llama_eval() call computes all logits, not just the last one (DEPRECATED - set llama_batch.logits instead)
 #     bool embedding;   // embedding mode only
 #     bool offload_kqv; // whether to offload the KQV ops (including the KV cache) to GPU
@@ -588,7 +587,6 @@ class llama_context_params(ctypes.Structure):
         cb_eval_user_data (ctypes.ctypes.c_void_p): user data for cb_eval
         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)
         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
         offload_kqv (bool): whether to offload the KQV ops (including the KV cache) to GPU
@@ -614,7 +612,6 @@ class llama_context_params(ctypes.Structure):
         ("cb_eval_user_data", ctypes.c_void_p),
         ("type_k", ctypes.c_int),
         ("type_v", ctypes.c_int),
-        ("mul_mat_q", ctypes.c_bool),
         ("logits_all", ctypes.c_bool),
         ("embedding", ctypes.c_bool),
         ("offload_kqv", ctypes.c_bool),
@@ -938,18 +935,6 @@ def llama_supports_gpu_offload() -> bool:
     ...
 
 
-# LLAMA_API DEPRECATED(bool llama_mmap_supported (void), "use llama_supports_mmap() instead");
-@ctypes_function("llama_mmap_supported", [], ctypes.c_bool)
-def llama_mmap_supported() -> bool:
-    ...
-
-
-# LLAMA_API DEPRECATED(bool llama_mlock_supported(void), "use llama_supports_mlock() instead");
-@ctypes_function("llama_mlock_supported", [], ctypes.c_bool)
-def llama_mlock_supported() -> bool:
-    ...
-
-
 # LLAMA_API const struct llama_model * llama_get_model(const struct llama_context * ctx);
 @ctypes_function("llama_get_model", [llama_context_p_ctypes], llama_model_p_ctypes)
 def llama_get_model(ctx: llama_context_p, /) -> Optional[llama_model_p]:
@@ -1158,47 +1143,6 @@ def llama_model_quantize(
     ...
 
 
-# // Apply a LoRA adapter to a loaded model
-# // path_base_model is the path to a higher quality model to use as a base for
-# // the layers modified by the adapter. Can be NULL to use the current loaded model.
-# // The model needs to be reloaded before applying a new adapter, otherwise the adapter
-# // will be applied on top of the previous one
-# // Returns 0 on success
-# LLAMA_API DEPRECATED(int32_t llama_apply_lora_from_file(
-#         struct llama_context * ctx,
-#                   const char * path_lora,
-#                        float   scale,
-#                   const char * path_base_model,
-#                      int32_t   n_threads),
-#         "use llama_model_apply_lora_from_file instead");
-@ctypes_function(
-    "llama_apply_lora_from_file",
-    [
-        llama_context_p_ctypes,
-        ctypes.c_char_p,
-        ctypes.c_float,
-        ctypes.c_char_p,
-        ctypes.c_int32,
-    ],
-    ctypes.c_int32,
-)
-def llama_apply_lora_from_file(
-    ctx: llama_context_p,
-    path_lora: Union[ctypes.c_char_p, bytes],
-    scale: Union[ctypes.c_float, float],
-    path_base_model: Union[ctypes.c_char_p, bytes],
-    n_threads: Union[ctypes.c_int32, int],
-    /,
-) -> int:
-    """Apply a LoRA adapter to a loaded model
-    path_base_model is the path to a higher quality model to use as a base for
-    the layers modified by the adapter. Can be NULL to use the current loaded model.
-    The model needs to be reloaded before applying a new adapter, otherwise the adapter
-    will be applied on top of the previous one
-    Returns 0 on success"""
-    ...
-
-
 # LLAMA_API int32_t llama_model_apply_lora_from_file(
 #         const struct llama_model * model,
 #                   const char * path_lora,
@@ -1220,7 +1164,7 @@ def llama_model_apply_lora_from_file(
     model: llama_model_p,
     path_lora: Union[ctypes.c_char_p, bytes],
     scale: Union[ctypes.c_float, float],
-    path_base_model: Union[ctypes.c_char_p, bytes],
+    path_base_model: Union[ctypes.c_char_p, bytes, None],
     n_threads: Union[ctypes.c_int32, int],
     /,
 ) -> int:
@@ -1571,11 +1515,11 @@ def llama_copy_state_data(
     ...
 
 
-# Set the state reading from the specified address
-# Returns the number of bytes read
+# // Set the state reading from the specified address
+# // Returns the number of bytes read
 # LLAMA_API size_t llama_set_state_data(
 #         struct llama_context * ctx,
-#                      uint8_t * src);
+#                const uint8_t * src);
 @ctypes_function(
     "llama_set_state_data",
     [llama_context_p_ctypes, ctypes.POINTER(ctypes.c_uint8)],
@@ -1647,72 +1591,6 @@ def llama_save_session_file(
 # //
 
 
-# // Run the llama inference to obtain the logits and probabilities for the next token(s).
-# // tokens + n_tokens is the provided batch of new tokens to process
-# // n_past is the number of tokens to use from previous eval calls
-# // Returns 0 on success
-# // DEPRECATED: use llama_decode() instead
-# LLAMA_API DEPRECATED(int llama_eval(
-#         struct llama_context * ctx,
-#                  llama_token * tokens,
-#                      int32_t   n_tokens,
-#                      int32_t   n_past),
-#         "use llama_decode() instead");
-@ctypes_function(
-    "llama_eval",
-    [
-        llama_context_p_ctypes,
-        llama_token_p,
-        ctypes.c_int32,
-        ctypes.c_int32,
-    ],
-    ctypes.c_int,
-)
-def llama_eval(
-    ctx: llama_context_p,
-    tokens: CtypesArray[llama_token],
-    n_tokens: Union[ctypes.c_int, int],
-    n_past: Union[ctypes.c_int, int],
-    /,
-) -> int:
-    """Run the llama inference to obtain the logits and probabilities for the next token(s).
-    tokens + n_tokens is the provided batch of new tokens to process
-    n_past is the number of tokens to use from previous eval calls
-    Returns 0 on success
-    DEPRECATED: use llama_decode() instead"""
-    ...
-
-
-# // Same as llama_eval, but use float matrix input directly.
-# // DEPRECATED: use llama_decode() instead
-# LLAMA_API DEPRECATED(int llama_eval_embd(
-#         struct llama_context * ctx,
-#                        float * embd,
-#                      int32_t   n_tokens,
-#                      int32_t   n_past),
-#         "use llama_decode() instead");
-@ctypes_function(
-    "llama_eval_embd",
-    [
-        llama_context_p_ctypes,
-        ctypes.POINTER(ctypes.c_float),
-        ctypes.c_int32,
-        ctypes.c_int32,
-    ],
-    ctypes.c_int,
-)
-def llama_eval_embd(
-    ctx: llama_context_p,
-    embd: CtypesArray[ctypes.c_float],
-    n_tokens: Union[ctypes.c_int, int],
-    n_past: Union[ctypes.c_int, int],
-    /,
-) -> int:
-    """Same as llama_eval, but use float matrix input directly.
-    DEPRECATED: use llama_decode() instead"""
-    ...
-
-
 # // Return batch for single sequence of tokens starting at pos_0
 # //
 # // NOTE: this is a helper function to facilitate transition to the new batch API - avoid using it
@@ -2247,35 +2125,6 @@ def llama_sample_apply_guidance(
     ...
 
 
-# LLAMA_API DEPRECATED(void llama_sample_classifier_free_guidance(
-#           struct llama_context * ctx,
-#         llama_token_data_array * candidates,
-#           struct llama_context * guidance_ctx,
-#                          float   scale),
-#           "use llama_sample_apply_guidance() instead");
-@ctypes_function(
-    "llama_sample_classifier_free_guidance",
-    [
-        llama_context_p_ctypes,
-        llama_token_data_array_p,
-        llama_context_p_ctypes,
-        ctypes.c_float,
-    ],
-    None,
-)
-def llama_sample_classifier_free_guidance(
-    ctx: llama_context_p,
-    candidates: Union[
-        CtypesArray[llama_token_data_array], CtypesPointerOrRef[llama_token_data_array]
-    ],
-    guidance_ctx: llama_context_p,
-    scale: Union[ctypes.c_float, float],
-    /,
-):
-    """Apply classifier-free guidance to the logits as described in academic paper "Stay on topic with Classifier-Free Guidance" https://arxiv.org/abs/2306.17806"""
-    ...
-
-
 # /// @details Sorts candidate tokens by their logits in descending order and calculate probabilities based on logits.
 # LLAMA_API void llama_sample_softmax(
 #         struct llama_context * ctx,
@@ -2474,28 +2323,6 @@ def llama_sample_temp(
     ...
 
 
-# LLAMA_API DEPRECATED(void llama_sample_temperature(
-#             struct llama_context * ctx,
-#           llama_token_data_array * candidates,
-#                            float   temp),
-#         "use llama_sample_temp instead");
-@ctypes_function(
-    "llama_sample_temperature",
-    [llama_context_p_ctypes, llama_token_data_array_p, ctypes.c_float],
-    None,
-)
-def llama_sample_temperature(
-    ctx: llama_context_p,
-    candidates: Union[
-        CtypesArray[llama_token_data_array], CtypesPointerOrRef[llama_token_data_array]
-    ],
-    temp: Union[ctypes.c_float, float],
-    /,
-):
-    """use llama_sample_temp instead"""
-    ...
-
-
 # /// @details Apply constraints from grammar
 # LLAMA_API void llama_sample_grammar(
 #         struct llama_context * ctx,