Merge branch 'utilityai:main' into main

Author: estokes

.github/workflows/llama-cpp-rs-check.yml

@@ -51,7 +51,7 @@ jobs:
       - name: Set up Docker Buildx
         uses: docker/setup-buildx-action@d70bba72b1f3fd22344832f00baa16ece964efeb
       - name: Build
-        uses: docker/build-push-action@v5
+        uses: docker/build-push-action@v6
         with:
           file: test-build.Dockerfile
           target: base-cuda

Cargo.lock

@@ -16,7 +16,7 @@ hf-hub = { version = "0.3.2" }
 criterion = "0.5.1"
 pprof = "0.13.0"
 bindgen = "0.69.4"
-cc = "1.0.99"
+cc = "1.0.100"
 anyhow = "1.0.86"
 clap = "4.5.4"
 encoding_rs = "0.8.34"

Cargo.toml

@@ -1,12 +1,12 @@
 [package]
 name = "embeddings"
-version = "0.1.59"
+version = "0.1.61"
 edition = "2021"
 
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
-llama-cpp-2 = { path = "../llama-cpp-2", version = "0.1.59" }
+llama-cpp-2 = { path = "../llama-cpp-2", version = "0.1.61" }
 hf-hub = { workspace = true }
 clap = { workspace = true , features = ["derive"] }
 anyhow = { workspace = true }

embeddings/Cargo.toml

@@ -1,7 +1,7 @@
 [package]
 name = "llama-cpp-2"
 description = "llama.cpp bindings for Rust"
-version = "0.1.59"
+version = "0.1.61"
 edition = "2021"
 license = "MIT OR Apache-2.0"
 repository = "https://github.com/utilityai/llama-cpp-rs"
@@ -10,7 +10,7 @@ repository = "https://github.com/utilityai/llama-cpp-rs"
 
 [dependencies]
 enumflags2 = "0.7.10"
-llama-cpp-sys-2 = { path = "../llama-cpp-sys-2", version = "0.1.59" }
+llama-cpp-sys-2 = { path = "../llama-cpp-sys-2", version = "0.1.61" }
 thiserror = { workspace = true }
 tracing = { workspace = true }
 

llama-cpp-2/Cargo.toml

@@ -339,6 +339,46 @@ impl LlamaContextParams {
         self.context_params.embeddings = embedding;
         self
     }
+
+    /// Set the evaluation callback.
+    ///
+    /// # Examples
+    ///
+    /// ```no_run
+    /// extern "C" fn cb_eval_fn(
+    ///     t: *mut llama_cpp_sys_2::ggml_tensor,
+    ///     ask: bool,
+    ///     user_data: *mut std::ffi::c_void,
+    /// ) -> bool {
+    ///     false
+    /// }
+    ///
+    /// use llama_cpp_2::context::params::LlamaContextParams;
+    /// let params = LlamaContextParams::default();
+    /// params.with_cb_eval(Some(cb_eval_fn));
+    /// ```
+    pub fn with_cb_eval(
+        mut self,
+        cb_eval: llama_cpp_sys_2::ggml_backend_sched_eval_callback,
+    ) -> Self {
+        self.context_params.cb_eval = cb_eval;
+        self
+    }
+
+    /// Set the evaluation callback user data.
+    ///
+    /// # Examples
+    ///
+    /// ```no_run
+    /// use llama_cpp_2::context::params::LlamaContextParams;
+    /// let params = LlamaContextParams::default();
+    /// let user_data = std::ptr::null_mut();
+    /// params.with_cb_eval_user_data(user_data);
+    /// ```
+    pub fn with_cb_eval_user_data(mut self, cb_eval_user_data: *mut std::ffi::c_void) -> Self {
+        self.context_params.cb_eval_user_data = cb_eval_user_data;
+        self
+    }
 }
 
 /// Default parameters for `LlamaContext`. (as defined in llama.cpp by `llama_context_default_params`)

llama-cpp-2/src/context/params.rs

@@ -294,8 +294,8 @@ impl ParseState {
                         type_: gre_type,
                         value: c as _,
                     });
-                    if rest.starts_with("-]") {
-                        let (c, r) = Self::parse_char(rest)?;
+                    if rest.starts_with("-") && rest.get(1..).is_some_and(|r| !r.starts_with("]")) {
+                        let (c, r) = Self::parse_char(&rest[1..])?;
                         rest = r;
                         rule.push(llama_grammar_element {
                             type_: llama_cpp_sys_2::LLAMA_GRETYPE_CHAR_RNG_UPPER,

llama-cpp-2/src/grammar.rs

@@ -55,3 +55,36 @@ fn check_parse_simple() {
         parse_state
     );
 }
+
+#[test]
+fn check_parse_char_range() {
+    let parse_state = ParseState::from_str(r#"root ::= [a-zA-Z]"#).unwrap();
+    assert_eq!(
+        ParseState {
+            symbol_ids: BTreeMap::from([("root".to_string(), 0),]),
+            rules: vec![vec![
+                llama_grammar_element {
+                    type_: llama_cpp_sys_2::LLAMA_GRETYPE_CHAR,
+                    value: 'a' as u32
+                },
+                llama_grammar_element {
+                    type_: llama_cpp_sys_2::LLAMA_GRETYPE_CHAR_RNG_UPPER,
+                    value: 'z' as u32
+                },
+                llama_grammar_element {
+                    type_: llama_cpp_sys_2::LLAMA_GRETYPE_CHAR_ALT,
+                    value: 'A' as u32
+                },
+                llama_grammar_element {
+                    type_: llama_cpp_sys_2::LLAMA_GRETYPE_CHAR_RNG_UPPER,
+                    value: 'Z' as u32
+                },
+                llama_grammar_element {
+                    type_: llama_cpp_sys_2::LLAMA_GRETYPE_END,
+                    value: 0
+                }
+            ]]
+        },
+        parse_state
+    );
+}

llama-cpp-2/src/grammar/tests.rs

@@ -1,7 +1,7 @@
 [package]
 name = "llama-cpp-sys-2"
 description = "Low Level Bindings to llama.cpp"
-version = "0.1.59"
+version = "0.1.61"
 edition = "2021"
 license = "MIT OR Apache-2.0"
 repository = "https://github.com/utilityai/llama-cpp-rs"

llama-cpp-sys-2/Cargo.toml

@@ -1,12 +1,12 @@
 [package]
 name = "simple"
-version = "0.1.59"
+version = "0.1.61"
 edition = "2021"
 
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
-llama-cpp-2 = { path = "../llama-cpp-2", version = "0.1.59" }
+llama-cpp-2 = { path = "../llama-cpp-2", version = "0.1.61" }
 hf-hub = { workspace = true }
 clap = { workspace = true , features = ["derive"] }
 anyhow = { workspace = true }