Commit b2b270ad authored by Devon Rifkin's avatar Devon Rifkin
Browse files

Merge branch 'main' into drifkin/array-head-count-simple

parents 20c5fd39 2bb69b40
...@@ -143,6 +143,25 @@ func TestGenerateChat(t *testing.T) { ...@@ -143,6 +143,25 @@ func TestGenerateChat(t *testing.T) {
} }
}) })
t.Run("missing thinking capability", func(t *testing.T) {
think := true
w := createRequest(t, s.ChatHandler, api.ChatRequest{
Model: "test",
Messages: []api.Message{
{Role: "user", Content: "Hello!"},
},
Think: &think,
})
if w.Code != http.StatusBadRequest {
t.Errorf("expected status 400, got %d", w.Code)
}
if diff := cmp.Diff(w.Body.String(), `{"error":"registry.ollama.ai/library/test:latest does not support thinking"}`); diff != "" {
t.Errorf("mismatch (-got +want):\n%s", diff)
}
})
t.Run("missing model", func(t *testing.T) { t.Run("missing model", func(t *testing.T) {
w := createRequest(t, s.ChatHandler, api.ChatRequest{}) w := createRequest(t, s.ChatHandler, api.ChatRequest{})
if w.Code != http.StatusBadRequest { if w.Code != http.StatusBadRequest {
......
...@@ -21,6 +21,8 @@ import ( ...@@ -21,6 +21,8 @@ import (
"testing" "testing"
"unicode" "unicode"
"github.com/gin-gonic/gin"
"github.com/google/go-cmp/cmp"
"github.com/ollama/ollama/api" "github.com/ollama/ollama/api"
"github.com/ollama/ollama/fs/ggml" "github.com/ollama/ollama/fs/ggml"
"github.com/ollama/ollama/openai" "github.com/ollama/ollama/openai"
...@@ -882,3 +884,87 @@ func TestFilterThinkTags(t *testing.T) { ...@@ -882,3 +884,87 @@ func TestFilterThinkTags(t *testing.T) {
} }
} }
} }
func TestWaitForStream(t *testing.T) {
gin.SetMode(gin.TestMode)
cases := []struct {
name string
messages []any
expectCode int
expectBody string
}{
{
name: "error",
messages: []any{
gin.H{"error": "internal server error"},
},
expectCode: http.StatusInternalServerError,
expectBody: `{"error":"internal server error"}`,
},
{
name: "error status",
messages: []any{
gin.H{"status": http.StatusNotFound, "error": "not found"},
},
expectCode: http.StatusNotFound,
expectBody: `{"error":"not found"}`,
},
{
name: "unknown error",
messages: []any{
gin.H{"msg": "something else"},
},
expectCode: http.StatusInternalServerError,
expectBody: `{"error":"unknown error"}`,
},
{
name: "unknown type",
messages: []any{
struct{}{},
},
expectCode: http.StatusInternalServerError,
expectBody: `{"error":"unknown message type"}`,
},
{
name: "progress success",
messages: []any{
api.ProgressResponse{Status: "success"},
},
expectCode: http.StatusOK,
expectBody: `{"status":"success"}`,
},
{
name: "progress more than success",
messages: []any{
api.ProgressResponse{Status: "success"},
api.ProgressResponse{Status: "one more thing"},
},
expectCode: http.StatusOK,
expectBody: `{"status":"one more thing"}`,
},
}
for _, tt := range cases {
t.Run(tt.name, func(t *testing.T) {
w := httptest.NewRecorder()
c, _ := gin.CreateTestContext(w)
ch := make(chan any, len(tt.messages))
for _, msg := range tt.messages {
ch <- msg
}
close(ch)
waitForStream(c, ch)
if w.Code != tt.expectCode {
t.Errorf("expected status %d, got %d", tt.expectCode, w.Code)
}
if diff := cmp.Diff(w.Body.String(), tt.expectBody); diff != "" {
t.Errorf("body mismatch (-want +got):\n%s", diff)
}
})
}
}
...@@ -8,6 +8,7 @@ import ( ...@@ -8,6 +8,7 @@ import (
"os" "os"
"reflect" "reflect"
"runtime" "runtime"
"slices"
"sort" "sort"
"strconv" "strconv"
"strings" "strings"
...@@ -132,11 +133,11 @@ func (s *Scheduler) processPending(ctx context.Context) { ...@@ -132,11 +133,11 @@ func (s *Scheduler) processPending(ctx context.Context) {
continue continue
} }
numParallel := int(envconfig.NumParallel()) numParallel := int(envconfig.NumParallel())
// TODO (jmorganca): mllama doesn't support parallel yet // `mllama` is a snowflake and uses an encoder cache which cannot be used with num_parallel > 1
// see https://github.com/ollama/ollama/issues/4165 // ref: https://github.com/ollama/ollama/issues/4165
if checkMllamaModelFamily(pending.model) && numParallel != 1 { if slices.Contains(pending.model.Config.ModelFamilies, "mllama") && numParallel != 1 {
numParallel = 1 numParallel = 1
slog.Warn("mllama doesn't support parallel requests yet") slog.Warn("mllama does not currently support parallel requests")
} }
for { for {
...@@ -386,6 +387,17 @@ func (s *Scheduler) processCompleted(ctx context.Context) { ...@@ -386,6 +387,17 @@ func (s *Scheduler) processCompleted(ctx context.Context) {
s.loadedMu.Unlock() s.loadedMu.Unlock()
runner.refMu.Unlock() runner.refMu.Unlock()
slog.Debug("duplicate expired event, ignoring", "runner", runner) slog.Debug("duplicate expired event, ignoring", "runner", runner)
} else if runner.pid != runnerToUnload.pid {
// If the pids do not match, we likely had multiple load
// failures for the same model in quick succession due to
// request context canceled and are draining the queue of
// events. Ensure the orphaned runner is properly shut down, but
// do not delete the mismatched loaded runner, or wait for VRAM
// convergence.
slog.Debug("orphaned runner shutting down", "orphan", runner, "loaded", runnerToUnload)
runner.unload()
s.loadedMu.Unlock()
runner.refMu.Unlock()
} else { } else {
slog.Debug("starting background wait for VRAM recovery", "runner", runner) slog.Debug("starting background wait for VRAM recovery", "runner", runner)
finished := runner.waitForVRAMRecovery() finished := runner.waitForVRAMRecovery()
......
...@@ -112,11 +112,7 @@ func newScenarioRequest(t *testing.T, ctx context.Context, modelName string, est ...@@ -112,11 +112,7 @@ func newScenarioRequest(t *testing.T, ctx context.Context, modelName string, est
b.ctx, b.ctxDone = context.WithCancel(ctx) b.ctx, b.ctxDone = context.WithCancel(ctx)
t.Helper() t.Helper()
f, err := os.CreateTemp(t.TempDir(), modelName) p, _ := createBinFile(t, ggml.KV{
require.NoError(t, err)
defer f.Close()
require.NoError(t, ggml.WriteGGUF(f, ggml.KV{
"general.architecture": "llama", "general.architecture": "llama",
"llama.context_length": uint32(32), "llama.context_length": uint32(32),
"llama.embedding_length": uint32(4096), "llama.embedding_length": uint32(4096),
...@@ -129,14 +125,14 @@ func newScenarioRequest(t *testing.T, ctx context.Context, modelName string, est ...@@ -129,14 +125,14 @@ func newScenarioRequest(t *testing.T, ctx context.Context, modelName string, est
}, []*ggml.Tensor{ }, []*ggml.Tensor{
{Name: "blk.0.attn.weight", Kind: uint32(0), Offset: uint64(0), Shape: []uint64{1, 1, 1, 1}, WriterTo: bytes.NewReader(make([]byte, 32))}, {Name: "blk.0.attn.weight", Kind: uint32(0), Offset: uint64(0), Shape: []uint64{1, 1, 1, 1}, WriterTo: bytes.NewReader(make([]byte, 32))},
{Name: "output.weight", Kind: uint32(0), Offset: uint64(0), Shape: []uint64{1, 1, 1, 1}, WriterTo: bytes.NewReader(make([]byte, 32))}, {Name: "output.weight", Kind: uint32(0), Offset: uint64(0), Shape: []uint64{1, 1, 1, 1}, WriterTo: bytes.NewReader(make([]byte, 32))},
})) })
require.NoError(t, err)
fname := f.Name()
model := &Model{Name: modelName, ModelPath: fname}
b.f, err = llm.LoadModel(model.ModelPath, 0)
require.NoError(t, err)
model := &Model{Name: modelName, ModelPath: p}
f, err := llm.LoadModel(model.ModelPath, 0)
if err != nil {
t.Fatal(err)
}
b.f = f
if duration == nil { if duration == nil {
duration = &api.Duration{Duration: 5 * time.Millisecond} duration = &api.Duration{Duration: 5 * time.Millisecond}
} }
......
{{- if or .Tools .System }}<|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|>
{{- if .Tools }}# Safety Preamble
The instructions in this section override those in the task description and style guide sections. Don't answer questions that are harmful or immoral.
# System Preamble
## Basic Rules
You are a powerful conversational AI trained by Cohere to help people. You are augmented by a number of tools, and your job is to use and consume the output of these tools to best help the user. You will see a conversation history between yourself and a user, ending with an utterance from the user. You will then see a specific instruction instructing you what kind of response to generate. When you answer the user's requests, you cite your sources in your answers, according to those instructions.
{{ if .System }}# User Preamble
{{ .System }}
{{- end }}
## Available Tools
Here is a list of tools that you have available to you:
{{- range .Tools }}
```python
def {{ .Function.Name }}(
{{- range $name, $property := .Function.Parameters.Properties }}{{ $name }}: {{ $property.Type }}, {{ end }}) -> List[Dict]:
"""{{ .Function.Description }}
{{- if .Function.Parameters.Properties }}
Args:
{{- range $name, $property := .Function.Parameters.Properties }}
{{ $name }} ({{ $property.Type }}): {{ $property.Description }}
{{- end }}
{{- end }}
"""
pass
```
{{- end }}
{{- else if .System }}{{ .System }}
{{- end }}<|END_OF_TURN_TOKEN|>
{{- end }}
{{- range .Messages }}
{{- if eq .Role "system" }}
{{- continue }}
{{- end }}<|START_OF_TURN_TOKEN|>
{{- if eq .Role "user" }}<|USER_TOKEN|>{{ .Content }}
{{- else if eq .Role "assistant" }}<|CHATBOT_TOKEN|>
{{- if .Content }}{{ .Content }}
{{- else if .ToolCalls }}
Action: ```json
[
{{- range .ToolCalls }}
{
"tool_name": "{{ .Function.Name }}",
"parameters": {{ .Function.Arguments }}
}
{{- end }}
]```
{{ continue }}
{{ end }}
{{- else if eq .Role "tool" }}<|SYSTEM_TOKEN|><results>
{{ .Content }}</results>
{{- end }}<|END_OF_TURN_TOKEN|>
{{- end }}
{{- if .Tools }}<|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|>Write 'Action:' followed by a json-formatted list of actions that you want to perform in order to produce a good response to the user's last input. You can use any of the supplied tools any number of times, but you should aim to execute the minimum number of necessary actions for the input. You should use the `directly-answer` tool if calling the other tools is unnecessary. The list of actions you want to call should be formatted as a list of json objects, for example:
```json
[
{
"tool_name": title of the tool in the specification,
"parameters": a dict of parameters to input into the tool as they are defined in the specs, or {} if it takes no parameters
}
]```
{{- end }}<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>
\ No newline at end of file
<|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|># Safety Preamble
The instructions in this section override those in the task description and style guide sections. Don't answer questions that are harmful or immoral.
# System Preamble
## Basic Rules
You are a powerful conversational AI trained by Cohere to help people. You are augmented by a number of tools, and your job is to use and consume the output of these tools to best help the user. You will see a conversation history between yourself and a user, ending with an utterance from the user. You will then see a specific instruction instructing you what kind of response to generate. When you answer the user's requests, you cite your sources in your answers, according to those instructions.
# User Preamble
You are a knowledgeable assistant. You can answer questions and perform tasks.
## Available Tools
Here is a list of tools that you have available to you:
```python
def get_current_weather(format: string, location: string, ) -> List[Dict]:
"""Get the current weather
Args:
format (string): The temperature unit to use. Infer this from the user's location.
location (string): The city and state, e.g. San Francisco, CA
"""
pass
```<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|USER_TOKEN|>What's the weather like today in Paris?<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>
Action: ```json
[
{
"tool_name": "get_current_weather",
"parameters": {"format":"celsius","location":"Paris, France"}
}
]```
<|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|><results>
22</results><|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>The current temperature in Paris, France is 22 degrees Celsius.<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|USER_TOKEN|>What's the weather like today in San Francisco and Toronto?<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|>Write 'Action:' followed by a json-formatted list of actions that you want to perform in order to produce a good response to the user's last input. You can use any of the supplied tools any number of times, but you should aim to execute the minimum number of necessary actions for the input. You should use the `directly-answer` tool if calling the other tools is unnecessary. The list of actions you want to call should be formatted as a list of json objects, for example:
```json
[
{
"tool_name": title of the tool in the specification,
"parameters": a dict of parameters to input into the tool as they are defined in the specs, or {} if it takes no parameters
}
]```<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>
\ No newline at end of file
{{- if or .System .Tools }}<|start_header_id|>system<|end_header_id|>
{{- if .System }}
{{ .System }}
{{- end }}
In addition to plain text responses, you can chose to call one or more of the provided functions.
Use the following rule to decide when to call a function:
* if the response can be generated from your internal knowledge (e.g., as in the case of queries like "What is the capital of Poland?"), do so
* if you need external information that can be obtained by calling one or more of the provided functions, generate a function calls
If you decide to call functions:
* prefix function calls with functools marker (no closing marker required)
* all function calls should be generated in a single JSON list formatted as functools[{"name": [function name], "arguments": [function arguments as JSON]}, ...]
* follow the provided JSON schema. Do not hallucinate arguments or values. Do to blindly copy values from the provided samples
* respect the argument type formatting. E.g., if the type if number and format is float, write value 7 as 7.0
* make sure you pick the right functions that match the user intent
Available functions as JSON spec:
{{- if .Tools }}
{{ .Tools }}
{{- end }}<|eot_id|>
{{- end }}
{{- range .Messages }}<|start_header_id|>
{{- if or (eq .Role "user") (eq .Role "assistant") (eq .Role "tool") }}{{ .Role }}
{{- end }}<|end_header_id|>
{{- if .Content }}{{ .Content }}
{{- else if .ToolCalls }} functools[
{{- range .ToolCalls }}{{ "{" }}"name": "{{ .Function.Name }}", "arguments": {{ .Function.Arguments }}{{ "}" }}
{{- end }}]
{{- end }}<|eot_id|>
{{- end }}<|start_header_id|>assistant<|end_header_id|>
\ No newline at end of file
<|start_header_id|>system<|end_header_id|>
You are a knowledgeable assistant. You can answer questions and perform tasks.
In addition to plain text responses, you can chose to call one or more of the provided functions.
Use the following rule to decide when to call a function:
* if the response can be generated from your internal knowledge (e.g., as in the case of queries like "What is the capital of Poland?"), do so
* if you need external information that can be obtained by calling one or more of the provided functions, generate a function calls
If you decide to call functions:
* prefix function calls with functools marker (no closing marker required)
* all function calls should be generated in a single JSON list formatted as functools[{"name": [function name], "arguments": [function arguments as JSON]}, ...]
* follow the provided JSON schema. Do not hallucinate arguments or values. Do to blindly copy values from the provided samples
* respect the argument type formatting. E.g., if the type if number and format is float, write value 7 as 7.0
* make sure you pick the right functions that match the user intent
Available functions as JSON spec:
[{"type":"function","function":{"name":"get_current_weather","description":"Get the current weather","parameters":{"type":"object","required":["location","format"],"properties":{"format":{"type":"string","description":"The temperature unit to use. Infer this from the user's location.","enum":["celsius","fahrenheit"]},"location":{"type":"string","description":"The city and state, e.g. San Francisco, CA"}}}}}]<|eot_id|><|start_header_id|><|end_header_id|>You are a knowledgeable assistant. You can answer questions and perform tasks.<|eot_id|><|start_header_id|>user<|end_header_id|>What's the weather like today in Paris?<|eot_id|><|start_header_id|>assistant<|end_header_id|> functools[{"name": "get_current_weather", "arguments": {"format":"celsius","location":"Paris, France"}}]<|eot_id|><|start_header_id|>tool<|end_header_id|>22<|eot_id|><|start_header_id|>assistant<|end_header_id|>The current temperature in Paris, France is 22 degrees Celsius.<|eot_id|><|start_header_id|>user<|end_header_id|>What's the weather like today in San Francisco and Toronto?<|eot_id|><|start_header_id|>assistant<|end_header_id|>
\ No newline at end of file
{{- if .Messages }}
{{- if or .System .Tools }}<|start_header_id|>system<|end_header_id|>
{{ .System }}
{{- if .Tools }} You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions. For each function call return a json object with function name and arguments within <tool_call></tool_call> XML tags as follows:
<tool_call>
{"name": <function-name>,"arguments": <args-dict>}
</tool_call>
Here are the available tools:
<tools>
{{- range .Tools }} {{ .Function }}
{{- end }} </tools>
{{- end }}
{{- end }}<|eot_id|>
{{- range .Messages }}
{{- if ne .Role "system" }}<|start_header_id|>{{ .Role }}<|end_header_id|>
{{ if eq .Role "user" }}{{ .Content }}
{{- else if eq .Role "assistant" }}
{{- if .Content }}{{ .Content }}
{{- else if .ToolCalls }}<tool_call>
{{ range .ToolCalls }}{"name": "{{ .Function.Name }}", "arguments": {{ .Function.Arguments }}}
{{- end }}
</tool_call>
{{- end }}
{{- else if eq .Role "tool" }}<tool_response>
{{ .Content }}
</tool_response>
{{- end }}<|eot_id|>
{{- end }}
{{- end }}<|start_header_id|>assistant<|end_header_id|>
{{ else }}
{{ if .System }}<|start_header_id|>system<|end_header_id|>
{{ .System }}<|eot_id|>{{ end }}{{ if .Prompt }}<|start_header_id|>user<|end_header_id|>
{{ .Prompt }}<|eot_id|>{{ end }}<|start_header_id|>assistant<|end_header_id|>
{{ end }}{{ .Response }}
{{- if .Response }}<|eot_id|>
{{- end }}
\ No newline at end of file
<|start_header_id|>system<|end_header_id|>
You are a knowledgeable assistant. You can answer questions and perform tasks. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions. For each function call return a json object with function name and arguments within <tool_call></tool_call> XML tags as follows:
<tool_call>
{"name": <function-name>,"arguments": <args-dict>}
</tool_call>
Here are the available tools:
<tools> {"name":"get_current_weather","description":"Get the current weather","parameters":{"type":"object","required":["location","format"],"properties":{"format":{"type":"string","description":"The temperature unit to use. Infer this from the user's location.","enum":["celsius","fahrenheit"]},"location":{"type":"string","description":"The city and state, e.g. San Francisco, CA"}}}} </tools><|eot_id|><|start_header_id|>user<|end_header_id|>
What's the weather like today in Paris?<|eot_id|><|start_header_id|>assistant<|end_header_id|>
<tool_call>
{"name": "get_current_weather", "arguments": {"format":"celsius","location":"Paris, France"}}
</tool_call><|eot_id|><|start_header_id|>tool<|end_header_id|>
<tool_response>
22
</tool_response><|eot_id|><|start_header_id|>assistant<|end_header_id|>
The current temperature in Paris, France is 22 degrees Celsius.<|eot_id|><|start_header_id|>user<|end_header_id|>
What's the weather like today in San Francisco and Toronto?<|eot_id|><|start_header_id|>assistant<|end_header_id|>
[
{
"role": "system",
"content": "You are a knowledgeable assistant. You can answer questions and perform tasks."
},
{
"role": "user",
"content": "What's the weather like today in Paris?"
},
{
"role": "assistant",
"tool_calls": [
{
"id": "89a1e453-0bce-4de3-a456-c54bed09c520",
"type": "function",
"function": {
"name": "get_current_weather",
"arguments": {
"location": "Paris, France",
"format": "celsius"
}
}
}
]
},
{
"role": "tool",
"tool_call_id": "89a1e453-0bce-4de3-a456-c54bed09c520",
"content": "22"
},
{
"role": "assistant",
"content": "The current temperature in Paris, France is 22 degrees Celsius."
},
{
"role": "user",
"content": "What's the weather like today in San Francisco and Toronto?"
}
]
{{- range $index, $_ := .Messages }}
{{- if eq .Role "user" }}
{{- if and (eq (len (slice $.Messages $index)) 1) $.Tools }}[AVAILABLE_TOOLS] {{ $.Tools }}[/AVAILABLE_TOOLS]
{{- end }}[INST] {{ if and (eq (len (slice $.Messages $index)) 1) $.System }}{{ $.System }}
{{ end }}{{ .Content }}[/INST]
{{- else if eq .Role "assistant" }}
{{- if .Content }} {{ .Content }}</s>
{{- else if .ToolCalls }}[TOOL_CALLS] [
{{- range .ToolCalls }}{"name": "{{ .Function.Name }}", "arguments": {{ .Function.Arguments }}}
{{- end }}]</s>
{{- end }}
{{- else if eq .Role "tool" }}[TOOL_RESULTS] {"content": {{ .Content }}}[/TOOL_RESULTS]
{{- end }}
{{- end }}
\ No newline at end of file
[INST] What's the weather like today in Paris?[/INST][TOOL_CALLS] [{"name": "get_current_weather", "arguments": {"format":"celsius","location":"Paris, France"}}]</s>[TOOL_RESULTS] {"content": 22}[/TOOL_RESULTS] The current temperature in Paris, France is 22 degrees Celsius.</s>[AVAILABLE_TOOLS] [{"type":"function","function":{"name":"get_current_weather","description":"Get the current weather","parameters":{"type":"object","required":["location","format"],"properties":{"format":{"type":"string","description":"The temperature unit to use. Infer this from the user's location.","enum":["celsius","fahrenheit"]},"location":{"type":"string","description":"The city and state, e.g. San Francisco, CA"}}}}}][/AVAILABLE_TOOLS][INST] You are a knowledgeable assistant. You can answer questions and perform tasks.
What's the weather like today in San Francisco and Toronto?[/INST]
\ No newline at end of file
{{- if (or .Tools .System) }}<extra_id_0>System
{{ if .System }}{{ .System }}
{{ end }}
{{- if .Tools }}
{{- range .Tools }}<tool> {{ . }} </tool>{{ end }}
{{ end }}
{{- end }}
{{- range $i, $m := .Messages }}
{{- $last := eq (len (slice $.Messages $i)) 1 -}}
{{- if eq .Role "user" }}<extra_id_1>User
{{ .Content }}
{{- if $last }}
<extra_id_1>Assistant
{{- end }}
{{ else if eq .Role "tool" }}<extra_id_1>Tool
{{ .Content }}
{{- if $last }}
<extra_id_1>Assistant
{{- end }}
{{ else if eq .Role "assistant" }}<extra_id_1>Assistant
{{- if .ToolCalls }}
{{ range .ToolCalls }}<toolcall> {"name": "{{ .Function.Name }}", "arguments": {{ .Function.Arguments }}} </toolcall> {{ end }}
{{ else }}
{{ .Content }}
{{- if not $last }}
{{ end }}
{{- end }}
{{- end }}
{{- end }}
\ No newline at end of file
<extra_id_0>System
You are a knowledgeable assistant. You can answer questions and perform tasks.
<tool> {"type":"function","function":{"name":"get_current_weather","description":"Get the current weather","parameters":{"type":"object","required":["location","format"],"properties":{"format":{"type":"string","description":"The temperature unit to use. Infer this from the user's location.","enum":["celsius","fahrenheit"]},"location":{"type":"string","description":"The city and state, e.g. San Francisco, CA"}}}}} </tool>
<extra_id_1>User
What's the weather like today in Paris?
<extra_id_1>Assistant
<toolcall> {"name": "get_current_weather", "arguments": {"format":"celsius","location":"Paris, France"}} </toolcall>
<extra_id_1>Tool
22
<extra_id_1>Assistant
The current temperature in Paris, France is 22 degrees Celsius.
<extra_id_1>User
What's the weather like today in San Francisco and Toronto?
<extra_id_1>Assistant
[
{
"type": "function",
"function": {
"name": "get_current_weather",
"description": "Get the current weather",
"parameters": {
"type": "object",
"properties": {
"location": {
"type": "string",
"description": "The city and state, e.g. San Francisco, CA"
},
"format": {
"type": "string",
"enum": [
"celsius",
"fahrenheit"
],
"description": "The temperature unit to use. Infer this from the user's location."
}
},
"required": [
"location",
"format"
]
}
}
}
]
{{- if .System }}{{ .System }}
{{ end }}
{{- range $i, $_ := .Messages }}
{{- if eq .Role "user" }}### Instruction:
{{- if and $.Tools (le (len (slice $.Messages $i)) 2) }}
[BEGIN OF TASK INSTRUCTION]
You are an expert in composing functions. You are given a question and a set of possible functions.
Based on the question, you will need to make one or more function/tool calls to achieve the purpose.
If none of the functions can be used, point it out and refuse to answer.
If the given question lacks the parameters required by the function, also point it out.
[END OF TASK INSTRUCTION]
[BEGIN OF AVAILABLE TOOLS]
{{ $.Tools }}
[END OF AVAILABLE TOOLS]
[BEGIN OF FORMAT INSTRUCTION]
The output MUST strictly adhere to the following JSON format, and NO other text MUST be included.
The example format is as follows. Please make sure the parameter type is correct. If no function call is needed, please make tool_calls an empty list '[]'.
```
{
"tool_calls": [
{"name": "func_name1", "arguments": {"argument1": "value1", "argument2": "value2"}},
... (more tool calls as required)
]
}
```
[END OF FORMAT INSTRUCTION]
[BEGIN OF QUERY]
{{ .Content }}
[END OF QUERY]
{{ else }}
{{ .Content }}
{{ end }}
{{- else if .ToolCalls }}### Response:
{"tool_calls": [{{ range .ToolCalls }}{"name": "{{ .Function.Name }}", "arguments": {{ .Function.Arguments }}}{{ end }}]}
<|EOT|>
{{ else if eq .Role "assistant" }}### Response:
{{ .Content }}
<|EOT|>
{{ end }}
{{- end }}### Response:
\ No newline at end of file
You are a knowledgeable assistant. You can answer questions and perform tasks.
### Instruction:
What's the weather like today in Paris?
### Response:
{"tool_calls": [{"name": "get_current_weather", "arguments": {"format":"celsius","location":"Paris, France"}}]}
<|EOT|>
### Response:
The current temperature in Paris, France is 22 degrees Celsius.
<|EOT|>
### Instruction:
[BEGIN OF TASK INSTRUCTION]
You are an expert in composing functions. You are given a question and a set of possible functions.
Based on the question, you will need to make one or more function/tool calls to achieve the purpose.
If none of the functions can be used, point it out and refuse to answer.
If the given question lacks the parameters required by the function, also point it out.
[END OF TASK INSTRUCTION]
[BEGIN OF AVAILABLE TOOLS]
[{"type":"function","function":{"name":"get_current_weather","description":"Get the current weather","parameters":{"type":"object","required":["location","format"],"properties":{"format":{"type":"string","description":"The temperature unit to use. Infer this from the user's location.","enum":["celsius","fahrenheit"]},"location":{"type":"string","description":"The city and state, e.g. San Francisco, CA"}}}}}]
[END OF AVAILABLE TOOLS]
[BEGIN OF FORMAT INSTRUCTION]
The output MUST strictly adhere to the following JSON format, and NO other text MUST be included.
The example format is as follows. Please make sure the parameter type is correct. If no function call is needed, please make tool_calls an empty list '[]'.
```
{
"tool_calls": [
{"name": "func_name1", "arguments": {"argument1": "value1", "argument2": "value2"}},
... (more tool calls as required)
]
}
```
[END OF FORMAT INSTRUCTION]
[BEGIN OF QUERY]
What's the weather like today in San Francisco and Toronto?
[END OF QUERY]
### Response:
\ No newline at end of file
...@@ -167,6 +167,10 @@ type Values struct { ...@@ -167,6 +167,10 @@ type Values struct {
api.Tools api.Tools
Prompt string Prompt string
Suffix string Suffix string
Think bool
// whether or not the user explicitly set the thinking flag (vs. it being
// implicitly false). Templates can't see whether `Think` is nil
IsThinkSet bool
// forceLegacy is a flag used to test compatibility with legacy templates // forceLegacy is a flag used to test compatibility with legacy templates
forceLegacy bool forceLegacy bool
...@@ -222,16 +226,20 @@ func (t *Template) Execute(w io.Writer, v Values) error { ...@@ -222,16 +226,20 @@ func (t *Template) Execute(w io.Writer, v Values) error {
system, messages := collate(v.Messages) system, messages := collate(v.Messages)
if v.Prompt != "" && v.Suffix != "" { if v.Prompt != "" && v.Suffix != "" {
return t.Template.Execute(w, map[string]any{ return t.Template.Execute(w, map[string]any{
"Prompt": v.Prompt, "Prompt": v.Prompt,
"Suffix": v.Suffix, "Suffix": v.Suffix,
"Response": "", "Response": "",
"Think": v.Think,
"IsThinkSet": v.IsThinkSet,
}) })
} else if !v.forceLegacy && slices.Contains(t.Vars(), "messages") { } else if !v.forceLegacy && slices.Contains(t.Vars(), "messages") {
return t.Template.Execute(w, map[string]any{ return t.Template.Execute(w, map[string]any{
"System": system, "System": system,
"Messages": messages, "Messages": messages,
"Tools": v.Tools, "Tools": v.Tools,
"Response": "", "Response": "",
"Think": v.Think,
"IsThinkSet": v.IsThinkSet,
}) })
} }
...@@ -241,9 +249,11 @@ func (t *Template) Execute(w io.Writer, v Values) error { ...@@ -241,9 +249,11 @@ func (t *Template) Execute(w io.Writer, v Values) error {
for _, m := range messages { for _, m := range messages {
execute := func() error { execute := func() error {
if err := t.Template.Execute(&b, map[string]any{ if err := t.Template.Execute(&b, map[string]any{
"System": system, "System": system,
"Prompt": prompt, "Prompt": prompt,
"Response": response, "Response": response,
"Think": v.Think,
"IsThinkSet": v.IsThinkSet,
}); err != nil { }); err != nil {
return err return err
} }
...@@ -286,9 +296,11 @@ func (t *Template) Execute(w io.Writer, v Values) error { ...@@ -286,9 +296,11 @@ func (t *Template) Execute(w io.Writer, v Values) error {
tree := parse.Tree{Root: nodes.(*parse.ListNode)} tree := parse.Tree{Root: nodes.(*parse.ListNode)}
if err := template.Must(template.New("").AddParseTree("", &tree)).Execute(&b, map[string]any{ if err := template.Must(template.New("").AddParseTree("", &tree)).Execute(&b, map[string]any{
"System": system, "System": system,
"Prompt": prompt, "Prompt": prompt,
"Response": response, "Response": response,
"Think": v.Think,
"IsThinkSet": v.IsThinkSet,
}); err != nil { }); err != nil {
return err return err
} }
......
package thinking
import (
"strings"
"unicode"
)
type thinkingState int
const (
// We're looking for the opening tag, but we haven't seen any non-whitespace
// characters yet
thinkingState_LookingForOpening thinkingState = iota
// We've seen the opening tag, but we haven't seen any non-whitespace
// characters yet (we want to eat any whitespace between the opening tag and
// the thinking content)
thinkingState_ThinkingStartedEatingWhitespace
// We've seen non-whitespace characters after the opening tag, but we haven't
// seen the closing tag yet
thinkingState_Thinking
// We've seen the closing tag, but we haven't seen any non-whitespace
// characters after the closing tag yet (we want to eat any whitespace between
// the closing tag and the content)
thinkingState_ThinkingDoneEatingWhitespace
// We've seen the closing tag and seen at least one non-whitespace character
// after it
thinkingState_ThinkingDone
)
func (s thinkingState) String() string {
switch s {
case thinkingState_LookingForOpening:
return "LookingForOpening"
case thinkingState_ThinkingStartedEatingWhitespace:
return "ThinkingStartedEatingWhitespace"
case thinkingState_Thinking:
return "Thinking"
case thinkingState_ThinkingDoneEatingWhitespace:
return "ThinkingDoneEatingWhitespace"
case thinkingState_ThinkingDone:
return "ThinkingDone"
default:
return "Unknown"
}
}
type Parser struct {
state thinkingState
OpeningTag string
ClosingTag string
acc strings.Builder
}
// AddContent returns the thinking content and the non-thinking content that
// should be immediately sent to the user. It will internally buffer if it needs
// to see more raw content to disambiguate
func (s *Parser) AddContent(content string) (string, string) {
s.acc.WriteString(content)
var thinkingSb, remainingSb strings.Builder
var thinking, remaining string
keepLooping := true
// we loop because we might pass through multiple parsing states in a single
// call to addContent, and we want to make sure callers don't have to wait for
// data that's already unambiguous
for keepLooping {
thinking, remaining, keepLooping = eat(s)
thinkingSb.WriteString(thinking)
remainingSb.WriteString(remaining)
}
return thinkingSb.String(), remainingSb.String()
}
// the additional bool return is true iff we should continue eating
func eat(s *Parser) (string, string, bool) {
switch s.state {
case thinkingState_LookingForOpening:
trimmed := strings.TrimLeftFunc(s.acc.String(), unicode.IsSpace)
if strings.HasPrefix(trimmed, s.OpeningTag) {
after := strings.Join(strings.Split(trimmed, s.OpeningTag)[1:], s.OpeningTag)
after = strings.TrimLeftFunc(after, unicode.IsSpace)
// after might contain more than just thinking tokens, so we continue
// parsing instead of returning it as thinking tokens here
s.acc.Reset()
s.acc.WriteString(after)
if after == "" {
s.state = thinkingState_ThinkingStartedEatingWhitespace
} else {
s.state = thinkingState_Thinking
}
return "", "", true
} else if strings.HasPrefix(s.OpeningTag, trimmed) {
// partial opening seen, so let's keep accumulating
return "", "", false
} else if trimmed == "" {
// saw whitespace only, so let's keep accumulating
return "", "", false
} else {
// didn't see an opening tag, but we have content, so thinking was skipped
s.state = thinkingState_ThinkingDone
// note that we use the original content, not the trimmed one because we
// don't want to eat any whitespace in the real content if there were no
// thinking tags
return "", s.acc.String(), false
}
case thinkingState_ThinkingStartedEatingWhitespace:
trimmed := strings.TrimLeftFunc(s.acc.String(), unicode.IsSpace)
s.acc.Reset()
if trimmed == "" {
return "", "", false
} else {
s.state = thinkingState_Thinking
s.acc.WriteString(trimmed)
return "", "", true
}
case thinkingState_Thinking:
acc := s.acc.String()
if strings.Contains(acc, s.ClosingTag) {
split := strings.Split(acc, s.ClosingTag)
thinking := split[0]
remaining := strings.Join(split[1:], s.ClosingTag)
remaining = strings.TrimLeftFunc(remaining, unicode.IsSpace)
s.acc.Reset()
if remaining == "" {
s.state = thinkingState_ThinkingDoneEatingWhitespace
} else {
s.state = thinkingState_ThinkingDone
}
return thinking, remaining, false
} else if overlapLen := overlap(acc, s.ClosingTag); overlapLen > 0 {
thinking := acc[:len(acc)-overlapLen]
remaining := acc[len(acc)-overlapLen:]
s.acc.Reset()
// keep track of the candidate closing tag. We have to buffer it until it
// becomes disambiguated
s.acc.WriteString(remaining)
return thinking, "", false
} else {
// purely just thinking tokens, so we can return them
s.acc.Reset()
return acc, "", false
}
case thinkingState_ThinkingDoneEatingWhitespace:
trimmed := strings.TrimLeftFunc(s.acc.String(), unicode.IsSpace)
s.acc.Reset()
// if we see non-whitespace, we're done eating the leading whitespace of the content
if trimmed != "" {
s.state = thinkingState_ThinkingDone
}
return "", trimmed, false
case thinkingState_ThinkingDone:
acc := s.acc.String()
s.acc.Reset()
return "", acc, false
default:
panic("unknown state")
}
}
// longest overlap between suffix of s and prefix of delim
func overlap(s, delim string) int {
max := min(len(delim), len(s))
for i := max; i > 0; i-- {
if strings.HasSuffix(s, delim[:i]) {
return i
}
}
return 0
}
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment