Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
OpenDAS
ollama
Commits
4b3507f0
Commit
4b3507f0
authored
Aug 08, 2023
by
Bruce MacDonald
Browse files
embeddings endpoint
Co-Authored-By:
Jeffrey Morgan
<
jmorganca@gmail.com
>
parent
5ebce03c
Changes
2
Show whitespace changes
Inline
Side-by-side
Showing
2 changed files
with
85 additions
and
31 deletions
+85
-31
api/types.go
api/types.go
+11
-0
server/routes.go
server/routes.go
+74
-31
No files found.
api/types.go
View file @
4b3507f0
...
@@ -42,6 +42,17 @@ type GenerateRequest struct {
...
@@ -42,6 +42,17 @@ type GenerateRequest struct {
Options
map
[
string
]
interface
{}
`json:"options"`
Options
map
[
string
]
interface
{}
`json:"options"`
}
}
type
EmbeddingRequest
struct
{
Model
string
`json:"model"`
Prompt
string
`json:"prompt"`
Options
map
[
string
]
interface
{}
`json:"options"`
}
type
EmbeddingResponse
struct
{
Embedding
[]
float64
`json:"embedding"`
}
type
CreateRequest
struct
{
type
CreateRequest
struct
{
Name
string
`json:"name"`
Name
string
`json:"name"`
Path
string
`json:"path"`
Path
string
`json:"path"`
...
...
server/routes.go
View file @
4b3507f0
...
@@ -38,35 +38,17 @@ var loaded struct {
...
@@ -38,35 +38,17 @@ var loaded struct {
options
api
.
Options
options
api
.
Options
}
}
func
GenerateHandler
(
c
*
gin
.
Context
)
{
// load a model into memory if it is not already loaded, it is up to the caller to lock loaded.mu before calling this function
loaded
.
mu
.
Lock
()
func
load
(
model
*
Model
,
reqOpts
map
[
string
]
interface
{},
sessionDuration
time
.
Duration
)
error
{
defer
loaded
.
mu
.
Unlock
()
checkpointStart
:=
time
.
Now
()
var
req
api
.
GenerateRequest
if
err
:=
c
.
ShouldBindJSON
(
&
req
);
err
!=
nil
{
c
.
JSON
(
http
.
StatusBadRequest
,
gin
.
H
{
"error"
:
err
.
Error
()})
return
}
model
,
err
:=
GetModel
(
req
.
Model
)
if
err
!=
nil
{
c
.
JSON
(
http
.
StatusBadRequest
,
gin
.
H
{
"error"
:
err
.
Error
()})
return
}
opts
:=
api
.
DefaultOptions
()
opts
:=
api
.
DefaultOptions
()
if
err
:=
opts
.
FromMap
(
model
.
Options
);
err
!=
nil
{
if
err
:=
opts
.
FromMap
(
model
.
Options
);
err
!=
nil
{
log
.
Printf
(
"could not load model options: %v"
,
err
)
log
.
Printf
(
"could not load model options: %v"
,
err
)
c
.
JSON
(
http
.
StatusInternalServerError
,
gin
.
H
{
"error"
:
err
.
Error
()})
return
err
return
}
}
if
err
:=
opts
.
FromMap
(
req
.
Opt
ion
s
);
err
!=
nil
{
if
err
:=
opts
.
FromMap
(
reqOpts
);
err
!=
nil
{
log
.
Printf
(
"could not merge model options: %v"
,
err
)
log
.
Printf
(
"could not merge model options: %v"
,
err
)
c
.
JSON
(
http
.
StatusInternalServerError
,
gin
.
H
{
"error"
:
err
.
Error
()})
return
err
return
}
}
if
model
.
Digest
!=
loaded
.
digest
||
!
reflect
.
DeepEqual
(
loaded
.
options
,
opts
)
{
if
model
.
Digest
!=
loaded
.
digest
||
!
reflect
.
DeepEqual
(
loaded
.
options
,
opts
)
{
...
@@ -83,21 +65,18 @@ func GenerateHandler(c *gin.Context) {
...
@@ -83,21 +65,18 @@ func GenerateHandler(c *gin.Context) {
llm
,
err
:=
llama
.
New
(
model
.
ModelPath
,
opts
)
llm
,
err
:=
llama
.
New
(
model
.
ModelPath
,
opts
)
if
err
!=
nil
{
if
err
!=
nil
{
c
.
JSON
(
http
.
StatusInternalServerError
,
gin
.
H
{
"error"
:
err
.
Error
()})
return
err
return
}
}
if
opts
.
NumKeep
<
0
{
if
opts
.
NumKeep
<
0
{
promptWithSystem
,
err
:=
model
.
Prompt
(
api
.
GenerateRequest
{},
""
)
promptWithSystem
,
err
:=
model
.
Prompt
(
api
.
GenerateRequest
{},
""
)
if
err
!=
nil
{
if
err
!=
nil
{
c
.
JSON
(
http
.
StatusInternalServerError
,
gin
.
H
{
"error"
:
err
.
Error
()})
return
err
return
}
}
promptNoSystem
,
err
:=
model
.
Prompt
(
api
.
GenerateRequest
{
Context
:
[]
int
{
0
}},
""
)
promptNoSystem
,
err
:=
model
.
Prompt
(
api
.
GenerateRequest
{
Context
:
[]
int
{
0
}},
""
)
if
err
!=
nil
{
if
err
!=
nil
{
c
.
JSON
(
http
.
StatusInternalServerError
,
gin
.
H
{
"error"
:
err
.
Error
()})
return
err
return
}
}
tokensWithSystem
:=
llm
.
Encode
(
promptWithSystem
)
tokensWithSystem
:=
llm
.
Encode
(
promptWithSystem
)
...
@@ -110,9 +89,8 @@ func GenerateHandler(c *gin.Context) {
...
@@ -110,9 +89,8 @@ func GenerateHandler(c *gin.Context) {
loaded
.
digest
=
model
.
Digest
loaded
.
digest
=
model
.
Digest
loaded
.
options
=
opts
loaded
.
options
=
opts
}
}
sessionDuration
:=
5
*
time
.
Minute
loaded
.
expireAt
=
time
.
Now
()
.
Add
(
sessionDuration
)
loaded
.
expireAt
=
time
.
Now
()
.
Add
(
sessionDuration
)
if
loaded
.
expireTimer
==
nil
{
if
loaded
.
expireTimer
==
nil
{
loaded
.
expireTimer
=
time
.
AfterFunc
(
sessionDuration
,
func
()
{
loaded
.
expireTimer
=
time
.
AfterFunc
(
sessionDuration
,
func
()
{
loaded
.
mu
.
Lock
()
loaded
.
mu
.
Lock
()
...
@@ -132,6 +110,32 @@ func GenerateHandler(c *gin.Context) {
...
@@ -132,6 +110,32 @@ func GenerateHandler(c *gin.Context) {
})
})
}
}
loaded
.
expireTimer
.
Reset
(
sessionDuration
)
loaded
.
expireTimer
.
Reset
(
sessionDuration
)
return
nil
}
func
GenerateHandler
(
c
*
gin
.
Context
)
{
loaded
.
mu
.
Lock
()
defer
loaded
.
mu
.
Unlock
()
checkpointStart
:=
time
.
Now
()
var
req
api
.
GenerateRequest
if
err
:=
c
.
ShouldBindJSON
(
&
req
);
err
!=
nil
{
c
.
JSON
(
http
.
StatusBadRequest
,
gin
.
H
{
"error"
:
err
.
Error
()})
return
}
model
,
err
:=
GetModel
(
req
.
Model
)
if
err
!=
nil
{
c
.
JSON
(
http
.
StatusBadRequest
,
gin
.
H
{
"error"
:
err
.
Error
()})
return
}
sessionDuration
:=
5
*
time
.
Minute
if
err
:=
load
(
model
,
req
.
Options
,
sessionDuration
);
err
!=
nil
{
c
.
JSON
(
http
.
StatusInternalServerError
,
gin
.
H
{
"error"
:
err
.
Error
()})
return
}
checkpointLoaded
:=
time
.
Now
()
checkpointLoaded
:=
time
.
Now
()
...
@@ -181,6 +185,44 @@ func GenerateHandler(c *gin.Context) {
...
@@ -181,6 +185,44 @@ func GenerateHandler(c *gin.Context) {
streamResponse
(
c
,
ch
)
streamResponse
(
c
,
ch
)
}
}
func
EmbeddingHandler
(
c
*
gin
.
Context
)
{
loaded
.
mu
.
Lock
()
defer
loaded
.
mu
.
Unlock
()
var
req
api
.
EmbeddingRequest
if
err
:=
c
.
ShouldBindJSON
(
&
req
);
err
!=
nil
{
c
.
JSON
(
http
.
StatusBadRequest
,
gin
.
H
{
"error"
:
err
.
Error
()})
return
}
model
,
err
:=
GetModel
(
req
.
Model
)
if
err
!=
nil
{
c
.
JSON
(
http
.
StatusBadRequest
,
gin
.
H
{
"error"
:
err
.
Error
()})
return
}
if
err
:=
load
(
model
,
req
.
Options
,
5
*
time
.
Minute
);
err
!=
nil
{
c
.
JSON
(
http
.
StatusBadRequest
,
gin
.
H
{
"error"
:
err
.
Error
()})
return
}
if
!
loaded
.
options
.
EmbeddingOnly
{
c
.
JSON
(
http
.
StatusBadRequest
,
gin
.
H
{
"error"
:
"embedding option must be set to true"
})
return
}
embedding
,
err
:=
loaded
.
llm
.
Embedding
(
req
.
Prompt
)
if
err
!=
nil
{
log
.
Printf
(
"embedding generation failed: %v"
,
err
)
c
.
JSON
(
http
.
StatusInternalServerError
,
gin
.
H
{
"error"
:
"failed to generate embedding"
})
return
}
resp
:=
api
.
EmbeddingResponse
{
Embedding
:
embedding
,
}
c
.
JSON
(
http
.
StatusOK
,
resp
)
}
func
PullModelHandler
(
c
*
gin
.
Context
)
{
func
PullModelHandler
(
c
*
gin
.
Context
)
{
var
req
api
.
PullRequest
var
req
api
.
PullRequest
if
err
:=
c
.
ShouldBindJSON
(
&
req
);
err
!=
nil
{
if
err
:=
c
.
ShouldBindJSON
(
&
req
);
err
!=
nil
{
...
@@ -381,6 +423,7 @@ func Serve(ln net.Listener, extraOrigins []string) error {
...
@@ -381,6 +423,7 @@ func Serve(ln net.Listener, extraOrigins []string) error {
r
.
POST
(
"/api/pull"
,
PullModelHandler
)
r
.
POST
(
"/api/pull"
,
PullModelHandler
)
r
.
POST
(
"/api/generate"
,
GenerateHandler
)
r
.
POST
(
"/api/generate"
,
GenerateHandler
)
r
.
POST
(
"/api/embeddings"
,
EmbeddingHandler
)
r
.
POST
(
"/api/create"
,
CreateModelHandler
)
r
.
POST
(
"/api/create"
,
CreateModelHandler
)
r
.
POST
(
"/api/push"
,
PushModelHandler
)
r
.
POST
(
"/api/push"
,
PushModelHandler
)
r
.
POST
(
"/api/copy"
,
CopyModelHandler
)
r
.
POST
(
"/api/copy"
,
CopyModelHandler
)
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment