mirror of
https://github.com/dogkeeper886/ollama37.git
synced 2025-12-12 00:37:04 +00:00
cbcbc9ae07
- Add Gemma3n model support with text generation capabilities - Add new CUDA mean operations for improved performance - Add macOS documentation and performance tests - Update LLAMA patches for ROCm/CUDA compatibility - Fix various model conversion and processing issues - Update CI workflows and build configurations - Add library model tests and Shakespeare test data 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>
52 lines
1.3 KiB
Go
52 lines
1.3 KiB
Go
package gemma3n
|
|
|
|
import (
|
|
"github.com/ollama/ollama/fs"
|
|
"github.com/ollama/ollama/kvcache"
|
|
"github.com/ollama/ollama/ml"
|
|
"github.com/ollama/ollama/model"
|
|
"github.com/ollama/ollama/model/input"
|
|
)
|
|
|
|
type Model struct {
|
|
model.Base
|
|
model.SentencePieceModel
|
|
|
|
*TextModel
|
|
}
|
|
|
|
// Forward implements model.Model.
|
|
func (m *Model) Forward(ctx ml.Context, batch input.Batch) (ml.Tensor, error) {
|
|
return m.TextModel.Forward(ctx, batch, m.Cache)
|
|
}
|
|
|
|
func New(c fs.Config) (model.Model, error) {
|
|
m := Model{
|
|
TextModel: newTextModel(c),
|
|
SentencePieceModel: model.NewSentencePieceModel(
|
|
&model.Vocabulary{
|
|
Values: c.Strings("tokenizer.ggml.tokens"),
|
|
Scores: c.Floats("tokenizer.ggml.scores"),
|
|
Types: c.Ints("tokenizer.ggml.token_type"),
|
|
AddBOS: c.Bool("tokenizer.ggml.add_bos_token", true),
|
|
BOS: []int32{int32(c.Uint("tokenizer.ggml.bos_token_id"))},
|
|
AddEOS: c.Bool("tokenizer.ggml.add_eos_token", false),
|
|
EOS: append(
|
|
[]int32{int32(c.Uint("tokenizer.ggml.eos_token_id"))},
|
|
c.Ints("tokenizer.ggml.eos_token_ids")...,
|
|
),
|
|
},
|
|
),
|
|
}
|
|
|
|
m.Cache = kvcache.NewWrapperCache(
|
|
kvcache.NewCausalCache(m.Shift),
|
|
kvcache.NewSWACache(int32(c.Uint("attention.sliding_window")), m.Shift),
|
|
)
|
|
return &m, nil
|
|
}
|
|
|
|
func init() {
|
|
model.Register("gemma3n", New)
|
|
}
|