mirror of
https://github.com/ollama/ollama.git
synced 2026-01-12 00:06:57 +08:00
33ee7168ba
* WIP - MLX backend with gemma3 * MLX: add cmake and go tag build toggles To build the new MLX backend code: cmake --preset MLX cmake --build --preset MLX --parallel cmake --install build --component MLX go build -tags mlx . Note: the main.go entrypoint for the MLX engine will change in a follow up commit. * add experimental image generation runtime * add experimental image generation runtime * MLX: wire up cuda build for linux * MLX: get dependencies correct and dedup This is still too large for a unified github artifact, but is now "correct" for the mlx_cuda_v13 directory. * fix relative link bug in dedup * Add darwin build and readme * add go build tag for mlx dependent code and wire up build_darwin.sh * lint cleanup * macos: build mlx for x86 This will be CPU only. * cuda build instructions and fix drift from mlx bump * stale comment * Delete agent helper doc * Clean up readme.md * Revise README for tokenizer clarity and details Updated README to clarify tokenizer functionality and removed correctness section. --------- Co-authored-by: jmorganca <jmorganca@gmail.com>
50 lines
1.7 KiB
Go
50 lines
1.7 KiB
Go
package convert
|
|
|
|
type gemma2Model struct {
|
|
gemmaModel
|
|
SlidingWindow uint32 `json:"sliding_window"`
|
|
AttentionLogitSoftcap float32 `json:"attn_logit_softcapping"`
|
|
FinalLogitSoftcap float32 `json:"final_logit_softcapping"`
|
|
}
|
|
|
|
func (p *gemma2Model) KV(t *Tokenizer) KV {
|
|
kv := p.ModelParameters.KV(t)
|
|
kv["general.architecture"] = "gemma2"
|
|
kv["gemma2.context_length"] = p.MaxPositionEmbeddings
|
|
kv["gemma2.embedding_length"] = p.HiddenSize
|
|
kv["gemma2.block_count"] = p.HiddenLayers
|
|
kv["gemma2.feed_forward_length"] = p.IntermediateSize
|
|
kv["gemma2.attention.head_count"] = p.NumAttentionHeads
|
|
kv["gemma2.attention.head_count_kv"] = p.NumKeyValueHeads
|
|
kv["gemma2.attention.layer_norm_rms_epsilon"] = p.RMSNormEPS
|
|
kv["gemma2.attention.key_length"] = p.HeadDim
|
|
kv["gemma2.attention.value_length"] = p.HeadDim
|
|
kv["gemma2.attention.sliding_window"] = p.SlidingWindow
|
|
kv["gemma2.attn_logit_softcapping"] = p.AttentionLogitSoftcap
|
|
kv["gemma2.final_logit_softcapping"] = p.FinalLogitSoftcap
|
|
kv["tokenizer.ggml.eot_token_id"] = uint32(107)
|
|
kv["tokenizer.ggml.middle_token_id"] = uint32(68)
|
|
kv["tokenizer.ggml.prefix_token_id"] = uint32(67)
|
|
kv["tokenizer.ggml.suffix_token_id"] = uint32(69)
|
|
return kv
|
|
}
|
|
|
|
func (p *gemma2Model) Replacements() []string {
|
|
return []string{
|
|
"model.embed_tokens", "token_embd",
|
|
"model.norm", "output_norm",
|
|
"model.layers", "blk",
|
|
"input_layernorm", "attn_norm",
|
|
"self_attn.q_proj", "attn_q",
|
|
"self_attn.k_proj", "attn_k",
|
|
"self_attn.v_proj", "attn_v",
|
|
"self_attn.o_proj", "attn_output",
|
|
"mlp.gate_proj", "ffn_gate",
|
|
"mlp.down_proj", "ffn_down",
|
|
"mlp.up_proj", "ffn_up",
|
|
"post_attention_layernorm", "post_attention_norm",
|
|
"pre_feedforward_layernorm", "ffn_norm",
|
|
"post_feedforward_layernorm", "post_ffw_norm",
|
|
}
|
|
}
|