Attempt at OpenElm #6986

joshcarp · 2024-04-29T18:28:03Z

Currently failing on line 821 of sgemm.cpp, still some parsing of ffn/attention head info needs to occur. Currently hard coded some stuff.

Fixes: #6868

Raising this PR as a draft because I need help. Will be adding comments to original source for reference purposes.

joshcarp · 2024-04-29T18:40:02Z

Okay I think it might have something to do with how i'm calculating the offesets from kqv

github-actions · 2024-04-29T19:44:20Z

📈 llama.cpp server for bench-server-baseline on Standard_NC4as_T4_v3 for phi-2-q4_0: 415 iterations 🚀

Expand details for performance related PR only

Concurrent users: 8, duration: 10m
HTTP request : avg=11355.74ms p(95)=31249.96ms fails=, finish reason: stop=361 truncated=54
Prompt processing (pp): avg=130.28tk/s p(95)=561.86tk/s
Token generation (tg): avg=27.12tk/s p(95)=34.82tk/s
ggml-org/models/phi-2/ggml-model-q4_0.gguf parallel=8 ctx-size=16384 ngl=33 batch-size=2048 ubatch-size=256 pp=1024 pp+tg=2048 branch=openelm-impl commit=8d2dead6819589281227694a6d75b8c1dc825936

More

---
config:
    xyChart:
        titleFontSize: 12
        width: 900
        height: 600
    themeVariables:
        xyChart:
            titleColor: "#000000"
---
xychart-beta
    title "llama.cpp bench-server-baseline on Standard_NC4as_T4_v3
 duration=10m 415 iterations"
    y-axis "llamacpp:prompt_tokens_seconds"
    x-axis "llamacpp:prompt_tokens_seconds" 1714489093 --> 1714489723
    line [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 344.19, 344.19, 344.19, 344.19, 344.19, 527.27, 527.27, 527.27, 527.27, 527.27, 551.35, 551.35, 551.35, 551.35, 551.35, 548.93, 548.93, 548.93, 548.93, 548.93, 559.08, 559.08, 559.08, 559.08, 559.08, 586.92, 586.92, 586.92, 586.92, 586.92, 617.48, 617.48, 617.48, 617.48, 617.48, 619.48, 619.48, 619.48, 619.48, 619.48, 613.81, 613.81, 613.81, 613.81, 613.81, 627.64, 627.64, 627.64, 627.64, 627.64, 643.58, 643.58, 643.58, 643.58, 643.58, 653.93, 653.93, 653.93, 653.93, 653.93, 653.9, 653.9, 653.9, 653.9, 653.9, 668.24, 668.24, 668.24, 668.24, 668.24, 665.63, 665.63, 665.63, 665.63, 665.63, 685.22, 685.22, 685.22, 685.22, 685.22, 687.47, 687.47, 687.47, 687.47, 687.47, 707.52, 707.52, 707.52, 707.52, 707.52, 680.47, 680.47, 680.47, 680.47, 680.47, 678.94, 678.94, 678.94, 678.94, 678.94, 682.54, 682.54, 682.54, 682.54, 682.54, 680.53, 680.53, 680.53, 680.53, 680.53, 691.88, 691.88, 691.88, 691.88, 691.88, 693.19, 693.19, 693.19, 693.19, 693.19, 693.61, 693.61, 693.61, 693.61, 693.61, 692.22, 692.22, 692.22, 692.22, 692.22, 691.96, 691.96, 691.96, 691.96, 691.96, 694.96, 694.96, 694.96, 694.96, 694.96, 697.38, 697.38, 697.38, 697.38, 697.38, 707.43, 707.43, 707.43, 707.43, 707.43, 697.6, 697.6, 697.6, 697.6, 697.6, 699.85, 699.85, 699.85, 699.85, 699.85, 699.77, 699.77, 699.77, 699.77, 699.77, 701.06, 701.06, 701.06, 701.06, 701.06, 703.41, 703.41, 703.41, 703.41, 703.41, 700.92, 700.92, 700.92, 700.92, 700.92, 699.29, 699.29, 699.29, 699.29, 699.29, 697.74, 697.74, 697.74, 697.74, 697.74, 697.98, 697.98, 697.98, 697.98, 697.98, 703.8, 703.8, 703.8, 703.8, 703.8, 704.52, 704.52, 704.52, 704.52, 704.52, 703.8, 703.8, 703.8, 703.8, 703.8, 706.52, 706.52, 706.52, 706.52, 706.52, 714.62, 714.62, 714.62, 714.62, 714.62, 720.04, 720.04, 720.04, 720.04, 720.04, 721.16, 721.16, 721.16, 721.16, 721.16, 725.44, 725.44, 725.44, 725.44, 725.44, 724.44, 724.44, 724.44, 724.44, 724.44, 723.61, 723.61, 723.61, 723.61, 723.61, 722.86, 722.86, 722.86, 722.86, 722.86, 724.41, 724.41, 724.41, 724.41, 724.41, 724.66, 724.66, 724.66, 724.66, 724.66, 712.67, 712.67, 712.67, 712.67, 712.67, 696.41, 696.41, 696.41, 696.41, 696.41, 691.32, 691.32, 691.32, 691.32, 691.32, 682.78, 682.78, 682.78, 682.78, 682.78, 679.57, 679.57, 679.57, 679.57, 679.57, 678.62, 678.62, 678.62, 678.62, 678.62, 678.16, 678.16, 678.16, 678.16, 678.16, 676.41, 676.41, 676.41, 676.41, 676.41, 679.58, 679.58, 679.58, 679.58, 679.58, 679.58]

More

---
config:
    xyChart:
        titleFontSize: 12
        width: 900
        height: 600
    themeVariables:
        xyChart:
            titleColor: "#000000"
---
xychart-beta
    title "llama.cpp bench-server-baseline on Standard_NC4as_T4_v3
 duration=10m 415 iterations"
    y-axis "llamacpp:predicted_tokens_seconds"
    x-axis "llamacpp:predicted_tokens_seconds" 1714489093 --> 1714489723
    line [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 36.2, 36.2, 36.2, 36.2, 36.2, 22.82, 22.82, 22.82, 22.82, 22.82, 20.79, 20.79, 20.79, 20.79, 20.79, 20.21, 20.21, 20.21, 20.21, 20.21, 19.23, 19.23, 19.23, 19.23, 19.23, 18.95, 18.95, 18.95, 18.95, 18.95, 20.03, 20.03, 20.03, 20.03, 20.03, 20.49, 20.49, 20.49, 20.49, 20.49, 21.82, 21.82, 21.82, 21.82, 21.82, 22.6, 22.6, 22.6, 22.6, 22.6, 22.88, 22.88, 22.88, 22.88, 22.88, 23.07, 23.07, 23.07, 23.07, 23.07, 23.07, 23.07, 23.07, 23.07, 23.07, 23.05, 23.05, 23.05, 23.05, 23.05, 22.77, 22.77, 22.77, 22.77, 22.77, 22.78, 22.78, 22.78, 22.78, 22.78, 22.36, 22.36, 22.36, 22.36, 22.36, 21.95, 21.95, 21.95, 21.95, 21.95, 21.92, 21.92, 21.92, 21.92, 21.92, 20.9, 20.9, 20.9, 20.9, 20.9, 20.98, 20.98, 20.98, 20.98, 20.98, 21.11, 21.11, 21.11, 21.11, 21.11, 21.2, 21.2, 21.2, 21.2, 21.2, 20.93, 20.93, 20.93, 20.93, 20.93, 20.89, 20.89, 20.89, 20.89, 20.89, 20.85, 20.85, 20.85, 20.85, 20.85, 20.67, 20.67, 20.67, 20.67, 20.67, 20.6, 20.6, 20.6, 20.6, 20.6, 20.52, 20.52, 20.52, 20.52, 20.52, 20.77, 20.77, 20.77, 20.77, 20.77, 20.8, 20.8, 20.8, 20.8, 20.8, 20.94, 20.94, 20.94, 20.94, 20.94, 21.1, 21.1, 21.1, 21.1, 21.1, 21.15, 21.15, 21.15, 21.15, 21.15, 21.18, 21.18, 21.18, 21.18, 21.18, 20.98, 20.98, 20.98, 20.98, 20.98, 20.9, 20.9, 20.9, 20.9, 20.9, 20.69, 20.69, 20.69, 20.69, 20.69, 20.65, 20.65, 20.65, 20.65, 20.65, 20.82, 20.82, 20.82, 20.82, 20.82, 20.94, 20.94, 20.94, 20.94, 20.94, 21.04, 21.04, 21.04, 21.04, 21.04, 21.19, 21.19, 21.19, 21.19, 21.19, 21.17, 21.17, 21.17, 21.17, 21.17, 21.14, 21.14, 21.14, 21.14, 21.14, 21.11, 21.11, 21.11, 21.11, 21.11, 21.02, 21.02, 21.02, 21.02, 21.02, 20.92, 20.92, 20.92, 20.92, 20.92, 20.95, 20.95, 20.95, 20.95, 20.95, 21.1, 21.1, 21.1, 21.1, 21.1, 21.24, 21.24, 21.24, 21.24, 21.24, 21.38, 21.38, 21.38, 21.38, 21.38, 21.48, 21.48, 21.48, 21.48, 21.48, 21.48, 21.48, 21.48, 21.48, 21.48, 21.12, 21.12, 21.12, 21.12, 21.12, 21.12, 21.12, 21.12, 21.12, 21.12, 21.02, 21.02, 21.02, 21.02, 21.02, 20.84, 20.84, 20.84, 20.84, 20.84, 20.79, 20.79, 20.79, 20.79, 20.79, 19.7, 19.7, 19.7, 19.7, 19.7, 19.65, 19.65, 19.65, 19.65, 19.65, 19.64]

Details

More

---
config:
    xyChart:
        titleFontSize: 12
        width: 900
        height: 600
    themeVariables:
        xyChart:
            titleColor: "#000000"
---
xychart-beta
    title "llama.cpp bench-server-baseline on Standard_NC4as_T4_v3
 duration=10m 415 iterations"
    y-axis "llamacpp:kv_cache_usage_ratio"
    x-axis "llamacpp:kv_cache_usage_ratio" 1714489093 --> 1714489723
    line [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.08, 0.08, 0.08, 0.08, 0.08, 0.23, 0.23, 0.23, 0.23, 0.23, 0.34, 0.34, 0.34, 0.34, 0.34, 0.25, 0.25, 0.25, 0.25, 0.25, 0.32, 0.32, 0.32, 0.32, 0.32, 0.16, 0.16, 0.16, 0.16, 0.16, 0.22, 0.22, 0.22, 0.22, 0.22, 0.13, 0.13, 0.13, 0.13, 0.13, 0.13, 0.13, 0.13, 0.13, 0.13, 0.13, 0.13, 0.13, 0.13, 0.13, 0.15, 0.15, 0.15, 0.15, 0.15, 0.22, 0.22, 0.22, 0.22, 0.22, 0.18, 0.18, 0.18, 0.18, 0.18, 0.3, 0.3, 0.3, 0.3, 0.3, 0.11, 0.11, 0.11, 0.11, 0.11, 0.28, 0.28, 0.28, 0.28, 0.28, 0.17, 0.17, 0.17, 0.17, 0.17, 0.32, 0.32, 0.32, 0.32, 0.32, 0.33, 0.33, 0.33, 0.33, 0.33, 0.15, 0.15, 0.15, 0.15, 0.15, 0.18, 0.18, 0.18, 0.18, 0.18, 0.12, 0.12, 0.12, 0.12, 0.12, 0.25, 0.25, 0.25, 0.25, 0.25, 0.27, 0.27, 0.27, 0.27, 0.27, 0.3, 0.3, 0.3, 0.3, 0.3, 0.3, 0.3, 0.3, 0.3, 0.3, 0.21, 0.21, 0.21, 0.21, 0.21, 0.15, 0.15, 0.15, 0.15, 0.15, 0.15, 0.15, 0.15, 0.15, 0.15, 0.22, 0.22, 0.22, 0.22, 0.22, 0.13, 0.13, 0.13, 0.13, 0.13, 0.11, 0.11, 0.11, 0.11, 0.11, 0.16, 0.16, 0.16, 0.16, 0.16, 0.14, 0.14, 0.14, 0.14, 0.14, 0.32, 0.32, 0.32, 0.32, 0.32, 0.37, 0.37, 0.37, 0.37, 0.37, 0.23, 0.23, 0.23, 0.23, 0.23, 0.16, 0.16, 0.16, 0.16, 0.16, 0.11, 0.11, 0.11, 0.11, 0.11, 0.11, 0.11, 0.11, 0.11, 0.11, 0.12, 0.12, 0.12, 0.12, 0.12, 0.15, 0.15, 0.15, 0.15, 0.15, 0.14, 0.14, 0.14, 0.14, 0.14, 0.14, 0.14, 0.14, 0.14, 0.14, 0.2, 0.2, 0.2, 0.2, 0.2, 0.17, 0.17, 0.17, 0.17, 0.17, 0.2, 0.2, 0.2, 0.2, 0.2, 0.13, 0.13, 0.13, 0.13, 0.13, 0.16, 0.16, 0.16, 0.16, 0.16, 0.08, 0.08, 0.08, 0.08, 0.08, 0.13, 0.13, 0.13, 0.13, 0.13, 0.11, 0.11, 0.11, 0.11, 0.11, 0.3, 0.3, 0.3, 0.3, 0.3, 0.43, 0.43, 0.43, 0.43, 0.43, 0.58, 0.58, 0.58, 0.58, 0.58, 0.6, 0.6, 0.6, 0.6, 0.6, 0.64, 0.64, 0.64, 0.64, 0.64, 0.56, 0.56, 0.56, 0.56, 0.56, 0.6, 0.6, 0.6, 0.6, 0.6, 0.31, 0.31, 0.31, 0.31, 0.31, 0.2, 0.2, 0.2, 0.2, 0.2, 0.25]

More

---
config:
    xyChart:
        titleFontSize: 12
        width: 900
        height: 600
    themeVariables:
        xyChart:
            titleColor: "#000000"
---
xychart-beta
    title "llama.cpp bench-server-baseline on Standard_NC4as_T4_v3
 duration=10m 415 iterations"
    y-axis "llamacpp:requests_processing"
    x-axis "llamacpp:requests_processing" 1714489093 --> 1714489723
    line [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 8.0, 8.0, 8.0, 8.0, 8.0, 4.0, 4.0, 4.0, 4.0, 4.0, 6.0, 6.0, 6.0, 6.0, 6.0, 8.0, 8.0, 8.0, 8.0, 8.0, 7.0, 7.0, 7.0, 7.0, 7.0, 6.0, 6.0, 6.0, 6.0, 6.0, 5.0, 5.0, 5.0, 5.0, 5.0, 8.0, 8.0, 8.0, 8.0, 8.0, 6.0, 6.0, 6.0, 6.0, 6.0, 3.0, 3.0, 3.0, 3.0, 3.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 6.0, 6.0, 6.0, 6.0, 6.0, 7.0, 7.0, 7.0, 7.0, 7.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 5.0, 5.0, 5.0, 5.0, 5.0, 8.0, 8.0, 8.0, 8.0, 8.0, 4.0, 4.0, 4.0, 4.0, 4.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 7.0, 7.0, 7.0, 7.0, 7.0, 8.0, 8.0, 8.0, 8.0, 8.0, 7.0, 7.0, 7.0, 7.0, 7.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 4.0, 4.0, 4.0, 4.0, 4.0, 5.0, 5.0, 5.0, 5.0, 5.0, 7.0, 7.0, 7.0, 7.0, 7.0, 5.0, 5.0, 5.0, 5.0, 5.0, 7.0, 7.0, 7.0, 7.0, 7.0, 8.0, 8.0, 8.0, 8.0, 8.0, 5.0, 5.0, 5.0, 5.0, 5.0, 8.0, 8.0, 8.0, 8.0, 8.0, 7.0, 7.0, 7.0, 7.0, 7.0, 3.0, 3.0, 3.0, 3.0, 3.0, 1.0, 1.0, 1.0, 1.0, 1.0, 6.0, 6.0, 6.0, 6.0, 6.0, 3.0, 3.0, 3.0, 3.0, 3.0, 0.0, 0.0, 0.0, 0.0, 0.0, 7.0, 7.0, 7.0, 7.0, 7.0, 8.0, 8.0, 8.0, 8.0, 8.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 6.0, 7.0, 7.0, 7.0, 7.0, 7.0, 8.0, 8.0, 8.0, 8.0, 8.0, 6.0, 6.0, 6.0, 6.0, 6.0, 3.0, 3.0, 3.0, 3.0, 3.0, 7.0, 7.0, 7.0, 7.0, 7.0, 6.0, 6.0, 6.0, 6.0, 6.0, 8.0, 8.0, 8.0, 8.0, 8.0, 6.0, 6.0, 6.0, 6.0, 6.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 1.0]

joshcarp · 2024-04-29T21:29:17Z

llama.cpp

+                // So because our original wo matrix wasn't 3x, the below function fails because there aren't enough elems in it.
+                // Got: [head_dim][n_tokens][n_head_v]
+                // Want: [n_embd_v_gqa(384)][n_tokens]
+                // I guess this means that i need to be able to able to repeat them


This is the problem, these things need to be repeated like they are in the python part on line 10806

if self.num_groups != 1: # GQA # [B, k_h, S, h] --> [B, q_h, S, h] // so, k=3 -> q=12 keys = keys.repeat_interleave(self.num_groups, dim=1) # [B, v_h, S, h] --> [B, q_h, S, h] // so, v=3 -> q=12

trying to do something similar to this

I think this is solved now

joshcarp · 2024-04-30T01:45:14Z

llama.cpp

+                LLM_NORM_RMS, cb, il);
+                cb(cur, "ffn_norm", il);
+
+                cur = llm_build_ffn(ctx0, cur,


Now i think i just need to implement LLM_FFN_SWIGLU which is slightly different to the existing activation functions i think:

joshcarp · 2024-04-30T03:23:19Z

Okay so this does the build purely because the tensor sizes all match up, which doesn't necesarily mean that it's the correct implementation shown by the fact that there's still a sigsev somewhere

joshcarp · 2024-04-30T21:10:44Z

Debugging this because some of the transformations are obvs wrong:

ggml_debug:                 inp_embd = (f32)   GET_ROWS(token_embd.weight{1280, 32000, 1, 1}, inp_tokens{3, 1, 1, 1}}) = {1280, 3, 1, 1}
                                     sum = 138204.000000
ggml_debug:                   norm-0 = (f32)   RMS_NORM(inp_embd{1280, 3, 1, 1}, }) = {1280, 3, 1, 1}
                                     sum = 138204.015625
ggml_debug:              attn_norm-0 = (f32)        MUL(norm-0{1280, 3, 1, 1}, blk.0.attn_norm.weight{1280, 1, 1, 1}}) = {1280, 3, 1, 1}
                                     sum = 138204.000000
ggml_debug:                   wqkv-0 = (f32)    MUL_MAT(blk.0.attn_qkv.weight{1280, 1152, 1, 1}, attn_norm-0{1280, 3, 1, 1}}) = {1152, 3, 1, 1}
                                     sum = 124380.968750
ggml_debug:            wqkv-0 (view) = (f32)       VIEW(wqkv-0{1152, 3, 1, 1}, }) = {64, 3, 12, 1}
                                     sum = 8722732.000000
ggml_debug:     wqkv-0 (view) (cont) = (f32)       CONT(wqkv-0 (view){64, 3, 12, 1}, }) = {64, 3, 12, 1}
                                     sum = 497453.906250
ggml_debug:                   norm-0 = (f32)   RMS_NORM(wqkv-0 (view) (cont){64, 3, 12, 1}, }) = {64, 3, 12, 1}
                                     sum = 497532.062500
ggml_debug:                   Qcur-0 = (f32)        MUL(norm-0{64, 3, 12, 1}, blk.0.attn_q_norm.weight{64, 1, 1, 1}}) = {64, 3, 12, 1}
                                     sum = 497535.562500
ggml_debug:        Qcur-0 (reshaped) = (f32)    RESHAPE(Qcur-0{64, 3, 12, 1}, }) = {64, 12, 3, 1}
                                     sum = 497535.593750
ggml_debug:                   Qcur-0 = (f32)       ROPE(Qcur-0 (reshaped){64, 12, 3, 1}, inp_pos{3, 1, 1, 1}}) = {64, 12, 3, 1}
                                     sum = 497535.593750
ggml_debug:                   Qcur-0 = (f32)      SCALE(Qcur-0{64, 12, 3, 1}, }) = {64, 12, 3, 1}
                                     sum = 497458.562500
ggml_debug:            wqkv-0 (view) = (f32)       VIEW(wqkv-0{1152, 3, 1, 1}, }) = {64, 3, 3, 1}
                                     sum = 1002128.312500
ggml_debug:     wqkv-0 (view) (cont) = (f32)       CONT(wqkv-0 (view){64, 3, 3, 1}, }) = {64, 3, 3, 1}
                                     sum = 62096.066406
ggml_debug:                   norm-0 = (f32)   RMS_NORM(wqkv-0 (view) (cont){64, 3, 3, 1}, }) = {64, 3, 3, 1}
                                     sum = 62041.996094
ggml_debug:                   ''Kcur-0 = (f32)        MUL(norm-0{64, 3, 3, 1}, blk.0.attn_k_norm.weight{64, 1, 1, 1}}) = {64, 3, 3, 1}
                                     sum = 62035.175781
ggml_debug:        Kcur-0 (reshaped) = (f32)    RESHAPE(Kcur-0{64, 3, 3, 1}, }) = {64, 3, 3, 1}
                                     sum = 62035.175781
ggml_debug:                  node_16 = (f32)       ROPE(Kcur-0 (reshaped){64, 3, 3, 1}, inp_pos{3, 1, 1, 1}}) = {64, 3, 3, 1}
                                     sum = 62035.175781
ggml_debug:                  node_17 = (f32)     REPEAT(node_16{64, 3, 3, 1}, }) = {64, 3, 12, 1}
                                     sum = 497318.656250
ggml_debug:            wqkv-0 (view) = (f32)       VIEW(wqkv-0{1152, 3, 1, 1}, }) = {64, 3, 3, 1}
                                     sum = 1002132.000000
ggml_debug:                   Vcur-0 = (f32)       CONT(wqkv-0 (view){64, 3, 3, 1}, }) = {64, 3, 3, 1}
                                     sum = 62100.000000
ggml_debug:                  node_20 = (f32)     REPEAT(Vcur-0{64, 3, 3, 1}, }) = {64, 3, 12, 1}
                                     sum = 497448.000000
ggml_debug:                      v-0 = (f16)       VIEW(cache_v_l0{655360, 1, 1, 1}, }) = {32, 64, 6, 1}
                                     sum = nan

uhlhosting · 2024-05-02T18:22:57Z

How is progress on this?

hunt-47 · 2024-05-03T17:44:21Z

mahn, we all are waiting for you...!!
@joshcarp nice work by the way..

joshcarp · 2024-05-03T17:55:15Z

Still slaving away at debugging,
Like i said before, first time in this codebase so things are a bit wack.
All of the parameter and tensor parsing/shaping i'm pretty certain is correct, i'm comparing against the reference implementation through corenet+huggingface

joshcarp · 2024-05-07T19:04:43Z

Closing because i'm not making any meaningful progress on this

Attempt at OpenElm

6e89d82

joshcarp mentioned this pull request Apr 29, 2024

Support for OpenELM of Apple #6868

Closed

4 tasks

Add nop feedforward length

5ba2143

thinkverse mentioned this pull request Apr 29, 2024

Add OpenELM ollama/ollama#3910

Open

Add comment

7c3c3eb

joshcarp commented Apr 29, 2024

View reviewed changes

Checkpoint

0084a2a

joshcarp commented Apr 30, 2024

View reviewed changes

Fix SwiGlu2

9858fd1

fix up

5eea11e

lvoz2 mentioned this pull request Apr 30, 2024

Error: llama runner process no longer running: -1 ollama/ollama#3904

Closed

joshcarp added 2 commits April 30, 2024 08:51

Update

896dee5

Remove comment on assert that was failing

8d2dead

compilade mentioned this pull request May 3, 2024

convert.py: add python logging instead of print() #6511

Merged

joshcarp added 6 commits May 6, 2024 15:14

WIP

98ba54e

WIP

16b8ecd

Update

308c817

Cleanup

92ff0de

Cleanup

04e2858

Cleanup

716886c

joshcarp closed this May 7, 2024

icecream95 mentioned this pull request May 18, 2024

OpenELM support #7359

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Attempt at OpenElm #6986

Attempt at OpenElm #6986

joshcarp commented Apr 29, 2024

joshcarp commented Apr 29, 2024

github-actions bot commented Apr 29, 2024 •

edited

Loading

joshcarp Apr 29, 2024

joshcarp Apr 29, 2024

joshcarp Apr 29, 2024

joshcarp Apr 30, 2024

joshcarp Apr 30, 2024

joshcarp commented Apr 30, 2024

joshcarp commented Apr 30, 2024

uhlhosting commented May 2, 2024

hunt-47 commented May 3, 2024 •

edited

Loading

joshcarp commented May 3, 2024

joshcarp commented May 7, 2024

Attempt at OpenElm #6986

Attempt at OpenElm #6986

Conversation

joshcarp commented Apr 29, 2024

joshcarp commented Apr 29, 2024

github-actions bot commented Apr 29, 2024 • edited Loading

joshcarp Apr 29, 2024

Choose a reason for hiding this comment

joshcarp Apr 29, 2024

Choose a reason for hiding this comment

joshcarp Apr 29, 2024

Choose a reason for hiding this comment

joshcarp Apr 30, 2024

Choose a reason for hiding this comment

joshcarp Apr 30, 2024

Choose a reason for hiding this comment

joshcarp commented Apr 30, 2024

joshcarp commented Apr 30, 2024

uhlhosting commented May 2, 2024

hunt-47 commented May 3, 2024 • edited Loading

joshcarp commented May 3, 2024

joshcarp commented May 7, 2024

github-actions bot commented Apr 29, 2024 •

edited

Loading

hunt-47 commented May 3, 2024 •

edited

Loading